1 " WI“???

:
(
I
t

-,.
PI‘
4.114....

V . —.—‘ ~
fluid
‘53

”3%";
.1: ..‘

3;-
17.2?

..
l.
.... ,
d‘l -
»_. l1"
.
Bum."
:-
54'6“.

...
~33

~ ‘5

m-

.,..,,.
U?
c’
on
f

m w
‘ ~ —,. -1“
. I"

“2:; ‘

“an.

am
“a...
w“.

a...
3:

t;- «ra
":3.
..., ‘ ;_ u
«.mmfl‘ 3.. , "'
lm‘ ‘ f “
." "5
J
U
‘ .

’3'“ _
:3.»

E
(ii '
“I,

i

 

: ‘-
"w,
-‘T.’
:—

«8&3

lllllllllll lllll‘l l \lll‘lll

31293 I'

 

This is to certify that the

dissertation entitled

Ecological Contexts of Delinquency: Examination of
an Ecological Model} of Delinquency
Using Meta-analytical Techniques

presented by

Ishien Li

has been accepted towards fulfillment
of the requirements for

 

 

 

 

Ph. D. degree inFamily and Child Ecology
fillet/Leia XMZM
Date May 30, 1996 Dissertation Director

 

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

LIBRARY

Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove We checkout from your record.
TO AVOID FINES return on or More dete due.

DATE DUE menus DATE DUE
‘ 1 @2003 l

 

      

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU le An Nflrmetive Action/Emmi Opportunity lnetMon
mus-m

 

ECOLOGICAL CONTEXTS OF DELINQUENCY: EXAMINATION OF
AN ECOLOGICAL MODEL OF DELINQUENCY
USING META-ANALYTICAL TECHNIQUES
By

Ishien Li

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

DOCTOR OF PHILOSOPHY

Department of Family and Child Ecology

1996

ABSTRACT
ECOLOGICAL CONTEXTS OF DELINQUENCY:
EXAMINATION OF AN ECOLOGICAL MODEL OF DELINQUENCY
USING META-ANALYTICAL TECHNIQUES
By

Ishien Li

This study is designed (1) to develop an explanatory model of
adolescent delinquency, attending to a diversity of ecological contexts, by
means of theory integration, and (2) to investigate the extent to which the
existing research supports the relationships of the variables proposed in the
theoretical model of delinquency.

The synthesis of the study results follow Becker and Schram’s (1994)
model-driven meta-analysis approach. Zero-order correlations among the
predictors of delinquency and their relationships with delinquency are
gathered fiom 50 independent studies. Based on the correlations, seven
standard regressions are conducted to investigate the intervening mechanisms
that protect adolescents from delinquent acts, or that translate social-
environmental influences into individual delinquent behavior.

The ecological model of delinquency includes five exogenous

variables, three intermediating variables, and the outcome of three types of

self-report juvenile delinquency -- index ofienses, status offenses, and
substance abuse. The background variables are family attachment, family
control, parental conventionality, family socio-economic status (SES), and
community trouble. The intermediating variables are the adolescents’
educational commitment, self conventionality, and peers’ conventionality.

Different combinations of significant predictors are found for different
types of delinquent behaviors. All three intermediating variables and family
attachment are significant predictors of status offenses. However, the
significant predictors of index offenses are self conventionality and
educational commitment; of substance abuse, peers’ conventionality, self
conventionality, and educational commitment.

Family attachment, on average, is an insignificant to barely significant
predictor of different types of delinquent behavior. However, it has an
indirect effect on delinquent behavior through adolescents’ educational
commitment, self conventionality, and peers’ conventionality. Family SES,
parental conventionality, community trouble, and family control are also

found to have indirect effects on delinquency.

ACKNOWLEDGMENTS

This dissertation was completed with the help of many individuals. I
would like to take this opportunity to express my appreciation to those
individuals who have offered help, support, fiiendship, and encouragement
over the course of the development of this dissertation.

I wish to thank the members of my dissertation committee. Many
thanks to Dr. Lawrence Schiamberg, my dissertation director, for his
unconditional support and guidance. Deep appreciation goes to Dr. Verna
Hildebrand, my major professor, for her continuous advice, support, and
monitoring. Special thanks goes to Dr. Rosemary Walker, for her
inspirational encouragement. Sincere appreciation is extended to Dr. Betsy
Becker, for her enthusiasm and professionalism in conducting this meta-
analysis.

Special gratitude is extended to sponsors of the department’s Verna
Lee and John Hildebrand Dissertation Fellowship. I wish to express my
appreciation of the award and my respects to Drs. Verna Hildebrand and John

Hildebrand.

There are several others who have been instrumental. I am thankful to
colleagues in the Long Term Training program, especially to the director Dr.
Esther Onaga, for their social and emotional support. I am also very grateful
to Kyle Fahrbach and Debra Davis for giving so freely of their time to assist
me with editing and preparation of the dissertation manuscript. I consider
myself most fortunate to have had their assistance.

I also want to thank all my fiiends, especially my Taiwanese fiiends.
They have made my stay at Michigan State a special experience, full of good
times and memorable events. Special thanks goes to Meng-Li, Su-Hao, Mei-
Chin, and Ching—Ju, from them I gathered much valuable information and
experiences in life.

Finally, I particularly want to thank my family: my parents, my
brothers, my husband and my children for the variety of support they have
given me. They are behind me at all times -- especially the bad ones. Very
special thanks goes to my parents for everything. Without them I would not

be who I am now.

TABLE OF CONTENTS

List of Tables ............................................................................................. ix
List of Figures ............................................................................................ x
Chapter 1. INTRODUCTION ................................................................... 1
Purpose of the Study ...................................................................... 4
Theoretical Approach ..................................................................... 6
Organization of the Study ................................................................ 11
Chapter 2. REVIEW OF RESEARCH ....................................................... 14
Juvenile Delinquency ....................................................................... 14
Definition .............................................................................. 14
Measurement ......................................................................... 15

Social Control Theory ..................................................................... 16
Differential Association Theory ........................................................ 21
Delinquent Values and Behavior ....................................................... 22

The Ecological Contexts ................................................................... 25
Family .................................................................................... 26

Peers ...................................................................................... 28

School .................................................................................... 29

Community ............................................................................. 32
Family-Peer Mesosystem Relationships ................................. 33

Family, Community, and School ............................................ 35

Gender Difierences in Delinquency ................................................. 37
Prevalence and Incidence of Delinquency .............................. 39

Patterns of Changes ............................................................... 40

The Proposed Ecological Model of Delinquency ............................. 44
Chapter 3. Methodology ............................................................................ 49
Literature Search ............................................................................. 49
Inclusion and Exclusion Criteria ...................................................... 51
Coding ............................................................................................ 54

Unit of Analysis and Synthesis ........................................................ 55

Study Characteristics ....................................................................... 56

Sample Characteristics .................................................................... 57
Research Variables .......................................................................... 5 8
Missing Data ................................................................................... 61
Most and Least Frequently Studied Relationships ............................ 65
Meta-Analytical Procedures ............................................................. 65
Overview ............................................................................... 65
Homogeneity Tests ................................................................ 67
Estimation of the Linear Models ............................................ 68
Hypotheses ...................................................................................... 70
Chapter 4. RESULTS ................................................................................ 72
Homogeneity Test Results ............................................................... 72
Average Correlations ....................................................................... 75
Standardized Regression .................................................................. 78
Revised Structural Models ............................................................... 91
Gender Differences in Average Correlations .................................... 91
Chapter 5. CONCLUSION, DISCUSSION, AND IMPLICATIONS .......... 94
Discussion of Major Findings .......................................................... 96
Prediction of Educational Commitment ................................. 99
Prediction of Peers’ Conventionality ..................................... 101
Prediction of Self Conventionality ........................................ 102
Prediction of General Delinquency ....................................... 103
Prediction of Index Offenses ................................................. 104
Prediction of Status Offenses ................................................ 105
Prediction of Substance Abuse .............................................. 105
Indirect Effects of Family Attachment ................................... 106
Indirect Effects of Family SES .............................................. 107
Co-occurance of Delinquent Behaviors ................................. 109
Gender Differences ............................................................... 110
Strengths and Limitations of the Study ............................................ 111
Implications .................................................................................... 1 15
Implications for Practice ....................................................... 115
Recommendations ...................................................... 1 19

Implications for Future Research .......................................... 121
Recommendations ...................................................... 124

vii

APPENDICES

Appendix A: Studies Included in this Meta-Analysis ....................... 127
Appendix B: Coding Frame ............................................................. 131
BIBLIOGRAPHY .................................................................................... 132

viii

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

LIST OF TABLES

page
Year of report of the studies under synthesis ........................ 56

Number of studies by gender and sample size distribution 57
Number of subjects by gender and sample size distribution 57

Number of studies by number of correlation coefficients

provided individually in the studies ....................................... 63
Relationships investigated with multiple male- and female-

only samples by number of studies ....................................... 63
The most and least studied relationships ............................... 64

The relationships with insignificant homogeneity test
results .................................................................................. 74

The random-effects results: Correlations, Z-tests, and
number of sample correlations ............................................. 76

The random-effect results: Gender differences in average

correlations (for the relationships that allow for gender
comparisons only) ................................................................ 93

ix

Figure 1
Figure 2
Figure 3
Figure 4

Figure 5
Figure 6

Figure 7
Figure 8
Figure 9
Figure 10
Figure 11

Figure 12

LIST OF FIGURES

page
An Ecological Model of Delinquency ................................... 48
Prediction of Educational Commitment ................................ 83
Prediction of Self Conventionality ........................................ 84
Prediction of Peers’ Conventionality ..................................... 84
Prediction of General Delinquency ...................................... 85
Prediction of Index Offenses ................................................. 85
Prediction of Status Offenses ............................................... 86
Prediction of Substance Abuse ............................................. 86
Revised Ecological Model of Index Offenses ........................ 87
Revised Ecological Model of Status Offenses ....................... 88
Revised Ecological Model of Substance Abuse .................... 89
Revised Ecological Model of Delinquency ........................... 90

CHAPTER 1

INTRODUCTION

About 7 million adolescents in the US. today (i.e., one fourth of the
current population of 10- to 17-year-olds) have limited potential for becoming
productive adults because they are at high risk of delinquency, substance
abuse, school failure, and teen parenthood (Dryfoos, 1990). These problem
behaviors ofien co-occur; most at-risk adolescents, to some extent, "do them
all" (Dryfoos, 1990). Also, these adolescents tend to be intractable in
treatment and have continued problem behaviors as adults (Lerman & Pottick,
1995). The large number of adolescents with multiple problems has become a
major concern because societal costs, both monetary and in human capital are
high.

Researchers attempt to explain the causes of delinquent behavior using
factors from different ecological contexts, such as family, school, peer
groups, and the local community. Some attribute the major cause of youth
crime in the US. to the dysfunction of the family. Their argument is that all
factors influencing delinquency are likely to be a consequence of parental
behavior. Others believe that influences from deviant peers and community

criminal subcultures are more important than influences from family factors;

family life is important only when there are delinquent patterns to copy. Few
studies of delinquent etiology, however, have studied the differential contexts
of influences simultaneously (F arrington, 1993). As a field, delinquency
research today presents neither a coherent theoretical position nor consistent
empirical evidence regarding what factors lead to delinquent behavior and
how the factors relate to each other.

The confusion stems from several sources. First, investigators define
delinquent behavior differently, often making distinctions between different
forms of delinquent behavior. Some study index offenses (illegal behaviors
that can be prosecuted whether conducted by adults or minors) and status
ofl'enses (behaviors defined as criminal only when committed by minors,
including truancy, curfew violation, and running away) separately (e. g.,

J ohnstone, 197 8; Thornton, 1982). Others study different combinations of
delinquent behavior (e.g., Larzelere & Patterson, 1990; Lipton & Smith,
1983). As a consequence, delinquency investigators have not effectively
addressed the question of whether there are different explanations for
different types of delinquent behavior.

Secondly, studies using official versus self-report data tend to generate

different conclusions. When official data are used as a measure of delinquent

3

behavior, the individuals who engage in delinquent behavior but are
undetected by law enforcement are not included. Also, some researchers
consider the use of official data problematic due to administrative and
procedural errors and police bias in the arrest decision-making process
(Siegal & Senna, 1988). On the other hand, research using self-report data to
measure delinquency addresses delinquent behavior prior to oflicial actions;
the focus is shifted from legal-judicial responses to the behaviors themselves
(Tolan & Lorion, 1988).

Finally, a significant source of confusion is that studies tend to
consider only one or two possible ecological contexts at a time. The historical
trend of limiting research to one or two possible ecological contexts, in part,
was the result of both the competition between different schools of thought on
crime and delinquency and shortcomings and limitations in analytic design.
However, research restricted to one or two possible ecological contexts are
inadequate because the influences from other contexts need to be considered
in order to address alternative explanations as well as to rule out possible
confounding effects. As Wikstrom (1993) pointed out, a key issue in future
studies of delinquent behavior is such an integration; simple identification of

likely intervention targets or the relative merits of the targets in limited

ecological contexts is insufficient. More knowledge is needed about what the
relative influences of the factors from different contexts are, and also how
these influences are related to each other. In other words, we need to know
how negative social-environmental influences induce delinquent behavior in
individuals, as well as how the positive influences dissuade individuals from
delinquent behavior.

The above mentioned issues provide a focus for the design of the
proposed study as an examination of the relationships among the commonly
studied factors on delinquency. Each factor of delinquent behavior will be put
in a broad ecological context and the conceptualization of theoretical
positions which underlay these factors will be specified. And the intervening
processes for different forms of delinquent behavior (measured with self-
report, instead of official data) will be examined and compared. The essential
goal of this study will be to investigate how family, peers, school, and

community contexts influence adolescents' delinquent behavior.

Purpose of the Study
The purposes of this study are (l) to develop an explanatory model of

delinquent behavior, attending to a diversity of ecological contexts, by means

of theory integration, and (2) to investigate the extent to which the primary
studies support the relationships of the variables proposed in the theoretical
model of delinquency. The test of the model will also focus on whether there
are different systematic levels of explanations for different types of delinquent
behavior.

The synthesis of the study results follow Becker and Schram’s (1994)
model-driven meta-analysis approach. A model-driven meta-analysis is an
investigation of multiple relationships in a theoretical model, through a
synthesis of primary study results. A path model that represents the theory
can then be estimated from the synthesized results.

The model-driven meta-analysis approach suggested by Becker and
Schram (1994) is an option, besides the traditional primary study, well suited
to test the integrated theory. Theories or theoretical models help guide the
conduct of primary studies, as well as research synthesis. For both primary
studies and meta-analysis, simple identification of differences or relationships
is rarely the sole research objective; it is insufficient to examine only
differences or bivariate relationships. To accomplish the goals of this study,
one can “piece together the existing studies” by synthesizing their statistical

results (Becker, 1992). Even though none of the existing studies has

examined all the relationships of interest, the subsets of the studies examining
different relationships often overlap to some extent and thus can be
synthesized.

Besides focusing on the presumed contribution of an ecological
perspective to our understanding in the social-environmental causes of
delinquency, this study will also investigate gender differences. The necessity
of a gender-specific theory of delinquency has been repeatedly emphasized
by some feminist criminologists (e.g., Chesney-Lind, 1989; Chesney-Lind &
Shelden, 1992; F igueira-McDonough, 1983; Leonard, 1982). However, other
researchers believe a better approach would be to begin by testing the utility
of contemporary theories separately for the two genders (J ang, 1992). To
investigate this controversy, this study aims to answer the questions below.
How are gender differences considered in the delinquency literature? Do the
interrelations among the social-ecological factors, proposed in the ecological

model of delinquency, differ by gender?

Theoretical Approach

In the theory integration, the focus is on Hirschi's social control theory

and Sutherland's differential association theory, the two dominant social-

psychological theories of delinquent behavior (Krohn, 1986; Massey &
Krohn, 1986). Hirschi's (1969) social control theory states that delinquent
behaviors result when an individual's bonding to the society is weak or
broken. Sutherland's (Sutherland & Cressey, 1978) differential association
theory conceptualizes delinquent behavior as a result of socialization to pro-
delinquent values. Some researchers treat the two theories as irreconcilable
(Jensen, 1972; Matsueda, 1982; Matsueda & Heimer, 1987). However,
others view the two theories as complementary rather than competitive,
emphasizing the close interrelationship between the two theories in
facilitating theory integration ( Conger, 1976; Elliott etal., 1985; Johnson et
al., 1987; Marcos et al., 1986; Massey & Krohn, 1986).

As each theory is often found to be incomplete in explaining
delinquency by itself, it is frequently proposed that the two theories would
benefit through integration. For example, Hirschi's social control theory
emphasizes that attachment to conventional others decreases the likelihood of
delinquent behavior (Hirschi, 1969: 140-141), but the effect of attaching to
delinquent others is left unexplained. Sutherland's differential association
theory, on the other hand, argues that deviant behavior is learned from deviant

others, and that attachment to deviant others increases the likelihood of

deviant behavior (Sutherland & Cressey, 197 8). Researchers also find it
economic to integrate the two theories because the foci of the two are
overlapping to a great extent-differential association theory does not attend
to settings other than the family, community, and peers, while the social
control theory emphasizes the social bonding to family, peers, school, and
community.

The major criticism the integrationists face, however, is the
incompatibility in the basic assumptions of social control and differential
association theory. For example, criminal motivation is assumed to be a
constant for control theory, but as a variable for differential association
theory. Social control theory assumes that every person has sufficient
motivation to act delinquently and that differential vulnerability to deviance is
determined by variation in the strength of social controls. As a result, social
control theory focuses on the strength of the social bonds; an adolescent's
bonding to family, school, community are the principle interest of social
control theorists. In contrast, differential association theory does not make the
assumption of an innate motivation to delinquent behavior, but rather views

delinquent behavior as learned through pro-delinquent socialization.

Consequently, differential association theory focuses more on variations in
pro-delinquent socialization.

Facing the concern of the incompatibilities in the above-mentioned
assumptions of social control and differential association theory, several
integrationists propose different solutions. The approach this study takes is to
modify Hirschi' s assumption to be more similar to Sutherland's assumption
about hmnan nature. In this study, human nature is assumed to be neutral;
goodness is not inherent, nor is badness. Also, it is assumed that one’s values
and attitudes are subject to the influences from the outside world.

After modifying Hirschi’s theory's assumption, one can explore how
the concepts and propositions of the two theories fit together, because not all
concepts and propositions are tightly related to assumptions (Liska et al.,
1989). F 0110ng the assumption modification, a middle range integration of
the two theories will be conducted, that is, only some of the theoretical
concepts will be selected in the attempt to integrate the two theories.

Regarding the selection of the theoretical concepts, the scope of the
this study is ecological. The variables include (1) family attachment/bonding,
(2) family control/supervision, (3) family socio-economic status (SES), (4)

educational commitment, (5) parents', peers', and self's values toward

10

conventional activities and antisocial behavior, and (6) community trouble.
However, it is important to emphasize that this ecological approach is
somewhat different from Shaw and McKay's social disorganization theory
(1942), the so-called ecological theory in criminology, which emphasizes the
influences of community variables on aggregated measures of delinquency.

Shaw and McKay (1942) argued that certain community characteristics
may influence adolescents’ relationships with parents, adolescents’ beliefs
about right and wrong, and adolescents’ peer association. These factors, in
turn, are considered to influence involvement in delinquent behavior.

To the degree that a community is considered to be socially disorganized,
adolescents are expected to display attenuated social bonds, a greater level of
association with delinquent peers, and a greater level of delinquent behavior
themselves.

The position taken in Shaw and McKay's (1942) social disorganization
theory is that characteristics of the community are the primary influence of the
initiation of delinquent behavior, independent of individual and fanrily factors.
They referred to the process by which community characteristics affect
individual behavior as "social disorganization." However, they did not make

explicit their arguments linking the community level and individual level

11

elements. This problematic stance has plagued social disorganization research
since its development by Shaw and McKay. Another reason for the scarcity
of empirical work in the field of social disorganization is that research in this
tradition is generally very labor intensive and time consuming.

In this study, the human ecological framework is used to address the
relationships among the family, school, peers, and community factors in
relation to adolescent delinquency. A human ecological approach emphasizes
the interdependence between the individual, the family, and their
environments (Bronfennbrenner, 1977, 1986, 1989, 1993, 1994; Bubolz &
Sontag, 1993; Lerner, 1991). The environments, or the ecological contexts,
include culture, socio-econornic system, community, school, peers, farrrily,
and other institutions. The strength of the approach is its description and
explanation of phenomena in the wholeness of interaction and

interdependence.

Organization of the Study

In Chapter 2, the proposed ecological model of delinquency is
developed based on the literature review. By means of human ecological

concepts which focus on the interrelations of family, school, peers, and

12

community environments and their influences on adolescent delinquent
behavior, Hirschi’s social control theory and Sutherland’s differential
association theory are integrated.

In Chapter 3, the principles and procedures of the research synthesis
(i.e., meta-analysis) are described. The rationale for using multivariate and
model driven meta-analysis is also provided. To investigate certain
phenomena, primary studies often apply statistical procedures to data
collected from subjects; with meta-analysis, this researcher studies
phenomena through analyzing the statistical results of primary studies. In
addition to gathering quantitative information about the zero-order
correlations among the predictors of delinquency and their relations with
delinquency, this study aims to investigate the intervening mechanisms that
translate social-environmental influences into individual delinquent behavior,
or the mechanisms that protect adolescents from conducting delinquent acts.

In Chapter 4, the results of the meta-analysis are presented, including:
(1) study characteristics, (2) sample characteristics, (3) homogeneity test
results, (4) common or average correlations, and (5) standardized regressions.
In the first two sections, the characteristics of the sample studies used in this

meta-analysis and of the subjects that are used in these sample studies are

l3

summarized. Next, with the results of the homogeneity tests, a decision is
made as to whether the relationships between the variables seem to differ in
strength across the studies after chance variation is taken into account.
Finally, results of the standardized regressions based on the estimated
parameters are presented. Because the regression coefficients are
standardized, different predictors’ influences on different types of delinquent
behavior will be compared. In addition, the comparisons of study results
based on gender-specific samples will be made.

In Chapter 5, major findings and implications for practice as well as
future research are discussed. The need for certain actions and continued
research are suggested, based on this study’s results. Comparisons are made
between this integrated model of delinquency and other models that integrate
social control and difl'erential association theories. Strengths and limitations

of this study are also provided.

CHAPTER 2

REVIEW OF RESEARCH

In this chapter, research is reviewed in the following areas: (1) the
definitions and measurements of juvenile delinquency, (2) social control
theory, (3) differential association theory, (4) delinquent values and behavior,
and (5) selected ecological contexts of delinquency -- specifically, family,
school, peers, and community. Also, an ecological model of delinquency will

be proposed.

Juvenile Delinquency

Definition

Juvenile delinquency is typically defined as behavior committed by
minors that violates the penal codes of the government. Definitions of
“minor” vary among states, but normally the term refers to a person below 16
or 17 years of age. Generally, all delinquent behaviors share the common
characteristic of being antisocial acts, with the potential to inflict harm on
others or their property. In addition to the legal codes regulating both
juveniles and adults, juveniles are also arrested for drinking, truancy, curfew

violation, and runaway

l4

15

behavior. Such offenses are called status offenses, as they would not be

defined as criminal if committed by adults (Bartollas, 1993).

Measurement

In the measurement of juvenile delinquency, one continuing debate
concerns the relative merits of official versus self-report data. Various
criticisms have been raised about the use of official records of arrests and
convictions. Major concerns about official records are their susceptibility to
class and race biases in rates of arrest, and their partial representation of
committed crimes (Cohen, 1986; Siegal & Senna, 1988; Elliot et al., 1989;
Huizinga & Elliott, 1987). When official records are used as aEasure of
delinquent behavior, the individuals who engage in delinquent behavior but
are undetected by law enforcement are not included. Research using self-
report data to measure delinquency, on the other hand, studies delinquent
behavior prior to or separate from official actions; the focus is shifted from
legal-judicial responses to the behaviors themselves (Tolan & Lorion, 1988).

Self-report data are believed to represent a more complete picture of

delinquent behaviors. However, there are also some possible drawbacks in

using self-report data, for example, self-report data may over-represent minor

16

offenses. Also, self-report data may incorporate systematic errors of

measurement across samples of differential characteristics, such as race

(Matsueda, 1982).

Social Control Theory

Hirschi is the preeminent spokesman of modern social control theory.
He assumes that humans are inherently antisocial and that deviant behaviors
are natural (Hirschi, 1969). The question for Hirschi is "Why don't they do
it?" rather than "Why do they do it?" Specifically, why do most people stay
out of trouble? Hirschi argues that humans are dissuaded, protected, or
insulated from committing crime by strong bonding to the conventional
society. And the elements of the social bond are attachment, commitment,
involvement, and belief in the conventional moral order. However, there are
no universally accepted precise meanings for these four elements; Hirschi's
descriptions about the four elements are rather "sketchy" (Marcos et al.,
1986)

Attachment, the first element of the social bond, refers to the ties of
affection adolescents have to parents, teachers, and fiiends. The stronger the

attachment to others, the more likely an individual will internalize norms,

l7

develop a conscience, and consider that a delinquent act might threaten, or
break, their ties of afl’ection. Attachment to parents is considered the most
important variable insulating an adolescent against delinquent behavior.
Hirschi (1969) argues that if an adolescent is alienated from the parent he or
she will not develop an adequate conscience or superego; if a adolescent is
attached to the parent he or she will be less likely to deviate from the norms
of society.

Commitment to conventional activities and values is the second
element of the social bond. Commitment refers to an individual's aspirations
for and behavior consistent with conventional activities, such as educational
or career goals. When a committed individual considers the cost of
delinquent behavior, he or she uses common sense and thinks of the risk of
losing the investment already made in conventional activities (Hirschi, 1969).
Therefore, the more an individual is committed to academic achievement or
has higher occupational expectations, the less likely he is to become involved
in delinquent acts.

Involvement in conventional activities is another bonding element
which protects an individual from conducting delinquent behavior;

involvement in conventional activities leaves no time for delinquency.

18

Hirschi (1969) reasons that "The person involved in conventional activities is
tied to appointments, deadlines, working hours, plans, and the like . . . so the
opportunity to commit deviant acts rarely arises to the extent that he is
engrossed in conventional activities."

Belief in the conventional social norms is the fourth element. The
absence of effective beliefs that forbid socially unacceptable behavior leads to
delinquency. According to Hirschi (1969), belief in “conventionality” is
derived from intimate relationships with other persons, especially parents. In
other words, there is a causal chain from parental attachment to belief in
conventionality to delinquency.

Hirschi’s concept of attachment is extremely broad and problematic. It
includes attachment to parents, school, and peers. Parental attachment is
considered as the most important attachment variable and is operationlized as
a single construct. School attachment is not clearly differentiated from, and
thus is overlapping and redundant with, the concepts of involvement and
commitment. Marcos et al. (1986) suggest that a broader concept of
educational influences, incorporating both attitudes and behavior related to
education, appears to be a reasonable solution to this problem. In this study,

this researcher will define a broad concept -- educational commitment --

19

which includes the emotional bonding to school, involvement in activities
related to educational goals, and academic achievement.

Hirschi's concept of attachment to peers is another source of difficulty.
This difficulty stems from his assumption of a single moral order in the
society. However, given the existence of conflicting and competing norms in
the modern society, Hirschi's assumption of a single conventional norm seems
questionable. Also, empirical research shows that parental attachment
decreases delinquency regardless of whether the values of the parent(s) are
pro- or anti-delinquent (Johnson, 197 6; Marcos et al., 1986). Nevertheless,
the influence of peer attachment on delinquency depends on whether there is
"an excess of definitions favorable to law-breaking" (Sutherland & Cressey,
197 8). Also, peer attachment per se is not considered as important as other
social bond variables (Marcos, Bahr, & Johnson, 1986). While Hirschi found
that attachment to peers slightly reduced the probability of delinquent
behavior, Hindelang (1973) found the opposite in his replication of Hirschi's
study. In general, there is very little support that the level of peer attachment
controls against delinquent acts, after holding peers values constant (Conger,

1976)

20

The interrelationships among the bonding elements are not clearly
specified in the social control theory, except for the relationship between
family attachment/bonding and the adolescent’s conventional beliefs. Hirschi
(1969) argues that intimate relationships with others, especially parents, lead
to an adolescent's belief in conventional order. As for the interrelations in
general, among attachment, commitment, involvement, and belief, he explains
that the more closely a person is tied to conventional society in one way, the
more closely he or she is likely to be tied in other ways.

To specify more firlly the interrelationships among the re-
conceptualized attachment elements, which are family attachment/bonding,
educational commitment, and conventional belief, this researcher propose to
organize these elements in a chronological order of an adolescent's
involvement in the socialization settings. Thus, it is supported that a causal
chain goes from family attachment/bonding to adolescent's educational
commitment, and then to strength of conventional beliefs. Another causal link
exists between family attachment/bonding and the adolescent's conventional
belief. Also, family attachment/bonding is directly linked with conventional
belief as described by Hirschi (1969); to the degree that adolescents

experience a warm family environment, they are expected to possess the

21

controls necessary to dissuade temptations toward committing delinquent

behavior.

Differential Association Theory

Sutherland's differential association theory is basically a learning
theory. In contrast to Hirschi’s (1969) social control theory which assumes
that humans are inherently deviant, Sutherland's theory does not posit any
assumption about human nature, implying that humans are born neutral and
not “good” nor “bad.” It views tendencies towards delinquent behavior as
learned in interaction with others, where the learning takes place within
intimate personal groups (Sutherland & Cressey, 197 8). The theory assumes
that there are conflicting norms in the society, with pro-delinquent and anti-
delinquent values as the two extremes in a continuum, and that each of us is
"programmed" by our significant others in both directions. In other words, we
may learn both anti- and pro-delinquent values from the people around us.

The learning process of delinquent behavior is invariant, but the
context of learning varies across people. The learning contexts, such as
parents, peers, and communities, tend to vary. Different parents, different

peers, different people living in the community may present different values

22

about the conventional moral order to the adolescent. Quality of home
environment also afiects the probability of an adolescent’s association with
anti-delinquent and pro-delinquent patterns outside the home. For example,
an unpleasant experience or relationship at home may drive the adolescent
from home and thus increase his or her association with delinquent patterns
(Shaw & McKay, 1942; Sutherland & Cressey, 197 8).

In sum, Sutherland's differential association theory assumes that
modern societies contain conflicting norms, behavior patterns, and definitions
of appropriate behavior, in contrast to Hirschi's assumption of a single
conventional moral order. According to Sutherland, delinquency is rooted in
normative conflicts; adolescents become delinquent because they experience
an environment that “favors” law-breaking. The theory implies that (1) the
values of parents, peers, and the community directly influence the
adolescent’s self conventionality, and (2) family attachment and family control

affect self conventionality indirectly through peers' conventionality.

Delinquent Values and Behavior
Differential association in Sutherland's theory refers to association with

law-breaking behavior patterns displayed by parents and peers whom are not

23

necessarily criminals. prarents or peers approve or fail to disapprove of
criminal patterns, then the adolescent’ 5 contact with pro-delinquent values
increases. There is also an increase in the ratio of the adolescent's definitions,
of what appropriate behavior is, favorable to law violation over law abidance
as well.

In clarifying the concepts in differential association theory, Cressey
states that:

“If you are programmed to know only honesty is the best

policy, then you will be honest. If you are programmed to

know only that it is all right to steal, then you will steal.

But the matter is not so simple because, starting when we

are at our mothers' knees, we are programmed in both

ways. What the theory says, then, is that whether you will

steal or not depends on the ratio of these two kinds of

behavior patterns that have been put in you” (Bartollas,

1993)

Differential association theory also emphasizes that most youth learn to
act delinquently through association with delinquent peers (Sutherland &
Cressey, 1978; Akers et al., 1979; Elliot et al., 1985). And it is an unpleasant
relationship at home that drives the adolescent from home and thus increases
the adolescent's association with delinquent patterns (Sutherland & Cressey,

1978; Shaw & McKay, 1942). In other words, differential association theory

implies that parental attachment protects the adolescent from the probability

24

of association with delinquent peers and thereby prevents delinquency.
Hirschi (1969), on the other hand, argues that parental attachment plays a role
of psychological monitoring, therefore, adolescents who are strongly attached
to their parents are less likely to deviate from the norms of society. This
model implies that parental attachment has a direct impact on delinquency.
Association with delinquent peers, according to differential association
theory, contributes to adolescents' acquisition of pro-delinquent values.
Association with delinquent peers in the contexts of low parental and low
school attachment drives the adolescent to acquire and refine more delinquent
values and behavior. As a consequence, researchers have observed that
association with delinquent peers is a powerful predictor of delinquent
behavior. Specifically, studies show that compared to the effects of
individual, family, and school factors, association with deviant peers is almost
always the strongest predictor of delinquent behavior (Cashwell, 1994;
Gardner & Shoemaker, 1989; Johnstone, 197 8; Jensen, 1972; Thompson et

al., 1984).

25

The Ecological Contexts

During adolescence, an individual can be simultaneously influenced by
parents, peers, school, and the community context. However, few attempts
have been made to study the combined influences and the interrelations
among their effects in relation to adolescent delinquent behavior. The scarcity
of cross-contextual ecological studies on delinquency may be due to two
reasons: ( 1) cross-contextual studies are more complicated and time-
consuming than studies focusing on a single context, and (2) the importance
of studying different social contexts simultaneously was not recognized until
Bronfenbrenner introduced the concept of ecological contexts
(Bronfenbrenner 1986, 1989, 1993, 1994; Bronfenbrenner & Crouter, 1983).

Bronfenbrenner (1986) outlined four levels to his ecological system--
microsystem, mesosystem, exosystem, and macrosystem -- which provide
contexts for human-social interaction. The microsystem is an individual's
immediate environment; an adolescent's microsystem includes the family, the
school, and the peer group. Adolescents tend to spend time in more than one
microsystem. The linkages or processes that occur between the settings are
called the mesosystem. For example, the linkages between home and school

as well as home and peers, have developmental significance for adolescents.

26

Settings or events that occur in a setting that do not include the adolescent but
have influences on the adolescent are called the exosystem, The macrosystem
is the broadest level of the environment -- the culture or subculture in which
the adolescent and his or her family are living. The remainder of this chapter
discusses the different levels of Brofenbrenner’s ecological system and
focuses on the ecological concepts relevant to the proposed delinquency

model.

Family

Both Hirschi's (1969) social control theory and Sutherland's
(Sutherland & Cressey, 1978) differential association theory emphasize the
importance of family life in an adolescent's involvement with delinquent acts.
Although Hirschi emphasized attachment more than supervision in his original
version of social control theory, he considers parental management skills --
supervision and nurturance -- as equally important in his more recent study
(Hirschi, 1986). And in theory integration attempts, most theorists
(Cemkovick & Giordano, 1987; Elliott et. al., 1979; Thomberry, 1987) rely
heavily on both parental attachment and supervision to explain delinquency.

Furthermore, the theoretical emphasis on the family process is supported by

27

consistent research findings that show that secure parent-adolescent
attachment and wise parental supervision prevents delinquency.

According to Hirschi's (1969) social control theory, family attachment
directly protects adolescents fi'om delinquent acts; a direct link exits between
family attachment/bonding and the adolescent's delinquent behavior.
Adolescents who have warm and secure relationships with their parents are
likely to consider their parents' reactions to the deviant act. This is the so-
called “psychological monitoring.” As Hirschi's theory assumes a single
moral order in the society, parental reactions will always be negative and thus
discourage delinquent acts. Additional research has also found that parental
attachment decreases delinquency regardless of whether parents’ values are
anti- or pro-delinquent (Johnson, 197 6).

Both Hirschi and Sutherland also consider the indirect effects of family
attachment and control on adolescent delinquency. According to Hirschi ’ s
theory, the indirect influences of family attachment exist because family
attachment may increase both the adolescent's educational comnritrnent and
belief in conventional order (self conventionality), which are negatively

related to delinquent behavior.

28

According to Sutherland’s theory, the indirect effects of family factors
include: (1) the influence parents’ values have on adolescents’ values which
in turn have a direct impact on adolescent delinquency, and (2) the influence
family attachment and family control have on adolescents’ values toward and
behaviors of delinquency, through changing the probabilities of adolescents’

association with pro-delinquent and anti-delinquent peers.

Peers

In both neutral and antisocial situations, peer influences are important
forces in adolescent development. Peer interactions contribute to the
adolescent’s acquisition of behavioral norms and moral values (Panella,
Cooper & Henggeler, 1982). Also, the peer group is an important source of
imitation. Association with delinquent peers tends to decrease an adolescent's
belief in the conventional moral order and increase the likelihood of an
adolescent's committing delinquent acts.

Adolescents’ propensity to associate with pro-delinquent peers as
opposed to anti-delinquent peers may be associated with (1) a lack of

nurturant home environment and family attachment, (2) a lack of parental

29

control and supervision, (3) a lack of educational commitment and (4) a lack
of positive role models in the community.

As Sutherland’s theory (Sutherland & Cressey, 1978) explains, family
attachment and family control influence adolescents’ differential association
with a delinquent pattern in two ways. First, an unpleasant experience and
relationship at home may drive the adolescent from home and thus increase
his or her association with delinquent patterns. Second, a lack of parents’
disapproval and wise control that prevents violation of family rules increases
adolescents’ association with delinquents as opposed to nondelinquents.

Besides the family factors, educational commitment and community
trouble may influence adolescents’ delinquent behavior through differential
association with delinquent peers; association with delinquent peers has been
considered as the most proximate cause of delinquent behavior, next to the

effect of adolescents’ own attitudes and beliefs.

School
The relationship between school experience and delinquency has
typically been explained in two ways. Hirschi’s social control theory explains

the association as resulting from (1) a lack of attachment to significant others

30

(e. g., parents) who might encourage school related goals and (2) a shortage of
attachment, involvement, and commitment to conventional goals, which
causes students to be less controlled and thus more likely to engage in illegal
activities. However, some researchers still hold to a “strain” explanation;
their argument is that poor educational attachment, involvement, and
commitment inhibit students from status and school-related opportunities,
leading them to search for compensatory achievement through illegal
activities (Figueira-McDonough, 1986).

Hirschi's (1969) discussion about education in relation to delinquent
involvement includes: the adolescent’s feelings about school and importance
of good grades, time spent on school work, grades received, and educational
expectations. Strong attachment to teachers helps the adolescent to internalize
conventional norms. High levels of commitment to and investment in
education decrease the likelihood of an adolescent’s jeopardizing his or her
investment by violating the law. Also, involvement in academic activities
simply limits the time available to become involved with illegal activities.

A concept of educational influences, incorporating both attitudes and
behaviors related to education, will be used in this study to represent the

social bonding between the individual and school that plays the role of

31

protecting adolescents from delinquent acts. This concept will be called
“educational commitment,” which includes the adolescent’s academic
achievement, attachment to teachers, involvement in activities related to
educational goals, and academic achievement.

The direct impact of adolescents’ educational comrrritment on
delinquency has received substantial empirical support; both cross-sectional
and longitudinal studies show that low educational commitment is associated
with delinquent behavior (for a review, see Henggeler, 1989). Also, school
failure is a common risk predictor of the multiple problem behaviors of
adolescents (Dryfoos, 1990). Dryfoos suggests that when school failure
happens for at-risk youth, different types of delinquent behavior begin to
arise.

Besides the direct effect of educational commitment on adolescent
delinquent behavior, indirect effects may exist between educational
commitment and delinquency, mediated by self conventionality and peers’
conventionality. As for the mediation by self conventionality, Hirschi (1969)
suggests that strong attachment to teachers, a component of educational
commitment, helps the adolescent to internalize conventional norms. The

indirect effect of adolescents’ low educational commitment on delinquency

32

through their association with delinquent peers is consistent with the generally
accepted view that social bonding has an indirect relationship to delinquency,
through association with delinquent peers (Cattarello, 1993); weakening or
breaking of the bond will not directly cause one to commit delinquent
behavior, but simply frees an adolescent to do so. An adolescent with
weakened social bonding (for example, low educational commitment) is more

likely to select or be recruited by deviant peers.

Community

Adolescents' attitudes toward unlawful behaviors and conventional
activities may be subject to the influence of the subcultures that exist in the
communities in which they live. According to differential association theory
(Sutherland & Cressey, 1978), delinquency is rooted in normative conflicts,
and some subcultures in modern industrial society demonstrate Sutherland's
"excess of definitions favorable to law-violating." The normative conflict can
be translated into group level criminal rates through differential social
organization -- the extent to which a group is organized for or against law-
violating behaviors. Community subculture, specifically community trouble as

it is considered in this study, represents a source of the environmental

33

pressure at the community level that influences an adolescent's
conventionality.

Besides the direct effect of community subculture on adolescents,
Sutherland’s theory also considers the indirect community effect through peer
association; because of the preponderance of delinquent peers in the troubled
communities, adolescents living in these communities have more chance to be

in contact with and socialized toward pro-delinquency values.

Family-Peer Mesosystem Relationships

Changes in family and peer relations mark the transition from
childhood to adolescence. During adolescence, the increase in emotional
autonomy from parents is accompanied by the increase of susceptibility to
peer influences in both antisocial and neutral situations (Steinberg &
Silverberg, 1986). The adolescent's susceptibility to negative peer influences
depends on the family environment and parental management skills.

According to the differential association theory (Sutherland & Cressey,
1978), secure family attachment and wise parental control decreases
adolescents' delinquent acts through decreasing their contact with pro-

delinquent definitions. The family variables -- lack of attachment and control

34

-- are related to delinquency through the processes described below.

The home may fail to train the adolescent to deal with outside world
situations in a law-abiding manner. Delinquent patterns may not be present in
the home, but the home may be neutral with respect to antisocial behaviors.
Whether such a "neutr " adolescent becomes delinquent or not will depend
on his or her association with delinquent and anti-delinquent patterns outside
the home. Also, an adolescent may be driven from the home by unpleasant
experiences or by the absence of pleasant experiences. Isolation from the
family is likely to increase an adolescent’s association with delinquent
behavior patterns and to decrease his or her association with anti-delinquent
behavior patterns.

Also, difl°erential association theory argues that the lack of direct and
indirect control by parents (i.e., attachment and supervision) is associated
with delinquent behavior only when there are delinquent patterns to copy. In
other words, the known relationship between quality of family life and
delinquency is thought to be associated through peer context. If poor quality
of family life makes the adolescent vulnerable to the influence of delinquent
peers, then peer pressure becomes more influential and most likely anti-

conventional or pro-delinquent.

35

Family, Community, and School

Educational commitment has a direct impact on adolescents' delinquent
behavior; low educational commitment can follow from particular family and
community influences, specifically (I) insufficient parental support, (2)
inadequate aspiration for education, and (3) lack of existence of role models
in the surrounding community. Adolescents from families with low
attachment, and low expectations for pursuing conventional goals such as
education, and who are from communities where graduation from high school
does not offer better occupational opportunities, are more likely to experience
school failure which in turn is more likely to lead to delinquent acts (Huston,
1991)

Also, educational commitment is among the most documented
correlates of both delinquency and family socio-economic status (SES)
(Lipton & Smith, 1983; McGarvey et al., 1981). Studies show strong
associations among family SES, family preparation for and attitude toward
the adolescent’s education, and the adolescent's educational experiences
(Alexander et al., 197 8; Lipton & Smith, 1983). Moreover, a strong and
consistent relationship between educational commitment and non-delinquency

has been well demonstrated as discussed previously. The question is now:

36

how do the three variables (i.e., family SES, educational comrrritrnent, and
delinquency) link with each other?

Given the existing evidence that (1) the relationships are strong
between educational comnritrnent and family SES, and educational
commitment and non-delinquency, and (2) the relation between family SES
and delinquency is consistently low, the following interrelationship of the
three variables is proposed: the links go from family SES (variable A) to
educational commitment (variable B) and then from educational comnritrnent
to non-delinquency (variable C). In other words, delinquent behavior is
caused in part by low educational comrrritrnent which is caused in part by the
negative effect of low family SES. The low correlation between family SES
and delinquency is thus explainable: as the causal chain flows from variable A
to B, and then from B to C, the correlation between A and C tends to be low
because it is the multiplication of the correlation of A and B by the correlation
of B and C.

This hypothesis that family SES predicts educational commitment,
which in turn precedes and predicts adolescent delinquent behavior, is
supported in the literature, although a much debated issue remains in the

criminology literature over whether family SES influences the emergence of

37

delinquent behavior. Some argue that low family SES predicts intellectual and
emotional deprivation which motivates later illegal acts. Others suggest that
regardless of family SES, intelligence is the chief determinant of criminal
behavior. The View that family SES influences educational performance,
which in turn precedes the development of criminal activity has been tested
and supported (Lipton & Smith, 1983; McGravey et al., 1981).

In sum, low family SES, low parent-adolescent attachment, low
parental aspiration for education, and low levels of collective well-being in
the community (e. g., high crime, high unemployment rates, and widespread
poverty) tend to increase the probability of delinquent acts by influencing

adolescents’ educational experiences.

Gender Differences in Delingjrency
Since the start of time, males and females have been treated differently,
partly because they are considered different in biological, social, and
psychological characteristics. However, while a variety of contemporary
sociological theories attempt to explain adolescents’ delinquent behavior in

general, our understanding of gender differences in this regard remains

38

limited. The explanations of gender difl’erences in delinquency have mainly
relied on biological and psychological theories.

However, a controversy exists over the causal impact of biological
differences in delinquent behavior. Some criminologists challenge the
appropriateness of using biological factors to explain gender differences in
delinquent behavior. For instance, hormonal explanations of gender
differences in aggression for adults cannot be applied to the explanation of
gender differences in aggression of prepubescent boys and girls (Anderson,
1988)

The question of whether gender-specific sociological theories of
delinquency are necessary also requires fruther consideration. While some
feminist criminologists advocate a construction of completely new theories of
delinquency relating to the unique experiences of women, others prefer to
first test the usefulness of contemporary theories to explain female
delinquency (J ang, 1992).

Although the sociological theories of delinquency do not totally
exclude females from their theory consideration (e.g., Sutherland & Cressey,
197 8), their theoretical development often relies heavily on male subjects

(e. g., Hirschi, 1969). Also, in most empirical research on delinquency, though

39

both males and females tend to be included in study samples, gender is
normally treated as an independent variable, leaving investigations of possible
gender effects on relationships between variables (e.g., interaction effects)
unattended. As a consequence, previous studies have consistently found
greater involvement in delinquency among males than females, but the
theoretical explanations of gender differences remain unexplored. Moreover,
existing research tend to focus on just one or two social contexts, without
attending to the full range of social contexts (e.g., peers, family, school, and
community) that are important to both male and female adolescents’ everyday
lives (Canter, 1982; Cemkovich & Giordano, 1987, 1992; Figueira-

McDonough, 1983, 1986; Giordano et al., 1986; Tolan, 1988).

Prevalence and Incidence of Delinquency

Are there more male delinquents than female delinquents? Do males
commit delinquent acts more frequently than females? Answers to these
questions seem to depend on the type of delinquent behavior under
consideration.

In general, male prevalence rate is higher than female prevalence rate,

especially when seriousness of offenses was taken into consideration (Canter,

40

1982; Cemkovich & Giordano, 1979; Deschenes et al., 1990; Gold, 1970).
Males are more likely to engage in property and violent crimes. However,
females are more likely to engage in status offenses such as nmaway, defying
parents’ authority, and prostitution. In addition, it seems that there are few
gender differences in the prevalence rate of substance abuse. For instance,
Canter (1982) found that the only significant gender difference was in rate of
alcohol abuse; rates of use of marijuana and hard drugs were similar between
males and females.

As is the case with prevalence rates of delinquency, consistent research
findings show higher incidence rates of male delinquency (Canter, 1982;
Cemkovich & Giordano, 197 9; Deschenes et al., 1990; Friedman &
Rosenbaum, 1988; Gold, 1970; Hindelang, 197 9). Although self-report data
seem to show smaller gender differences than official data, significant gender
differences in incidence rate exist, no matter which type of data are used

(Datesman & Scarpitti, 1980; Hindelang, 1971).

Patterns of Changes
Besides the above mentioned issues of gender differences, patterns of

changes in gender differences in delinquency over time are also intriguing.

41

Changes in both historical patterns and developmental patterns are the two
mostly studied topics.

The so-called “emancipation hypothesis” or “liberation hypothesis” is
the central focus of the studies investigating changes in the historical pattern.
It is hypothesized that since the modernization and the movement of women’s
liberation have made socialization patterns increasingly similar for males and
females, and have brought women more opportunities to comrrrit delinquent
acts than before, the magnitude of gender differences in both prevalence rate
and incidence rate of delinquency are expected to increase over time.

The emancipation thesis has generated a series of heated debates
among researchers. Research findings in this area are mixed. While a
considerable amount of studies finds increase in arrest rates of females,
especially for larceny (e. g., Alder, 1975; Simon, 1975; Streffensmeier &
Streffensmeier, 1980) and for substance abuse (e.g., Canter, 1982; Gold &
Reimer, 1975), controversies exist as to whether the women’ s movement is
responsible for the increase in female crime.

In the explanation of increased arrest rates of females, Simon’s (1975)
and Adler’s (1975 , 1980) works are the most often cited, concerning the

gender convergence. Simon (197 5) argues that the liberation increases

42

females’ likelihood to engage in the types of crimes, for which their works
outside the home provide them with increased opportunities, such as larceny.
Adler (1975), on the other hand, proposes that the gap between male and
female delinquency has been narrowed in both type and frequency of
delinquent acts due to increasing unisex role expectations and socializations.

In response to the liberation argument of Simon (1975) and Adler
(1975), a series of studies were generated that provided only limited support
for the notion that the women’s movement is responsible for the increase in
female crime (e.g., Giordano & Cemkovich, 197 9; Thornton, 1982).
Insignificant associations (accompanying a mixed findings of positive and
negative relationships) between female liberated attitudes and self-reported
delinquency involvement were reported. Some attribute the insignificant
findings to (1) possible oversirnplification of the notion of liberation of
women (i.e., viewing liberalization as masculirrization), (2)
multidimensionality of the concept of gender roles, and (3) differential
meaning of gender roles for individuals of different race or social classes
(Giordano & Cemkovich, 197 9).

Similarly, in response to Simon’s (1975) hypothesis, Figueira-

McDonough (1984) argues that the influence of the women’s movement on

43

delinquent acts is “far more complex, less linear, and much more tenuous”
than suggested by Simon. Leonard (1982) argues that Alder’s attribution of
gender convergence in delinquency is problematic; she stresses that Alder’s
assumption of equality among males and females is exaggerated, and
criticizes Alder’s equating of liberation with masculinization. Still other
investigators attribute the increased rates of female delinquency to the
influence of the women’s movement on the treatment of women by criminal
justice authorities (Datesman et al., 1975 ; Moulds, 1980) or to a greater
willingness and improved methods to detect less serious ofl°enses such as
shop-lifting (Steflensmeier & Steffensmeier, 1980).

Developmental changes in gender differences in delinquency is another
interesting research topic. While there are some discussions on changes in
developmental patterns of gender differences (e.g., Famworth, 1984), existing
research tends to focus on the influence of age, rather than developmental
stage, on delinquent behavior itself or gender differences in this issue. In
addition, controversies exist concerning whether male and female delinquent
involvements will diverge or converge over time.

Although adolescents tend to enjoy greater freedom from family

control over, and intervention into, their lives as they grow older (McCarthy

44

& Hagan, 1987), some studies show this process takes place only for males
and the opposite in the case with females (Chesney-Lind & Shelden, 1992). If
such differential treatment exists for males and females, gender differences in
delinquent involvement would diverge. However, Famworth (1984) argues
that male and female delinquency will converge, not diverge, as socialization
and social relationships grow increasingly gender-neutral. A quantitative
review to investigate these two conflicting hypotheses is not possible at this

time, due to the insufficient amount of primary studies addressing this issue.

The Proposed Ecological Model of Delinquency

In light of the above review of juvenile delinquency variables and their
contextual nature in families, schools, and communities, an ecological model
of delinquency is proposed as a promising strategy for addressing this
complexity. The ecological model of delinquency (Figure 1) assumes: (1)
human beings are born neutral in terms of values which in turn are subject to
influences from the environment, (2) conflicting norms and definitions about
appropriate behavior exist in the U. S. society, and thus (3) criminal

motivation is a variable and not a constant.

45

The model has three groups of variables -- four background variables,
four intervening variables selected from Sutherland’s differential association
and Hirschi’s social control theories, and one outcome variable of different
types of self-reported delinquent behavior. The four background variables are
family SES, community trouble, family attachment/bonding, and family
control/supervision. The four intervening variables are educational
commitment and the values toward antisocial behavior (conventionality) of
parents, peers, and adolescents. The outcome variable, delinquency, is
categorized into (1) official criminal offenses (i.e., index ofi’enses) -- crimes
against property and people, (2) status offenses, (3) substance abuse, and (4)
general delinquency, which is a combination of the different types of
delinquent behavior.

One important issue is the arrangement of the variables. Because the
proposed model is primarily an integration of Sutherland’s (Sutherland &
Cressey, 1978) and Hirschi’s (1969) theories, the arrangement of the
variables is set to reflect the propositions of the theories. When both theories
are not explicit about the relationships, the arrangement will be based on the
chronological order of an adolescent’s involvement in the socialization

settings; for example, family is considered the first socialization setting,

46

school the second, and then the peer group factor comes in to compete with
family and school influences. Also, the human ecological perspective,
specifically the interrelations between family, school, peer groups, and
community, provides a special focus for the construction of the model.

Hirschi's four elements of social bonding (i.e., attachment,
commitrrrent, involvement, and belief) are reconceptualized into three
categories--fami1y attachment/bonding, educational commitment, and the
adolescent's belief in the conventional moral order. The interrelations among
these three categories of social bonding and their relationships to delinquency
may be summarized by the following research questions: (1) does family
attachment have a direct impact on adolescent delinquency? (2) does
educational commitment have a direct influence on adolescent delinquency?
(3) does adolescents’ self conventionality have a direct efiect on adolescent
delinquency? (4) is there a direct path flowing fi'om family attachment to
educational commitment? (5) is there a direct link between family attachment
and self conventionality? (6) is there a direct link going from educational
commitment to self conventionality?

Based on Sutherland's differential association theory, the following

research questions are generated: (1) is there a direct relationship between

47

peers’ conventionality and adolescent delinquency? (2) is there a direct
relationship between adolescent’s self conventionality and adolescent
delinquency? (3) do family factors (including family attachment and family
control) directly influence the likelihood of the adolescent's association with
delinquent peers? (4) does community trouble have a direct effect on peers’
conventionality? (5) does parental conventionality have a direct influence on
the adolescent’s self conventionality, (6) is there a direct relationship between
peers’ conventionality and the adolescent’s self conventionality?

The research literature reviewed suggests that this model should
propose that adolescents' differential association with delinquent peers is
affected by educational commitment, although differential association theory
states that differential association is determined solely by family and peer
factors. In addition, adolescents’ educational commitment are suggested to be
influenced by community trouble, family SES, parental conventionality
(parents’ adherenceto values of education is a component of this measure),

besides by family attachment which is proposed by Hirschi (1969).

48

Parental Community
Conventionality Trouble

  

 

 

 

Family ‘ Educational

SES ' Commitment ----------------------------- ~> Conventionality
Family L Peers’
Control ' Conventionality’i_~

 

   

Family
Attachment """""""""""""""" " Delinquency

Figure 1. An Ecological Model of Delinquency
Note:___., Links derived fi'om Sutherland’s differential association theory
............ _, Links derived fi'om Hirschi’s social control theory
__, Links added based on human ecological theory

CHAPTER 3

METHODOLOGY

This chapter describes the meta-analysis conducted to investigate the
conceptual model derived from a review of existing research and presented in
Figure 1. It provides details regarding the literature search, the criteria used
for inclusion and exclusion of the relevant studies, the information recorded
from the studies, study and sample characteristics, missing data, and
procedures used to examine the proposed social-ecological model of
delinquency. The descriptions of research variables and hypotheses derived

from the conceptual model are also presented.

Literature Search
This study’s literature search employed the approaches suggested by
Cooper (1989). The search targeted studies examining relationships between
measures of delinquent behavior and social-ecological causes such as family
management skills, parents’ and peers’ attitudes toward the conventional
moral order, school environments, and community subcultures. The

population of interest consisted of all US. adolescents.

49

50

Ninety-eight reports were located from on-line or manual searches of a
number of indices: Criminal Justice Abstract, Criminal Justice Publication
Index, Dissertation Abstract Index, ERIC (Educational Resource Information
Center), NCJRS (National Criminal Justice Resource Services), PsycInfo,
SocioFiles, and Social Science Citation Index. The first groups of key words
used were: aggression, alcohol use, alcoholism, antisocial behavior,
cigarette smoking, delinquency, delinquent behavior, deviance, deviant
behavior, drinking, drug abuse, drug use, externalizing disorder, juvenile
delinquency, violent behavior, problem behavior, violence, substance use,
and substance abuse. The search results were further narrowed down, by
intersecting the above set of references, with the following key words, one at
a time: family, parent, peer, school, education, community, SES, and social
class.

Thirty-six candidate reports were added by skimming through the
references of the reports discovered by the computerized search. Fifteen more
relevant reports were found by investigating studies that cited the same
reference(s) as did the first 98 studies found. Finally, 9 additional references
were provided through discussion with Rolf Loeber, a professor at the
University of Pittsburgh, who has carried out several meta-analyses on the

relation of family factors to delinquency (e. g., Loeber & Dishion, 1983;

51

Loeber & Stouthamer—Loeber, 1986). Thus, these procedures retrieved a total
of 158 initial sources that seemed to be relevant. Each report found was then

tested against a variety of criteria.

Inclusion and Exclusion Criteria

Before its results were entered into the analysis, a report needed to
meet the following criteria: (1) the report was made between 1970 and 1995,
(2) the report concentrated on relationships among socio-ecological factors
(as opposed to biological or psychological factors) and delinquent outcomes,
(3) delinquency was measured as a continuous variable, (4) enough
information was provided about the definitions or measurements of the
variables of interest so that construct validity could be examined, (5) zero-
order correlation coefficients of the interrelationships among the variables of
interest were either available or accessible, (6) samples were drawn from the
average US. adolescent population, i.e., the study did not employ clinical
samples or samples for comparison purposes (e.g., delinquent vs. non-
delinquent group), and (7) the delinquency measure was based on self-report
data rather than official reports.

The first two criteria were employed for the following reasons. First,

restricting the collection of studies to the past 25 years eliminated studies

52

that may no longer be relevant to understanding modem-day delinquency.
Second, this meta-analysis focused on the social-ecological causes of
delinquency because biological studies of delinquency have not provided
enough information, especially in terms of quantity, for research synthesis,
and current theories have not attended to the simultaneous considerations of
biological, psychological, and social-environmental factors. Limiting the
focus of this meta-analysis to the social-ecological causes does not deny the
importance of biological and psychological factors in explaining delinquent
behavior.

The third and fourth criteria enhanced the compatibility of the studies
for synthesis. Special care is needed in combining studies that use different
designs or measurements to answer the same research question. It is not
appropriate to combine data that measure delinquent behavior as dichotomous
with data that measure delinquent behavior as continuous. Also, it is not
appropriate to use broad categories that subsume numerous causes and
outcomes of delinquent behavior. More information about the operational
definitions used in this meta-analysis is presented later in the Research
Variables section, in this chapter.

The fifth criterion was necessary because zero-order correlations are

the indices of study outcome in this research synthesis. In order to be

53

included into this meta-analysis, the sample studies need to provide the
correlation coefficients or other information (e.g., a variance-covariance
matrix) that can be converted to correlation coeflicients. Techniques that
utilize less information, such as p-value syntheses (see, e.g., Becker &
Schram, 1994) would be able to utilize more studies, as the inclusion criteria
would be less strict. However, this would preclude multivariate analyses or
the estimation of effect magnitudes.

The last two criteria were used to ensure the validity of this meta-
analysis. Many studies were excluded from this meta-analysis due to their use
of comparison samples instead of average US. adolescent samples: some
compared delinquents with non-delinquents, and some compared clinical
adolescents with non-clinical adolescents. Inclusion of studies that use
comparison samples would be problematic because the purpose of this meta-
analysis is to generate and test models that predict delinquent behavior in the
overall population of adolescents. While the use of comparison samples is
appropriate in explicating a set of variables relating to delinquent and
nondelinquent populations, it is not appropriate for predicting delinquent
behavior.

Other studies were excluded because they used official data rather than

self-report data as measures of delinquency. The validity of official data is

54

questionable due to administrative and procedural errors and police bias in
arrest decision-making (Siegal & Senna, 1988). Also, the use of official data
excludes adolescents whose unlawful acts are not detected.

Using the above mentioned criteria to screen the 15 8 reports, 36
research reports (containing 50 separate studies) remained for further
analyses and syntheses. A bibliography of these reports is available in

Appendix A.

Coding

Information of interest was extracted from the collection of 36 selected
research reports. An initial version of the coding form was pilot-tested with
the first ten reports obtained. The coding form was then modified to eliminate
the problems encountered in the pilot-test and to facilitate subsequent
computer entry and data analysis. Appendix B contains the coding form used
in recording the information found in the 36 research reports.

The information coded included: (1) name(s) of the author(s), title and
source of the report, and year of publication/report, (2) sampling procedures
used in the report and characteristics of the sample(s), (3) the

conceptualization and measurement methods of the variables of interest, (4)

55

reliability of the measurements, and (5) zero-order correlations (r’s) among
the variables of interest.

Sample characteristics of interest were the distributions of gender, age,
ethnicity, and geographic location of the residences of the participants in the
studies. As the examination of gender differences was of particular interest in
this study, when the rs reported were calculated based on mixed gender
samples, the author(s) of the report were requested to provide a breakdown of
the correlations by gender. Five out of 21 breakdown requests were provided.

The coding was completed by two persons -- Meng-Li Yang, a Ph. D.
candidate in Measurement and Quantitative Methods (MQM) Program at
Michigan State University, and this researcher. The coding reliability (i.e.,
percent agreement between coders) ranged from 75% to 100%. All

discrepancies in coding were resolved.

Unit of Analysis and Synthesis
The 423 correlation coefficients (the data points) used in the present
meta-analysis were collected from 50 separate studies using independent
samples in 36 research reports. A report refers to a written description of
research, while a study refers to an element of the report based on an

independent sample. Some reports used more than one independent sample

56

and thus contain more than one study. For example, a research report might
provide correlation coefficients separately for boys and girls or for blacks and

nonblacks, and thus would equal two studies.

Study Characteristics

Years of report of the studies are summarized in Table 1. The studies
were reported during a time span of 25 years. The majority (40%) of the
studies were reported between 1986 and 1990. Over half of the studies had
appeared in last 10 years.

Table 1. Year of report of the studies in the synthesis.

 

Year of report Number of studies Percentage

 

1991-1995 11 22
1986-1990 20 40
1981-1985 13 26
1976-1980 4 8
1971-1975 2 4

Total 50 100

 

Numbers of studies and of subjects are summarized by gender and by
sample size distribution in Table 2 and Table 3. Of the 50 studies, 22 (44%)
studied mixed samples of boys and girls, while 17 (34%) studied boys and 11
(22%) studied girls separately. Eighteen studies (36%) used sample sizes

larger than 900, five studies (10%) had sample sizes between 601 and 900,

57

twelve studies (24%) had sample sizes between 301 and 600, and fifteen

studies (30%) had sample sizes less than 300.

Table 2. Numbers of studies by gender and sample size distribution
Sample size distribution All girls All boys Mixed Total

 

< 300 5 7 3 15
310-600 3 5 4 12
601-900 2 O 3 5

> 900 1 5 12 18

Total 1 1 17 22 50

 

Table 3. Number of subjects by gender and sample size distribution
Sample size distribution All girls All boys Mixed Total

 

< 300 1,026 1,231 337 2,594
310-600 1,024 2,012 1,683 4,719
601-900 1,702 0 2,046 3,748

> 900 1,374 5,708 50,074 57,156

Total 5,126 8,951 54,140 68,217

 

Sample Characteristics
A total of 68,217 US. adolescents were included in the 50 studies.
About one-fourth of the adolescents were from national representative
samples; one-half were selected as state representative samples; and the rest
were sampled from specific rural, urban, and suburban areas across the U. S..
The ages of the adolescents ranged from 10 to 20 years, with the majority in

the range of 11 to 17 years. Based on the information provided in the sample

58

studies, about 77% of the total sample was Caucasian American and 53%

was male.

Research Variables

Research variable used in this research synthesis are often broad
categories that could have more than one measures. When multiple measures
are used in the primary studies, the correlations of each measure with another
variable had to be statistically combined before being entered into the meta-
analytical procedure. To obtain a composite correlation, the Olkin & Siotani
(1976) formula was first used to compute the covariance among the individual
subcategorical (component) correlations, then the weights for combining the
component correlations were generated using the procedures described by
Gleser & Olkin (1994). The more the component correlation correlates with
other component correlations, the higher the weight it receives. A summary of
how the variables are conceptualized in this study is presented below.

Family bonding/attachment (variable 1) is a measure of the social
bonding between the adolescents and their parents; measurements under this
construct include the adolescents’ response to questions asking about the

levels of (1) time spent with their parents, (2) caring and support from their

59

parents (e.g., parents understand and listen to them), and (3) affection ties (or
rejection) between themselves and their parents.

F arnily control/supervision (variable 2) is a measure of discipline and
monitoring from parents toward the adolescents; this variable includes the
measures of parents’ (1) awareness of the adolescents’ whereabouts and daily
plans, (2) setting of rules (e. g., of curfew, manner, appearance) and follow-up
of threatened punishments, and (3) letting their adolescents get away with
misbehavior.

Educational comnritrnent (variable 3) is a measure incorporating both
attitudes and behaviors of adolescents toward education. Specifically, this
variable reflects adolescents’ educational aspiration, emotional bonding to
school, involvement in activities related to educational goals, and academic
achievement. The questions asked include: (1) how important the adolescents
think education is, (2) how much the adolescents like or enjoy their schools
and teachers, (3) how much time the adolescents spend in doing homework
and activities related to educational goals, and (4) how well do they do in
school academically (e. g., GPA).

Community troubles (variable 4) are measures of the community
subcultures that focus on the types of resources, role models, and cultural

norms existing in the community; these factors imply the aspirations and

60

expectations of the community. Community characteristics measured to
reflect community troubles include (1) low group socio-economic level and
(2) the extent to which people in the community “get into trouble” (e.g.,
selling drugs, police arrests).

The measure of adolescents’ conventionality (variable 5) focuses on
their belief in conventional morality, such as their willingness to respect
others, obey the law, follow the rules, and stay out of trouble. Parental
conventionality (variable 6) is a measure of the parents’ adherence to such
values as hard work, deferred gratification, and formal education. Peers’
conventionality (variable 7) is a measure of peers’ attitudes toward delinquent
behavior in general, which ranges from the merely troublesome to the illegal.

General delinquency (variable 8) is a measure which combines
different types of delinquent behavior, including substance abuse, index
offenses (illegal behavior whether conducted by adults or minors), and status
offenses (behavior that is illegal only when conducted by minors). Although
some researchers consider index offenses, status offenses, and substance
abuse as distinct types of delinquent behavior, others tend to combine the
subcategories, often giving more weight to both more serious and more

frequent behavior, while they study delinquency.

61

Variable 9, 10, and 11 measure the frequency of adolescents’
involvement in index offenses, status offenses, and substance abuse
respectively.

The combination of parents’ educational level and class position is
used to measure the family’s socio-economic status (variable 12). Parents’
educational level and occupation are considered as more accurate measures of
social class than family income, because most parents seldom discuss salary

matters with their adolescents.

Missing Data

In order to have optimal results of model-driven meta-analysis (to
“piece together the existing studies” by synthesizing their statistical results
and to evaluate a theoretical model with the synthesized results), one needs
the relationships, being analyzed in the existing studies, to overlap as much as
possible. Existence of “missing” data, depending on the amount and the
nature of missing, often results in the inability to fully consider the
multivariate relationships or to investigate the interaction effects. Specifically,
the problem of “missing” data forces this model-driven meta-analysis (1) to
take a random-effects model approach, instead of using a fixed-effects model

approach to investigate the moderator effects, and (2) to look at gender

62

differences on limited numbers of relationships between the social-ecological
variables of delinquency and delinquent outcomes, rather than to evaluate the
whole model for both gender separately.

This meta-analysis studied 12 variables and their 66 possible
intercorrelations. With the 50 studies used in this meta-analysis, the total
number of correlation coefficients possible is 3300. However, none of the 50
studies contain information about all of the 66 relationships. A considerable
amount of “missing” data exist; this meta-analysis gathered only 12.8% of the
possible number of correlation coefficients (423 out of 3300).

A summary of number of studies by the number of intercorrelations
provided in each study is presented in Table 4. The majority (40%) of the
studies individually contributed only one to three correlation coefficients to
this meta-analysis. Only three studies (6%) provided more than 15 correlation

coefficients.

63

Table 4. Number of studies by number of correlation coefficients provided in

 

 

each study.
Number of Correlations Provided in Number of Studies
each Study
1 to 3 20
4 to 6 10
7 to 9 6
10 to 12 5
13 to 15 6
more than 15 3
Total 50

 

Table 5 . Relationships investigated with multiple male- and female-only
samples by number of studies

Relationships # of Studies: for Females # of Studies: for Males

 

an&2
Va1&8
le&H)
Va1&ll
Vm2&8
Vu2&9
Var.2&10
Vu5&8
Vn8&d2

NNUJNQAANDO

 

Note: Var. 1 = Family Attachment
Var. 2 = Family Control
Var. 5 = Self Conventionality
Var. 8 = General Delinquency
Var. 9 = Index Offenses
Var. 10 = Status Offenses
Var. 11 = Substance Abuse
Var. 12 = Family SES

64

Table 6. The most and least studied relationships.
# of studies Relationships studied (# of studies)

 

More than 10 Family attachment and general delinquency (20 studies)
Family attachment and family control (16 studies)
Family control and general delinquency (15 studies)
Family attachment and status oflenses (14 studies)
Family attachment and substance abuse (11 studies)

1 to 2 Family control and educational commitment (2 studies)
Family control and community trouble (2 studies)
Family control and self conventionality (2 studies)
Community trouble and index oflenses (2 studies)
Self conventionality and index oflenses (2 studies)
General delinquency and status oflenses (2 studies)
Parental conventionality and family SES (1 study)
General delinquency and index oflenses (1 study)
Substance abuse and family SES (1 study)

Missing Family control and parental conventionality (0 studies)
Community trouble and parental conventionality (0 studies)

 

Limited number of studies looked at the relationships between
variables separately for the two genders (see, Table 5). Only 9 out of 66
relationships were investigated with multiple male- and female-only samples.
And the focus of the investigation seemed to restrict on the relationships

between family factors and delinquent outcomes.

65

Most apd Least Frequently Studied Relationships
A summary of the most and least studied relationships is provided in
Table 6. For detailed information regarding the extent to which the 50 studies
addressed the 66 possible relationships among the 12 variables used in the

proposed conceptual model, see Table 7.

Meta-Analytical Procedures

Overview

The correlations (rs) collected from the 50 studies were analyzed
using methods proposed by Becker (1992) and refined by Becker and
Fahrbach (1994). Before conducting further analyses, the rs were first
transformed into Fisher zs (Fisher, 1928). The purpose of this transformation
was to normalize the distribution of the correlations, as well as simplify their
variance. Statistical analyses that rely on assumptions of normality can then
be safely applied. After analyses were completed, estimates were transformed
back into the r-metric to provide interpretability.

In all analyses, each study was weighted depending on its sample size;

studies using larger sample sizes are presumably more precise in the

66

estimations of relationships, and thus deserve more weight in research
syntheses than studies using smaller sample sizes.

Both univariate and multivariate homogeneity tests were then
conducted to investigate whether the sample correlations seemed to derive
from a single homogeneous population. Multiple univariate homogeneity tests
were conducted on the samples of bivariate correlations, while one
multivariate homogeneity test was used to consider all the correlations
simultaneously. Finally, based on the results of the homogeneity tests, a
decision was made to use either a fixed- or random-effects model for
examination of the proposed social-ecological model of delinquency.

The population correlations between delinquency measures and their
social-ecological predictors (i.e., the estimation of central tendency) were
then estimated. The test of homogeneity of the correlations helped determine
whether population variation needs to be considered a factor in the
consideration of the relationship. Finally, the estimated population
correlations and their variances were used to examine the proposed

conceptual model of delinquency.

67

Homogeneity Tests

A homogeneity test was used to decide if the observed variance in the
sample correlations can be fully explained by sampling error; in other words,
to test whether it seemed plausible that the variation observed in the sample
observations was due to chance alone GIedges & Olkin, 1985).

If the expected sampling error variance and the observed variance are
similar, it is safe to conclude that the relation between the pair of variables is
constant across different samples (with differences in gender, race, social
class, geographic locations, and other unmeasured characteristics). In other
words, a series of sample correlations are said to be homogeneous (i.e., to
share a common population correlation) if the variance in the sample
correlations can be explained by the sampling error. It is important to note
that this inference of homogeneity of population correlations only applies to
the types of variation in study characteristics that were observed in the
synthesis.

Two models exist for explaining why the distribution of population
correlations seems to be heterogeneous. The first is a fixed-effects model that
relies on moderator analysis to explain the extra variation. Population
correlations are thought to differ systematically depending on study or sample

characteristics. In contrast, the second model, the random-effects model, does

68

not attempt to explain away the variation by relying on sample or study
characteristics. Instead, the simple random-efi‘ects model assumes that the
population correlations have a normal distribution. For detailed discussions of
the fixed- and random-effects models, see Hedges (1994) and Raudenbush
(1994)

In this study, missing data problems forced the use of a random-effects
model to explain the surplus variation that is not explained by sampling error.
When the studies are considered random, two approaches are available for
the estimation of the parameters: the “method of moments” and the
“interative maximum likelihood” approaches. The “method of moments”
approach was employed in this study for the estimation of the population

parameters and firrther estimation of the linear models.

Estimation of the Linear Models

The estimated population correlations and their variances were then
used to examine the proposed social-ecological model of delinquency (see,
Figure 1) by using methods proposed by Becker (1992) and Becker &
Schram (1994). The method for estimating linear models outlined by Becker
(1992) was used to conduct the 7 standardized regressions with which the

multiple relationships among the social-environmental factors and delinquent

69

behaviors, proposed in Figure 1, were specified. Using model-guided meta-
analysis, this study aimed to generate an empirical social-environmental
model of delinquency that fits the data in the existing research.

Using models in research synthesis allows the simultaneous
examination of multiple relationships (Becker & Schram, 1994). Primary
studies tend to focus on only one or several paths in a model. Like the blind
man’s description of the elephant, the primary studies may provide many
conclusions, but little understanding of the big picture under study (Eccles et
al., 1983). In order to piece together the information in the primary studies to
provide a more complete explanation of the phenomenon, one needs to
employ the approach of model-driven meta-analysis.

The above mentioned model-driven meta-analysis procedures are
programmed in SAS IML (Statistical Analysis System, Interactive Matrix
Language) by Kyle Fahrbach, a Ph.D. candidate in the Measurement and

Quantitative Methods program at Michigan State University.

7O

Hypotheses

Given the review of research (which culminated in a proposed model
of delinquency and a set of related research questions -- see Chapter 2) as
well as the above discussion of the samples, the variables under
consideration, and the meta-analytical procedures, the proposed ecological
model of delinquency presented in Figure 1 (Chapter 2) can be tested. The
test of the proposed ecological model of delinquency will involve seven
standardized regressions which reflect the following hypotheses:

Hypothesis 1: Educational commitment is predicted by family
bonding/attachment, community trouble, parental conventionality, and
family SES.

Hypothesis 2: Adolescent’s self conventionality is predicted by family
bonding/ attachment, educational commitment, community trouble,
parental conventionality, and peers’ conventionality.

Hypothesis 3: Peers’ conventionality is predicted by family
bonding/attachment, family control/supervision, and educational
commitment.

Hypothesis 4: General delinquency is predicted by family
bonding/ attachment, educational commitment, self-conventionality,

and peers’ conventionality.

71

Hypothesis 5: Index offenses are predicted by family bonding/attachment,
educational commitment, self-conventionality, and peers’
conventionality.

Hypothesis 6: Status offenses are predicted by family bonding/attachment,
educational commitment, self-conventionality, and peers’
conventionality.

Hypothesis 7: Substance abuse is predicted by family bonding/attachment,
educational commitment, self-conventionality, and peers’

conventionality.

Because an insufficient number of studies existed to evaluate the
proposed model separately for males and females, gender differences will be
addressed only through the comparison of study results based on gender-
specific samples. For every relationship between variables that allows for an

investigation of gender differences, a general hypothesis is stated as:

Hypothesis 8: The average correlation, representing the relationship between
variables in the model (e.g., between family attachment and status

offenses), is not significantly different between males and females.

CHAPTER 4
RESULTS

This meta-analysis analyzes 50 studies, extracted from 26 published
and 10 unpublished research reports. The total number of correlation
coefficients under synthesis is 423. The total sample includes 68,217 US.
adolescents. The results presented here include: (1) homogeneity tests, (2)
average correlations, (3) standardized regressions, and (4) the revised models
of delinquent behavior based on the integration of the results of the

standardized regressions.

flgmggeneity Test Resfitp
Both multivariate and univariate homogeneity tests were used to

investigate whether the sample correlations collected from the 50 studies
were derived from a homogeneous population. There was only one
multivariate homogeneity test, which considered whether the study
correlations taken together as 12 x 12 correlation matrices seemed to share an
underlying population correlation matrix. This was a test of the fit of the
fixed-effects model. There were 66 possible univariate tests (i.e., the first test

was conducted on r1,1.2, , r50,1.2, the second on r2,1.3, , r50,1.3, up until

72

73

the 66th on m, 11.12, ,1'50,11o12; where rk, m... is the correlation between
variables m and n in the kth study). However, as none of the 50 studies
studied the relationship between parental conventionality and fam ily control
or between parental conventionality and community trouble, only 64
univariate tests of homogeneity were conducted. Each test was conducted to
determine if the sample correlations, for each relationship, shared an
underlying population correlation.

The multivariate homogeneity test statistic (QB) has approximately a
chi-square distribution with degrees of freedom (dj) equal to [(k-l) x p'] - #
of missing correlations, where k is the number of studies under synthesis and
p' refers to the number of possible relationships within a study. This meta-
analysis analyzed 50 studies and 66 possible relationships. So, the QB value
for the multivariate test of homogeneity was compared to the critical value of
the chi-square distribution, at the 5% significance level with df = 357 . For
these data, the multivariate Q; value was 4915.19 (p < 0.001) indicating that
it was unlikely that the sample correlations contributed by the 50 studies
shared a common population correlation matrix. The rejection of the
multivariate null hypothesis suggested that the fixed-efi‘ects model did not fit

the data.

74

The univariate QE’s, which test individual cells’ homogeneity, have
approximate chi-square distributions with degrees of freedom equal to the
number of studies that contained a sample correlation for the cell in question
minus one. Ten of the univariate homogenous test results were not significant

at the 0.05 significance level (see, Table 7).

Table 7. The relationships with insignificant homogeneity test results8

 

Relationships QB p-value of Q1; df
Family attachment & general delinquency 26.10 0.13 19
Family control & status ofienses 11.37 0.12 7
Family control & substance abuse 3.59 0.46 4
Family control & family SES 5.92 0.20 4
Educational commitment & index offenses 5.27 0.07 2
Community trouble & index offenses 0.62 0.43 1
Self conventionality & family SES 4.13 0.53 5
Peers’ conventionality & status oflenses 6.62 0.36 6
General delinquency & substance abuse 1.10 0.29 1
Substance abuse & family SES 2.85 0.24 2

 

Note: a = at 0.05 significance level, the sample correlations of each of these
relationships seem to share a homogeneous underlying population correlation.

Average Correlations

Because the homogeneity tests showed that a random-effects model,

assuming variations in the population correlations, fits the data better than a

fixed-effects model, further analyses were based on the random-effects

75

model. When the null hypotheses of homogeneity are rejected, one might still
prefer to use a fixed-effects model, to assume a single population parameter,
and to seek explanation for the variation with sample and study
characteristics. However, as a substantial amount of missing data exists in
this meta-analysis, it seemed that a random-effects model was the only
option. Unfortunately, because there was a considerable amount of missing
data, there were not enough degrees of freedom to estimate the covariances
well. Thus, the random-effects model utilized considers only the variances
among the population correlations. In other words, for statistical reasons, a
fully multivariate random effects analysis could not be conducted; all random
effects analysis presented later on are univariate in nature.

Table 8 shows (1) the mean correlations, (2) the estimated 95%
confidence interval for each mean correlation, and (3) a 2 test for whether the

mean correlation was significantly different fiom zero.

76

Table 8. The random-effects results: Correlations, Z-tests, number of sample
correlations, and number of sub'ects.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Lower Mean Upper Z score Z p- Number Number
Limit Value Limit value of of Total

Sample Subjects

r’s

Var. 1 & 2 0.17 0.25 0.32 6.00 0.00 16 7247
Var. l & 3 0.23 0.29 0.35 9.09 0.00 4 35437
Var. 1 & 4 -0.20 -0.10 -0.01 -2.09 0.04 4 4559
Var. 1 & 5 0.26 0.34 0.40 8.65 0.00 10 35549
Var. 1 & 6 0.23 0.31 0.39 6.90 0.00 4 5040
Var. 1 & 7 0.22 0.27 0.31 11.01 0.00 10 8801
Var. l & 8 -0.28 -0.26 -0.24 -27.82 0.00 20 10865
Var. 1 & 9 -0.30 -0.21 -0.12 -4.58 0.00 7 7053
Var. 1 & 10 -0.31 -0.28 -0.24 -l6.75 0.00 14 34255
Var. 1 & 11 -0.26 -0.23 -0.21 -19.30 0.00 11 34920
Var. 1 & 12 -0.05 0.005 0.14 0.99 0.32‘I 6 3477
Var. 2 & 3 0.20 0.22 0.24 23.29 0.00 2 10852
Var. 2 & 4 -0.23 -0.13 -0.03 -2.61 0.01 2 2589
Var. 2 & 5 0.03 0.18 0.32 2.31 0.02 2 2589
Var. 2 & 6 missing missing missing missing missing 0 O
Var. 2 & 7 0.22 0.32 0.41 6.12 0.00 5 4382
Var. 2 & 8 -0.24 -0. l9 -0. 14 -6.83 0.00 15 9692
Var. 2 & 9 -0.21 -0. 12 -0.04 -2.83 0.01 6 2641
Var. 2 & 10 -0.29 -0.25 -0.22 -13.38 0.00 8 2751
Var. 2 & 11 -0.34 -0.30 -0.26 -13.29 0.00 5 1920
Var. 2 & 12 -0.06 -0.01 0.03 -0.66 0.511‘ 5 3788
Var. 3 & 4 -0.28 -0.09 -0.09 -0.99 0.32‘I 3 2218
Var. 3 & 5 0.16 0.28 0.40 4.38 0.00 9 35486
Var. 3 & 6 0.10 0.26 0.40 3.19 0.00 4 4863
Var. 3 & 7 0.24 0.34 0.44 6.04 0.00 6 5186
Var. 3 & 8 -0.38 -0.27 -0.15 -4.41 0.00 6 6342
Var. 3 & 9 -0.28 -0.26 -0.23 -l7.38 0.00 3 4351
Var. 3 & 10 -0.38 -0.31 -0.24 -8.31 0.00 9 33486
Var. 3 & 11 -0.27 -0.25 -0.23 -22.27 0.00 10 12803
Var. 3 & 12 0.07 0.27 0.48 2.27 0.02 4 5459
Var. 4 & 5 -0.20 -0.12 -0.04 -2.88 0.00 3 4077
Var. 4 & 6 missing missing missing missing missing 0 0
Var. 4 & 7 -0.24 -0.15 -0.05 -3.04 0.00 6 6044
Var. 4 & 8 -0.03 0.12 0.26 1.56 0.12’ 3 3142
Var. 4 & 9 0.05 0.10 0.14 4.14 0.00 2 1790
Var. 4 & 10 -0.36 -0.15 0.07 -l.31 0.19' 4 2285
Var. 4 & 11 -0.28 -0.11 0.07 -1.18 0.24' 4 3896

 

 

77

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Var. 4 & 12 -0.44 -0.22 0.03 -1.70 0.09‘ 4 2387
Var. 5 & 6 0.33 0.36 0.38 25.91 0.00 3 4863
Var. 5 & 7 0.22 0.39 0.54 4.19 0.00 8 7293
Var. 5 & 8 -0.42 -0.33 -0.24 -6.84 0.00 0 8791
Var. 5 & 9 -0.54 -0.36 -0.15 -3.28 0.00 2 3798
Var. 5 & 10 -0.46 -0.34 -0.21 —4.89 0.00 5 30684
Var. 5 & 11 -0.40 -0.37 -0.33 -20.22 0.00 5 31781
Var. 5 & 12 -0.59 -0.04 -0.01 -2.84 0.00 6 7036
Var. 6 & 7 0.32 0.36 0.40 16.61 0.00 3 1419
Var. 6 & 8 -0.40 —0.32 -0.26 -8.41 0.00 3 2458
Var. 6 & 9 -0.30 -0.22 -0.14 -5.23 0.00 4 5523
Var. 6 & 10 -0.36 -0.30 -0.24 -8.96 0.00 7 5700
Var. 6 & 11 -0.32 -0.25 -0.18 -6.62 0.00 4 3936
Var. 6 & 12 0.28 0.22 0.16 7.29 0.00 1 1065
Var. 7 & 8 -0.59 -0.49 -0.38 -7.91 0.00 0 5802
Var. 7 & 9 -0.38 -0.27 -0.15 -4.39 0.00 5 3146
Var. 7 & 10 -0.44 -0.42 -0.39 -27.22 0.00 7 3641
Var. 7 & 11 -0.49 -0.42 -0.33 -9.06 0.00 6 5499
Var. 7 & 12 -0.12 -0.02 0.08 -0.48 0.63‘ 4 2855
Var. 8 & 9 0.71 0.75 0.78 22.82 0.00 1 553
Var. 8 & 10 0.61 0.76 0.85 6.88 0.00 2 110
Var. 8 & 11 0.78 0.84 0.90 12.50 0.00 4 864
Var. 8 & 12 -0.12 -0.07 -0.02 -2.91 000 0 6715
Var. 9 & 10 0.28 0.38 0.47 6.90 0.00 3 2593
Var. 9 & 11 0.28 0.39 0.48 6.61 0.00 3 1970
Var. 9 & 12 -0.16 -0.06 0.03 -1.26 0.21' 3 1790
Var. 10 & 11 0.46 0.51 0.54 19.61 0.00 6 4341
Var. 10 & 12 -0.12 -0.06 -0.00 -2.11 0.04 l 1237
Var. 11 & 12 -0.05 -0.00 0.04 -0.15 0.88a 3 1859

 

 

 

 

 

 

 

 

 

Note: a = correlations not different from zero at 1% significance level.

Var. 1 = Family Attachment.

Var. 2 = Family Control

Var. 3 = Educational Commitment
Var. 4 = Community Trouble
Var. 5 = Self Conventionality
Var. 6 = Parental Conventionality
Var. 7 = Peers’ Conventionality
Var. 8 = General Delinquency
Var. 9 = Index Offense

Var. 10 = Status Offense

Var. 11 = Substance Abuse

Var. 12 = Family SES

 

78

Standardized Regpessions

Using the average correlations, which take into account both sampling
error and population variation, the social-environmental model of delinquency
was examined by conducting seven standardized regressions discussed in
Chapter 3. The independent-dependent relations specified in these regression
models were derived from Sutherland’s differential association, Hirschi’s
social control theories, and the delinquency literature reviewed in Chapter II.
The results of the standardized regressions are presented in Figures 2 to 8.

Because none of the 50 studies examined the relationship between
family control (var. 2) and parental conventionality (var. 6), or community
trouble (var. 4) and parental conventionality (var. 6), values needed to be
imputed for the two estimates of the population correlation coefficients (i.e.,
p26 and p46), as well as for their population variances. The following
describes how the values were determined. An examination of the
relationships between variables 2, 6, and all other variables implied that it
seemed likely that 926 .>. 0. For example, flz’s 95% confidence interval is
[0.17, 0.32], andiw’s 95% confidence interval is [0.22, 0.39]. It seemed
reasonable to assume that if the relationship between variables 2 and 6 was

positive, it was only moderately strong. Thus, it was decided to impute two

79

different values -- 0.00 and 0.30, for the estimated correlation between
parental control and community trouble. For the relationship between variable
4 and variable 6 (i.e., community trouble and parental conventionality), the
same procedure was followed, leading to trial values of 0.00 and -0.30 for
E16.

Realistic values of population variances for 926 and p45 also needed to
be picked. The smaller the number, the higher the precision of estimation is
assumed. Arbitrarily, 0.005 and 0.09 were selected to represent two difierent
conditions -- precise and not so precise. The variance of 0.005 gives a
population standard deviation of 0.07, and 0.09 gives 0.3.

Experimentation with different combinations of the estimated
population correlations and their population variances for p26 and p46 across
the seven regression models led to very similar results. Generally, the
estimations of the standardized betas were within 0.04 of each other.

Because the results of the standardized regressions were so similar,
only one set of numbers is presented. For the presented model, the correlation
between family control and parental conventionality is assumed to be zero

(p26=0), the correlation between community trouble and parental

80

conventionality is also assumed to be zero (p46=0), and the population
variances for these two correlations are assumed to be 0.09.

The first regression model (Figure 2) predicts educational commitment
with family attachment, community trouble, parental conventionality, and
family SES. Educational commitment is best predicted by family SES (0 =
0.33, SD([3) = 0.06), followed by parental conventionality ([3 = 0.27, SD03)
= 0.06), and family attachment ([3 = 0.20, SD(6) = 0.06). For community
trouble predicting educational commitment (0 = 0.00, SD03) = 0.11), the
95% confidence interval of the regression coefiicientfi was 0.00 i- 0.22. This
last estimate implies that the relationship between community trouble and
educational commitment is generally negligible when the effects of family
attachment, parental conventionality, and fam ily SES were taken into
account.

The second regression model (Figure 3) predicts self conventionality
with family attachment, educational commitment, community trouble,
parental conventionality, and peers ’ conventionality. The three significant
predictors of self conventionality were: peers ’ conventionality (j; = 0.25,
SD(/[i) = 0.05), parental conventionality (E = 0.19, SD43) = 0.07), and family

attachment (B = 0.18, SD(§) = 0.05). The effects of educational commitment

81

and community trouble were not significant, with the 95% confidence
intervals of [3 = 0.09 i 0.18 and/l3 = —0.05 i 0.10 respectively.

The third regression model (Figure 4) is with peers’ conventionality
predicted by family attachment, family control, educational commitment, and
community trouble. All four predictors turn out to be significant; educational
commitment was the most important predictor (f3 = 0.26 , SD63) = 0.06),
family control was the second (0 = 0.22, SD63) = .05) and family attachment
was the third ([3: 0.13, so (ii) = .05), while community trouble has a small
but statistically significant relationship with peers ' conventionality. Adding
the predictor community trouble increased the explanatory power of the
model (R2) from 13.1% to 21.3%.

The fourth, fifth, sixth, and seventh regression models (Figures 5, 6, &
7) were with family attachment, educational commitment, self
conventionality, and peers ’ conventionality predicting, respectively, general
delinquency, index oflense, status oflense, and substance abuse. It was found
that general delinquency, status oflense, and substance abuse were better
explained than was index offense. The amount of variation explained in these
models by the four predictors were, respectively, 31.8%, 26.5%, 25.2%, and

18.0%.

82

Adolescents’ self con ventionality toward moral values was a
significant predictor for every type of delinquent measure. Among the
different types of delinquent behavior, self conventionality had the highest
impact on index oflense; the 95% confidence interval of the/0 coefficient is
{-0.36, -0.20]. With self conventionality predicting general delinquency,
status oflense, and substance abuse, thefi coefficients’ 95% confidence
intervals were [-0.21, -0.01], {-0.24, -0.08], and {-0.30, -0.14] respectively.

Peers’ conventionality had significant impacts on all types of
delinquent behavior, except for index oflense. The 95% confidence interval of
the regression coefficients, for peers ’ conventionality predicting general
delinquency, status oflense, and substance abuse, were
{-0.49, -0.41], {-0.36, -0.24], and {-0.36, -0.26] respectively.

Educational commitment had moderate but statistically significant
influence on index offense, status oflense, and substance abuse (95 %
confidence intervals of the regression coefficients were [-.25, -.01], [-.28,
-.02], and [-.31, -.05] respectively), while its influence on general
delinquency was not significant.

Also, family attachment had a small but significant effect on status

oflenses only, after holding adolescents’ educational commitment, self

83

conventionality, and peer conventionality constant. However, family
attachment did not have significant influence on other types of delinquent

behavior.

Parental Conventionality .27 i .12

A
My

Family Attachment .20 i .12 Educational Commitment

L
V

.00 rag? ........ »

 

 

Community Trouble

Figure 2. Prediction of Educational Commitment, R2 = 21.1%.
Note: 1. Statistically insignificant path coefficient (p < 0.05).
2. B i 2* SD(B) represents the 95% confidence interval of a path coefficient.

 

 

   

Peers’ Conventionality .25 i .10
.19 i .14
Parental Conventionality ’
18 i .10
Family Attachment ’ Self Conventionality
09 i .181 ______
Educational Commitment """"""""""""
- 05 i .10.‘ ....... 2'

Community Trouble

Figure 3. Prediction of Self Conventionality, R2 = 29.6%.
Note: 1. Statistically insignificant path coefficient (p < 0.05).
2. B i 2* SD(B) represents the 95% confidence interval of a path coefficient.

Educational Commitment .26 i .12

Family Control 22 i -10 ,

     

Family Attachment .13 iJO

-.08 :1: .06
Community Trouble /'

Peers’ Conventionality

V

Figure 4. Prediction of Peers’ Conventionality, R2 = 21.3%.
Note: [3 i 2* SD(B) represents the 95% confidence interval of a path coefficient.

85

Family Attachment --...,,”'-08 i .121

Educational Commitment -.06 i .121

.............................. _,
Peers’ Conventionality -.45 i .04 > General Delinquency

-.1 1 i .10
Self Conventionality

Figure 5. Prediction of General Delinquency, R2 = 31.8%.
Note: 1. Statistically insignificant path coefficient (p < 0.05).
2. B i 2* SD(B) represents the 95% confidence interval of a path coefficient.

 

Family Attachment ------- .,,_._-..05 i .101

u
a.
a
n
I.
-
c
c.
'.
.
o.
n
e
.
.
e.
-

Educational Commitment _-.13_:L_12_.

Self Conventionality -.28 :t .08 : Index Offenses

40:14} ........ 2

 

Peers’ Conventionality

Figure 6. Prediction of Index Offenses, R2 = 18.0%.
Note: 1. Statistically insignificant path coefiicient (p < 0.05).
2. B i 2* SD(B) represents the 95% confidence interval of a path coefficient.

86

Family Attachment -.10 i .09

\

Educational Commitment -. 14 i .12 :

Self Conventionality -.16 i .08 A

-.30 i .06
Peers’ Conventionality /

Figure 7. Prediction of Status Offenses, R2 = 26.5%.
Note: [3 i 2* SD(B) represents the 95% confidence interval of a path coefficient.

 

Family Attachment ----- .._H-.06 i .091

""" \.
Educational Commitment -.06 21:.11 -

Self Conventionality -.22 i .08 ‘ Substance Abuse

-.31 d: .05
Peers’ Conventionality

Figure 8. Prediction of Substance Abuse, R2 = 25.2%.
Note: 1. Statistically insignificant path coefficient (p < 0.05).
2. B :1: 2* SD(B) represents the 95% confidence interval of a path coefficient.

 

 

‘=|

87

Community Trouble

   
  
 

331.12 Family SES

    
 
 
 
    

Family Control
Parental Conventionality

.13i.10

  

 
 

 

  

 

l. 1 Si. 10
Educational Self A .2521; 10 Peers’
Commitment Conventionality ‘ Conventionality
\ .2621;

 

-.13:.12‘ l-.28i.08‘

 

 

 

Index Offenses

 

 

 

 

 

 

Figure 9. Revised Ecological Model of Index Offenses.

Note: 1. Educational Commitment and Self Conventionality each has a statistically
significant direct impact on Index Ofl’enses (p < 0.05).
2. B j: 2* SD(B) represents the 95% confidence interval of a path coefficient.
3. Statistically insignificant paths are not shown in this path model.

88

Community Trouble

   
  
 

.33i.12 Family SES
Family Control

Parental Conventionality

Family Attachment

ll8i.10

Self A
Conventionality

   
  
  

.13:lz.10

Peers’
Conventionality

\;.10:.09l I -.16i.08‘ -.30i.06l

   
  

 

 

Educational 25:. 10

Commitment

 

V

        

\ .26i.

 

 

 

  

-.14s_.12l

 

 

Status Offenses

 

 

 

 

 

 

Figure 10. Revised Ecological Model of Status Offenses.

Note: 1. Educational Commitment, Family Attachment, Self Conventionality, and Peers’
Conventionality each has a statistically significant direct impact on Index Offenses
(p < 0.05).
2. B i 2* SD(B) represents the 95% confidence interval of a path coefficient.
3. Statistically insignificant paths are not shown in this path model.

89

Community Trouble

   
 

33:12 Family SES

    
  
     
 

Family Control

Parental Conventionality

   
   

Peers’
Conventionality

  
 

Educational
Commitment

 

   

 

I -.22_t.08' -.31:.05'

 

 

Substance Abuse

 

 

 

 

 

 

Figure 11. Revised Ecological Model of Substance Abuse.
Note: 1. Educational Commitment, Self Conventionality, and Peers’ Conventionality
each has a statistically significant direct impacts on Index Ofi‘enses (p < 0.05).
2. B i 2* SD(B) represents the 95% confidence interval of a path coefficient.
3. Statistically insignificant paths are not shown in this path model.

90

Parental Community
Conventionality Trouble

  

 

 

 

 

 

 

Family A Educational
SES 7 Commitment ------------------------------ —> Conventionality
Family _ Peers’
Control ' Conventionahty__
(4)
Family (1)
Attachment """"""""""""""""" " Delinquency

Figure 12. Revised Ecological Model of Delinquency
Note:_, Links derived from Sutherland’s differential association theory
............ _, Links derived from Hirschi’s social control theory
. Links added based on human ecological theory
(1) a significant path for status offenses only
(2) a significant path for status offenses and substance abuse
(3) a significant path for index offenses and status offenses (barely significant)
(4) for index offenses, status offenses, and substance abuse

91

Revised Structu_r§ Model
Figures 9, 10, and 11, the revised models from the original model
(Figure 1), represent the intervening mechanisms that translate socio-
ecological factors into index oflenses, status oflenses, and substance abuse
respectively. The model revisions are based on the integration of the results of
the standardized regressions (presented in Figures 2 through 8); paths that are
statistically insignificant are excluded. These revised ecological models of

different types of adolescent delinquency are summarized in Figure 12.

Gender Differences in Avemge Correlations

For the investigation of gender differences, the original intention was to
test the proposed model separately for both genders, first using a sample of
female-only studies, and then a sample of male-only studies. Unfortunately,
since an insufficient number of studies existed to do either of these analyses,
t-tests were used on all relationships that contained at least 2 male- and 2
female-only studies. Only samples for 9 relationships met these criteria. On
average, none of the relationships under investigation were significantly

different by gender (see, Table 9)

92

In conducting the t-tests, transformations for the standard errors and
degrees of freedoms were necessary because a typical t-test assumes equal
variances in small samples. However, the estimated variances for males and
females among the studies were unequal. The transformations in this analysis

used procedures described in Hays (1973).

Table 9. The random-effect results: Gender differences in average

93

correlations ( for the relationships that allow for gender comparisons only).

 

 

Variables Female Male Gender Differences
Mean r Number Mean r Number Standard t df" p
of studies of studies error
Var. l & 2 0.216 6 0.223 10 0.100 0.150 7 0.885
Var. l & 8 -0.250 9 -0.254 11 0.020 -0.166 14 0.871
Var. l & 10 -0.236 4 -0.254 6 0.040 -0.445 5 0.675
Var. 1 & 11 -0.159 4 -0.113 4 0.072 0.633 1 0.641
Var. 2 & 8 -0.187 6 -0.145 7 0.076 0.559 8 0.592
Var. 2 & 9 -0.077 2 -0.145 4 0.071 -0.992 2 0.426
Var. 2 & 10 -0.236 3 -0.224 5 0.052 0.228 4 0.831
Var. 5 & 8 -0.238 2 -0.259 5 0.064 -0.328 2 0.774
Var. 8 & 12 -0.153 2 -0.169 6 0.047 -0.333 3 0.761

 

Note: dj‘ = adjusted degrees of freedom

Var. 1 = Family Attachment
Var. 2 = Family Control

Var. 5 = Self Conventionality
Var. 8 = General Delinquency

Var. 9 = Index Offenses
Var. 10 = Status Offenses
Var. 11 = Substance Abuse
Var. 12 = Family SES

CHAPTER 5

CONCLUSION, DISCUSSION, AND IMPLICATIONS

Through an examination of several theoretical and empirical works, an
integrative model of delinquency was developed. The integrated model
combined propositions of Hirschi’s control theory, Sutherland’s differential
association theory, and Bronfenbrenner’s human ecological theory. The
proposed interrelations, based on the theories, were tested by analyzing
statistical results from the delinquency literature. This chapter presents (1) a
summary of the principles of the integrated theories and the proposed
hypotheses, (2) a discussion of major findings, and (3) implications for
practice and future research.

Hirschi’s theory emphasizes that adolescents are dissuaded, protected,
and insulated from committing delinquent acts through bonding to
conventional society. In contrast to Hirschi’s assumption of single
conventional moral order, Sutherland’s theory argues that because modern
societies contain conflicting norms, definitions, and patterns of “appropriate”
behavior, an adolescent becomes delinquent because he or she experiences an

environment that “favors” law-breaking. Bronfenbrenner, a developmentalist,

94

95

stresses that (1) human behavior is a result of the interplay between an
individual and the key environments in which he or she is embedded, and (2)
the links between the key environments have developmental significance for
the individuals. Linrited by the nature of the data, this study did not analyze
the nonrecursive relationship between adolescent behavior and ecological
contexts. Instead, the focus of this study is on the links between the key
environments of the adolescents (i.e., the intervening mechanisms) that
translate social-environmental influences into individual behaviors, including
adolescents’ commitment to educational goals, association with delinquent
peers, and involvement in delinquent acts.

This integrated model proposes that adolescent delinquency is largely
the product of attenuated bonding to conventional individuals, activities, and
values. Besides adolescents’ attitudes toward the conventional moral order,
their family attachment, educational commitment, and peers’ conventionality
are viewed as social-environmental factors that have direct impacts on
preventing or encouraging delinquent behavior.

The statistically significant links between key social-environmental
factors that influence adolescent’s educational commitment, peers’
conventionality, and self conventionality are summarized below. First,

adolescents’ educational commitment is influenced by family SES, family

96

attachment, and parental conventionality. Second, self conventionality is
influenced by family attachment, parental conventionality, peers ’
conventionality, and educational commitment. And finally, peers ’
conventionality is influenced by family attachment, family control,
educational commitment, and community trouble. Family attachment, family
control, parental conventionality, family SES and community trouble are
considered as an exogenous (background) variables that are not subject to

prediction in this model.

Discussion of Major Findings

Overall, the results of the meta-analysis provide more support for
Sutherland’s differential association theory than for Hirschi’s social control
theory. It was found that family attachment and educational commitment
(i.e., the components of social bonding) have insignificant to barely
significant relationships with delinquency. In general, the influences of fam ily
attachment and educational commitment are mainly mediated by the
differential learning of definitions favorable to law breaking through
socialization with pro-delinquent peers. This finding is consistent with
previous research results supporting Sutherland’s theory rather than Hirschi’s

(Thompson et al., 1984; Matsueda & Heimer, 1987).

97

Specifically, Sutherland’s argument that family life is relevant to
delinquency only when “delinquent patterns” are available to copy is partially
supported; family attachment itself is not a significant predictor for different
measures of adolescent delinquent behavior (except for a barely significant
relationship between family attachment and status oflenses), while significant
indirect links exist between family attachment and adolescent delinquency
(except for index oflenses) through peer conventionality.

Although Sutherland’s theory is better supported, this study finds
support for both theories when considering the impact of adolescents’ self
conventionality (belief about the conventional moral order) on adolescent
delinquency. Self conventionality is the second most important predictor for
status oflenses, and substance abuse, and first importance in predicting index
offenses.

Another finding of this study is that the zero-order correlations and
standardized regressions offer different answers in the explanation of
delinquency. The correlations between the four predictors of adolescent
delinquency (i.e., family attachment, educational commitment, self
conventionality, and peers ’ conventionality) and different measures of
delinquent behavior (i.e., general delinquency, index oflenses, status

offenses, and substance abuse) are all significantly different from zero. The

98

zero-order correlations support (1) Hirschi’s observation that adolescents
with stronger attachment to the conventional society are more likely to be
dissuaded, protected, and insulated from committing delinquent acts, and (2)
Sutherland’s argument that adolescents’ attitude toward the conventional
moral order (i.e., self conventionality) and their peers ’ conventionality are
related to the adolescents’ delinquent behavior.

However, when the standardized regressions are considered, it seems
that Hirschi’s suggestion that family attachment and educational
commitment, the major components of social bonding, may directly dissuade
adolescents from delinquent acts does not always hold true. On average,
family attachment does not have significant direct impact on delinquent
behavior of any type (except for a barely significant influence on status
oflenses). Educational commitment is not a significant predictor of general
delinquency and substance abuse and only a marginal significant predictor of
index oflenses and status oflenses, when other variables are held constant, as
proposed by Hirschi.

Sutherland’s theory, on the other hand, is better supported;
adolescents’ attitude toward the conventional moral order (i.e., self
conventionality) and their peers ’ conventionality are the strongest predictors

of all types of delinquent behavior with only one exception -- peers’

99

conventionality is not a significant predictor for index oflenses, after
adolescents’ self conventionality and educational commitment are taken into
account.

This implies that for adolescents with equal levels of self
conventionality and educational commitment, their peers ’ conventionality
has significant direct effects on general delinquency, status offenses, and
substance abuse, but NOT on index oflenses. Perhaps because of the
seriousness of index offenses, adolescents with high educational commitment
and high self conventionality do not choose to commit this type of crime even

when they are pressured to do so by their peers.

Prediction of Educational Commitment

The strongest predictor for educational commitment is family SES,
followed by parental conventionality and family attachment; community
trouble is not a significant predictor of educational commitment. Given that
parental conventionality is a measure of parents’ attitudes toward hard work,
education, and unlawful acts, the following interpretation is reached: for
adolescents with similar levels of family attachment and parental
conventionality, family SES has positive efiects on their educational

commitment. In other words, this study shows that low family SES has a

100

significant negative influence on adolescent’s educational commitment,
holding constant fam ily attachment and parental con ventionality (a measure
including parents’ adherence to such values as hard work and education).

The finding of this influence is as expected; an extensive literature
suggests that there is a strong association between the two: the lower the
adolescent’s family SES, the less likely he or she is to succeed or even want
to succeed in school (Lipton & Smith, 1983; Huston, 1991). However, the
results of the standardized regressions seem to disagree with some
researchers’ argument that this relationship is due to the association between
family SES and parents’ attitudes toward education (a component measured
in parental conventionality).

Although there is a positive relationship between family SES and
parental conventionality (r = 0.22, SD(r) = 0.03), after the effect of parental
conventionality is controlled for, the relationship between family SES and
adolescents’ educational commitment still exists. This study did not directly
examine family aspiration and preparation for education and thus cannot rule
out the influence of family aspiration and preparation on the relationship
between family SES and educational commitment. However, from a human
ecological perspective, it is important for future research to consider the

mediating effect between family SES and adolescents’ educational

101

adolescents’ educational commitment still exists. This study did not directly
examine family aspiration and preparation for education and thus cannot rule
out the influence of family aspiration and preparation on the relationship
between family SES and educational commitment. However, from a human
ecological perspective, it is important for future research to consider the
mediating effect between family SES and adolescents’ educational
commitment within a broader context; besides family aspiration and
preparation, factors such as teacher expectation, communication between
family and school, and peer influence need to be taken into account

simultaneously.

Prediction of Peers’ Conventionality

The strongest predictor of peers ’ conventionality is educational
commitment, followed by family control, family attachment and community
trouble, and all the regression coemcients are statistically significant. The
regression coefficients of educational commitment (B = 0.26, SD(B) = 0.06),
family control (B = 0.22, SD(B) = 0.05), family attachment (B = 0.13, SD(B)
= .05, and community trouble (B = -0.08, SD(B) = 0.03) are significantly

different from zero (a = 0.05). Given that the strength of the direct

102

relationship between family attachment and delinquency ranges from non-
existent to barely significant, this finding of an indirect efiect of family
attachment and control on delinquency (through peers’ conventionality),
support Sutherland’s argument that family life is important only when there

are “delinquent patterns” to copy.

Prediction of Self Conventionality

Peers’ conventionality, parental conventionality, and family
attachment are significant predictors of adolescents’ self conventionality (i.e.,
belief about the conventional moral order). Educational commitment and
community trouble, on the other hand, are not significant predictors. This
finding supports Sutherland’s View that the values and attitudes of intimate
personal groups influence adolescent definitions about what “appropriate”
behavior is.

Besides peers’ conventionality and parental conventionality, family
attachment is a significant predictor of adolescents’ attitudes toward
conventional moral order (i.e., self conventionality). One explanation is that
secure family attachment increases the likelihood of adolescent
internalization of the conventional definitions. This is consistent with

Hirschi’s theory stating that the stronger the adolescent’s attachment to

103

Hirschi’ s theory stating that the stronger the adolescent’s attachment to
others, especially parents, the more likely he or she will internalize
conventional social norms.

Overall, this regression model suggests that regardless of the level of
community trouble and the level of adolescents’ educational commitment, an
appeal to family attachment and influence from intimate group (i.e., parental
and peers ’ conventionality) can socialize adolescents toward the direction of

conventional norms.

Prediction of General Delinquency

Perhaps because researchers use more items inquiring about general
delinquency than about specific delinquent acts such as index oflense, status
oflense, or substance abuse, the proposed model explains general
delinquency better than specific delinquent behaviors. Theoretically, the more
frequently the items are asked, the lower the measurement error is. However,
as this study finds that different delinquent behaviors have different
predictors, combining different types of delinquent behavior together is

questionable.

104

Prediction of Index Offenses

Index oflenses are best predicted by adolescents’ levels of educational
commitment and self conventionality (belief toward conventional moral
order). Different from other types of delinquent behavior, index oflenses are
not predicted by peers ’ conventionality; when the effects of adolescents’
educational commitment and self conventionality are taken into account,
differential association with peers who provide definitions favorable to law
breaking does not increase adolescents’ involvement in index oflenses.

In other words, for adolescents with equal amounts of self
conventionality and educational commitment, peers’ conventionality affects
their status oflenses, and substance abuse, but not their index Offenses.
Perhaps because of the seriousness of index oflenses, adolescents’
educational commitment and self conventionality prevent adolescents’
involvement in this type of crime even when they are pressured to do so by
peers. Note, though, that although peer conventionality is not important
enough to have a direct impact on index oflenses, it still has an indirect effect

through self conventionality.

105

Prediction of Status Offenses

All predictors of status oflenses in this model are significant. Family
attachment is a direct predictor of status oflenses, but not of other types of
adolescent delinquency. However, its significance is marginal, and B is small
(-0. 10).

The influence of peers ’ conventionality on status oflenses is stronger
than that of the social bonding factors --family attachment, educational
commitment, and self-conventionality, after the influences of these social
bonding factors are taken into account. In other words, when adolescents’
family attachment, educational commitment, and self-conventionality are
held constant, peers ’ conventionality still has a strong negative effect on

status oflenses.

Prediction of Substance Abuse

The literature has long demonstrated a relationship between differential
association and adolescent substance abuse. Consistent with previous
research (e.g., Marcos et a1. 1986), this study finds that peers ’
conventionality is the strongest predictor of adolescent substance abuse,
holding constant other variables. Besides peers ’ conventionality, adolescents’

self conventionality is a significant determinant of substance abuse. Family

106

attachment and educational commitment, on the other hand are not
significant predictors. This result implies adolescents’ self conventionality is
more strongly and directly related to substance abuse than family attachment

and educational commitment.

Indirect Effects of Family Attachment

Family attachment, on average, is not a significant predictor of
delinquent behavior of any type, except for status oflenses (the 95 %
confidence interval for the regression coefficient is [ -0. 15, .05] for index
ofi’enses, [-0. 19, -.01] for status oflenses, and [-0.l6, .04] for substance
abuse). However, one should not conclude that family attachment has no
influence on adolescent delinquency. The importance of the indirect effect of
family attachment on delinquency needs to be emphasized; family attachment
is a significant predictor of adolescents’ educational commitment, self
conventionality, and peers ’ conventionality -- the three variables that have
significant direct impacts on delinquency. If an adolescent has higher family
attachment, he or she is more likely to have higher educational commitment,
to associate with conventional peers, and to adhere to conventional values,

and thus is less likely to be delinquent.

107

The following discussion about the predictions of the three variables --
educational commitment, self conventionality, and peers ’ conventionality --
that have significant impacts on delinquency is organized around the results of
the three standardized regressions (see Figures 2, 3, & 4). The first regression
predicted educational commitment with family attachment, community
trouble, parental conventionality, and family SES. The second regression
predicted selfconventionality with family attachment, educational
commitment, parental conventionality, and peers ’ conventionality. The third
regression predicted peers ’ conventionality with family attachment, family

control, educational commitment, and community trouble.

Indirect Effects of Family SES

The role of adolescents’ educational commitment as an intervening
variable between family SES and delinquency is also supported by this study.
A positive path coefficient exists between family SES and adolescents’
educational commitment, which is negatively associated with both index and
status oflenses. As (1) the relationship between family SES (var A) and
adolescents’ educational commitment (var B) is significant and moderate (2)

the relationship between educational commitment (var. B) and delinquency

108

,A
(var. C) is significant, although small (B = -0.13 for index ofi‘enses, -0.14 for
A
status oflenses; SD(B) = 0.12 for both).

The low interrelationship between family SES and delinquency,
extensively observed in the literature, is explainable. As the path goes from
variable A, to B, then to C, and no direct relationship between A and C, the
zero-order correlation between A and C is theoretically the product of the
correlation between A and B and the correlation between B and C; in other
words, rAc = rAB * me. The actual rAc (using substance abuse for example),
empirically measured in the range between -0.05 to 0.04, is close to the
theoretical computation: rAc = rAB * ch = (0.27 :1: 0.21) * (-0.25 i 0.02),
which is in the range between -0.13 and 0.00.

Therefore, the small relationship between family SES and delinquency
should not be considered as an error or a source of confusion as generally
viewed in the literature. As Tittle et al. (197 9) describe in their work,

...there does seem to be an empirical relationship between
class origin and academic performance in high school. There

also seems to be a consistent and strong association between

academic performance and delinquency... Therefore, it should

follow that there would be a strong class origin/delinquency
association, but of course, our paper shows that in general such
relationship has not been demonstrated. Either the

origin/performance or the performance/ delinquency association
is in error... (p. 670)

109

The weak association between family SES and delinquency is found in
this study and in many other studies. Traditional narrative reviews have stated
that a considerable amount of studies have found a slight negative to
nonexistent relationship between SES and delinquency (e.g., Tittle et al.,

197 9; Larzelere & Patterson, 1990). In a meta-analysis (bivariate), Loeber
and Dishion (1983) found that SES is the weakest predictor compared to the
other predictors of delinquency.

Findings of a weak association between family SES and delinquency,
however, are not in error. Instead, they suggest the existence of a third
variable serving as a link between the two. As adolescents’ educational
commitment is related to both family SES and delinquency, it seemed logical
to treat educational commitment as the mediating link between family SES

and delinquency.

Co-occurance of Delinquent Behaviors

Significant and strong to modest zero-order correlations were found
between status oflenses and substance abuse (r = 0.50, SD(r) = 0.02),
between index oflenses and substance abuse (r = 0.39, SD(r) = 0.05), and

between status oflenses and index oflenses (r = 0.37, SD(r) = 0.05). These

110

study results show a significant association between different types of
delinquent behavior, commonly found in the literature, suggesting that a
similar explanation exists to account for the varied behaviors (Dryfoos, 1990;
Lerman & Pottick, 1995). However, as mentioned previously, this study
shows that similar but not identical predictors exist for different types of

delinquent behaviors.

Gender Differences

Most studies included in this meta-analysis that addressed female
delinquency focused primarily on family factors and ignored possible peer,
school, and community influences. The primacy of the family context in
explaining female delinquency is perhaps due to the traditional view that
females are more likely than males to be influenced by family factors such as
attachment and control.

Although only a limited number of bivariate relationships could be
investigated, the null hypotheses that there would be no differences in the
strengths of the relationships that depended on gender were not rejected. One
possible explanation for this finding is that the t-tests conducted had fairly
low power because there was population variation within the male and female

samples, and the sample sizes (i.e., the numbers of studies that looked at

111

males and females separately) were often small. However, an examination of
the mean correlations of the male and female samples indicate that differences
between the genders on these relationships were substantively small.

In conclusion, because not enough studies had looked at males and
females separately, testing the utility of the proposed model separately for the
two genders was not possible at this time. Whether specific theories of
delinquency with respect to gender, ethnicity, social status, and geographic

residence are needed may well be a subject for future research.

Strengths and Limitations of the Study
The strengths of this study include the combination of theory

integration, quantitative research synthesis, and multiple regression approach.
First, unhke studies which focus on only one theory (e.g., Gardner &
Shoemaker, 1989), this study compared, contrasted, and finally integrated the
major theories of delinquency. Second, this integrated model was tested using
meta-analytical techniques. These techniques allow for both specific
quantitative estimates of relationships and tests of whether the strength of the
relationships differ across the diversity of ecological contexts the primary

studies tend to focus on separately.

112

Finally, unlike previous meta-analyses (e. g., Loeber & Dishion, 1983),
this study did not address only the bivariate relationships. The standardized
regression approach which was used allowed partial relationships to be
estimated. For example, while the mean bivariate correlation between fam ily
attachment and different types of delinquency was about 0.25, it was found in
the specific path model tested that the direct partial effect of family
attachment on delinquent behaviors was generally not substantially different
from 0. Therefore, unlike the previous meta-analyses, the results of this study
indicate that, while family attachment has significant and substantial
relationships with factors that have causal influences on delinquency, the
direct relationships between family attachment and various measures of
delinquent behavior seem to be nearly 0. Thus this approach allows for testing
of both direct and indirect effects while holding important variables constant.
This methodological strategy is unique to this study, differentiating it fiom
other investigations which address delinquency.

However, this meta-analysis has three major limitations. First, while
measurement error exists in the primary studies (i.e., many of the instruments
had imperfect reliability), few studies reported the reliability of the
instruments they used. Thus, measurement error could not be accounted for

statistically. Second, specific cross-population model comparisons could not

113

be made due to “missing data.” For instance, the conceptual model could not
be estimated separately for boys and girls; instead, only differences in the
strength of certain bivariate relationships could be compared. Third, the vast
amount of “missing data” led to an inability to employ fixed-effects models
which could investigate moderator variable effects. The only option remaining
was to attribute surplus variance (i.e., variance over and above that expected
due to sampling error) to variation in the population correlations.

This study synthesized the statistical results of primary studies that rely
on the self-reports of adolescent respondents, and thus depends on
adolescents’ willingness to share their experiences regarding environments,
social relations, and delinquent behaviors. Not all adolescents might have
wanted to share this information, due to desires for privacy, or fears of
disclosure. When self reports are less than accurate, potential for
measurement errors and biases exist. To control statistically for such bias, one
needs the information about reliability of the measurement scales. Because
only a small proportion of the primary studies included in this meta-analysis
reported this information, the results of this meta-analysis are subject to the
influence of measurement errors.

Specific model comparisons across sub-populations are not available.

If there were enough studies (more than 2) that focused on a specific

114

adolescent sub-population (e.g., boys or girls, whites or nonwhites) for each
of the possible intercorrelations studied in this meta-analysis, then it would be
possible to test whether the proposed model fits the data for different sub-
populations, or to compare the magnitude of the intercorrelations and path
coefficients across the sub-populations. Unfortunately, sufficient data on all
correlations within each sub-populations were not available. Partial
comparisons were done based on the data available; only gender differences
in some relationships between family factors and delinquency can be
examined, because family factors seem to be the only variables commonly
focused on in the primary studies examining female delinquency.

As the hypotheses of homogeneity were rejected, suggesting the
existence of variation in the population correlations in the studies subject to
synthesis, one might elect to employ the fixed-effects model to use sample or
study characteristics, as moderators, to explain the variation. However,
because not enough studies had studied the relationships using samples from
sub-populations (e.g., males vs. females, whites vs. non-whites, rural vs.
urban), this meta-analysis necessarily employed a random-effects model
instead, which assumed a normal distribution of population parameters in

order to explain surplus variation.

115

Implications

Implications for Practice

According to Henggeler and Borduin (1990), adolescent delinquency
prevention and treatment programs based on broad human ecological
perspectives are more appropriate and effective than most existing programs,
which focus only on the individual adolescent in one context. In order to
make the existing programs more effective, practitioners who work with
adolescents at risk should consider simultaneously the eflects of family,
peers, school, and community.

In general, current etiologic theories of delinquency fail to consider a
diversity of causes of adolescent delinquency. As a consequence, the theories
fail to drive the prevention and treatment programs effectively. Most
prevention and treatment programs have been influenced by the narrower
perspective of non-ecological theories; trained in reduction, those who work
clinically with delinquents tend to dismiss too quickly diverse, multiple
factors.

Reported by Kotlowitz (1991), the following case study illustrates the
importance of using a human ecological approach to understand the needs of

adolescents at risk.

116

Two boys, Pharoah and Lafayette, and their mother live in an inner-city
ghetto, struggling for a better life. There are no banks, hospitals, or public
libraries where they live. The newborn morality rate is higher than that of the
third world. There are frequent gun shootings, and adults as well as children
are very uncertain about their future.

The boys’ mother believed that education would offer some promise of
a better future and decided to spend more time with her two youngest
children in order to give them support and advice. She felt she had made
mistakes with her older children, and vowed not to repeat these mistakes of
being too busy to supervise her children’s progress in school and their
whereabouts after school. Her older children had ended up in jail, for which
she blames herself.

Pharoah and Lafayette each had strong ties to their mother. They
depended on her; she was responsible for them. She was a caring mother and
a soft-spoken person known for her warmth and generosity not only to her
children, but also to her children’s fiiends, because she knew that most
families were facing the same problems. Pharoah and Lafayette’s father,
however, had let them down. Drugs had ruined his life and destroyed his
relationship with his wife and sons. Ifthey had had his full income and

personal support, they would have moved out of the community.

117

The community had become a ghetto of poor families, mothers and
children without fathers. An estimated 85 percent of the households were
headed by women. The decay of the community was attributed to the
structural changes in the economy, the loss of manufacturing jobs, and the
transient population. Most people’s lives in this community were influenced
by drugs and drug-related violence. Gangs seemed to control the community
and rival drug gangs frequently had shoot outs. These gangs also recruited
young children to work for them. Many children in the community joined
gangs, sold drugs, and were sometimes involved in homicides.

The fact that their mother valued education helped Lafayette and
especially Pharoah do well in school. Pharoah studied very hard and placed
near the top of his class. His teacher liked him very much and he appreciated
the rewards of school. Lafayette, on the other hand, did not like school as
much as Pharoah did, although he was also a smart adolescent. Later, a good
relationship with a new teacher helped him take school more seriously, but he
had already been held back one year because of a D average, had a poor
attendance record, and secretly wished his mother would push him more.

As a strategy to prevent being pulled into gangs, Lafayette decided to
have as few fiiends as possible. However, as he grew older he started to

change, and began hanging out with delinquent peers.

118

Lafayette’ 8 mother felt he was slipping away, but she did not know
what to do. One day, Lafayette broke the law because of his delinquent peers’
insistence. Had someone looked at Lafayette’s relationships with his parents,
teachers, and fiiends somehow, and supported his initial efforts against
delinquent peer pressure and the community’s criminal subcultures, Lafayette
might not have broken the law. Had the school system provided more
opportunities for the development of commitment to conventional goals,
instead of focusing on his poor grades, Lafayette might have liked school
more and not played truant as much.

Perhaps, Lafayette’ s developmental trajectory would have been toward
a healthier direction, if he had taken school more seriously either because his
mother had pushed him harder or his relationships with school teachers had
improved earlier . . . if his father had not been addicted to drugs and given
more attention to his family helping them to move out of the ghetto . . . if the
families in the community had joined together to counter the criminal forces

. . . if these poor families had not been ghettoized . . . if . . .

119

Recommendations

The results of this present study suggest the need for certain actions
and continued research. Below are some suggestions for practitioners, based
on the findings of this study based on an overall 68,217 cases.

First, the family, as a source of social bonding and differential
socialization, should not be ignored as an important component of effective
prevention and treatment pro gram. This suggestion is generated by the
findings that family factors (including fam ily attachment, family control, and
parental conventionality) affect (1) the likelihood of an adolescent’s
association with delinquent peers (peers ’ conventionality), which then has a
direct impact on delinquency and (2) an adolescent’s attitudes toward
conventional moral order (self conventionality). Both paths have the tendency
of socializing an adolescent towards pro- or anti-delinquent behaviors.

Second, adolescent delinquency prevention and treatment programs
need to be more actively involved in enhancing the community’s potential
power to influence its adolescents to reject delinquency. Although community
trouble does not have a direct effect on an adolescent’s behavior, it increases
adolescent delinquent behavior by increasing the likelihood of association

with delinquent peers (peers ’ conventionality). In other words, certain

120

community characteristics (e.g., availability of drugs and acceptance of
delinquent behaviors) may influence adolescents’ peer association.

Third, to assist adolescents to reject delinquent behavior, family and
school need to work together to make sure adolescents have many
opportunities to develop commitment to conventional goals (e.g., educational
commitment). Perhaps a broader definition of success in school, instead of the
narrow focus on academic performance, may help adolescents to build their
social bonds to the conventional society. The connection between family and
school is important, because this study shows that family attachment
influences adolescent delinquent behavior indirectly through its impact on
adolescents’ educational commitment.

Finally, this study’s results suggest (based on the prediction of self
conventionality) that regardless of the level of community trouble and the
level of adolescents’ educational comnritrnent, an appeal to the importance of

family attachment and influence from intimate groups (i.e., parents ’ and
peer ’s conventionality) can socialize adolescents toward the direction of
conventional norms. This finding lends credence to the most recent trend of
intervention programs in schools, which target the student body as a whole

rather than particular subsets of students (e.g., drug-using students).

121

Implications for Future Research

The integration of social control and differential association theories
has taken different forms; several models have been generated, with different
interpretations of the two theories. In general, these models agree that: (1)
association with delinquent peers is a proximate cause of delinquency, and (2)
the formation of social bonding (e. g., family attachment and educational
commitment) is generally “set” temporally prior to differential association
with peers. Controversy exists, however, regarding whether social bonding,
an element crucial to social control theories, has direct impacts on adolescent
delinquency.

Some assert that social bonding has only an indirect impact on
delinquency through association with delinquent peers (Elliott et al., 1985);
weakening or breaking the social bonds does not cause adolescents to commit
delinquent acts, but simply makes them more vulnerable, or susceptible to the
influence of delinquent peers. With a national probability sample of 1,725
adolescents, Elliott and his colleagues found empirical support for their
model; the bonding constructs (family and school involvement) had only
indirect effects on delinquency, through association with delinquent peers, as

they had hypothesized.

122

However, unlike Elliott et al. (1985), Massey and Krohn (1986) found
lack of educational commitment to be a direct predictor of delinquency, in
addition to previous delinquent behavior and association with delinquent
peers, when they used panel data from 1065 adolescents to test a variant of
Elliott and associates’ (1985) model. Massey and Krohn (1986) thus claimed
that their finding confirmed Hirschi’s argument that the impact of the social
bond is not fully mediated by differential association.

Partially consistent with the previous models (e. g., Elliott et al., 1985;
Massey & Krohn, 1986), this study confirmed both direct and indirect effects
of social bonding on adolescent delinquency. Educational commitment had a
direct impact on index oflenses, status oflenses, and substance abuse. Also,
self conventionality prevented all of these three types of adolescent
delinquent behavior directly. Family attachment, on the other hand, dissuaded
adolescents from delinquent behaviors mostly by helping them internalize
conventional social norms, reject influences of delinquent peers, and build
educational comrrritrrrent; the indirect impact of family attachment on
delinquency was mediated by self conventionality, peers ’ conventionality,
and educational commitment. Educational commitment also prevented
adolescents’ delinquent behaviors indirectly, by influencing their association

with delinquent peers (peers ’ conventionality).

123

However, the ecological model of delinquency proposed in this study
differs from other models in several ways. First, adolescents’ belief in
conventional moral order (i.e., self conventionality), considered as a bonding
construct in the social control theory, is shown to be a proximate prevention
of delinquency. Second, differential association theory is not confirmed for
index offenses; association with delinquent peers does not have direct impact
on index oflenses. In this model, the two proximate predictors of index
oflenses are adolescents’ belief in the conventional moral order and their
commitment to conventional goals -- that is, education.

One of the biggest differences between the model studied here and
others is that “past delinquent behavior” was not a predictor of interest. The
focus of this model is on the social-environmental influences on adolescent
delinquency. The goal is to find out why some adolescents become
delinquents while others don’t, rather than to confirm that they are
delinquents because they were delinquents in the past. Also, different studies
define “past” differently and thus prohibit the synthesis of their results
regarding this variable.

Not using adolescents’ past delinquent behavior as a predictor in this
model is not to disagree with the argument of continuity in individual

behavior, but using past delinquent behavior as a predictor requires a more

124

sophisticated study design. Elliott and associates’ (Elliott etal., 1985) and
Massey and Krohn’s (1986) models included adolescents’ past delinquent
behavior as a predictor and thus brought their models more explanatory
power. However, their designs suffered from a serious logical flaw:
delinquency at time one (i.e., past history of delinquent acts) was proposed to
be independent of every other variable in their models, including the social
factors measured at time one that are proposed to predict delinquent behavior
at time two. Their assumption of the independence between the social factors
and delinquency in time one is problematic, and thus the accuracy of their
model is in question, even though the inclusion of past delinquency in their

model is a sound idea.

Recommendations
Based on this study’s results, some suggestions are generated for future
empirical work on delinquency. Potential benefits exist for both primary
studies and meta-analysis, if the following points are taken into consideration.
First, the human ecological perspective is necessary. Influences fi'om
different ecological contexts need to be considered in order to rule out
possible confounding effects. Simple identification of the influences of factors

in limited ecological contexts is insufficient; factors from different contexts

125

are related to each other and their interrelationships and translation
mechanisms have important implications for intervention programs of
delinquency.

Second, scale reliability needs to be measured and reported.
Interrelations and path coefiicients tend to be attenuated by measurement
errors, and to correct for attenuation, one needs the information about
reliabilities of the measurement scales (Hunter & Schmidt, 1990). It also
cannot be assumed that there are no differences in reliability across sub-
groups of the subjects (e.g., gender, race, age); separate reliability
information needs to be generated for each sample studied.

Third, to make future efforts in research synthesis more efficient,
researchers of primary studies need to report zero-order correlations between
the variables measured. Preferably, delinquency will be measured as a
continuous variable. Also, the conceptualization of different constructs needs
to be explicit. As this study shows, model driven meta-analysis is another
option, besides traditional primary studies, for studying social phenomena.
However, the success of meta-analysis depends on the thorough reporting of
the statistical results of the primary studies (e.g., reliability information, zero-
order correlations for full sample, and for sub-groups as well if sub-group

differences exist.)

126

Fourth, this study’s results suggest that different explanations exist for
different types of delinquent behavior. Thus, future efforts in understanding
the etiology of delinquency need to measure and examine different types of
delinquency separately. Researchers thus need to come to a general
consensus on how to differentiate different types of delinquent behavior, so
that their empirical results may lend themselves to meta-analyses.

Finally, cross-cultural examination of the usefuhress of this social-
environmental model of delinquency is desirable. Adolescents living in
another culture might have different socialization patterns and different value
systems. Adolescents in Taiwan for example, are socialized differently from
adolescents in the US. and are subject to influences of a different set of
social-environmental factors. Cross-cultural tests of this social-environmental
model might contribute to confirmation of the model’s universality or provide
contradictions that might be used to develop a more sophisticated model of

delinquency.

APPENDICES

APPENDIX A

127

Appendix A
Studies Included in this Meta-Analysis
Published Reports:

Astone, N. M. & McLanahan, S. S. (1991). Family structure, parental
practices and high school completion. American Sociological Review,
56, 309-320.

Canter, R. J. (1982). Family correlates of male and female delinquency.
Criminology, 20, 149-167.

Cemkovich, S. (197 8). Evaluating two models of delinquency causation:
Structural theory and control theory. Criminology, 16, 335-3 52

Cemkovich, S. & Giordano, P. C. (1987). Family relationships and
delinquency. Criminology, 25, 295-319.

Gove, W. R. & Crutchfield, R. D. (1982). The family and juvenile
delinquency. The Sociological Quarterly, 23: 301-319.

Gray-Ray, P. & Ray, M. C., (1990). Juvenile delinquency in the black
community. Youth and Society, 22(1), 67-84.

Hill, G. D. & Atkinson, M. (1988). Gender, family control, and delinquency.
Criminology, 26(1), 127-149.

Jensen, G. F., 1972. Parents, peers, and delinquent action: A test of the
differential association perspective. American Journal of Sociology,
78, 562-575.

J ohnstone, J. W. C. (1978). Juvenile delinquency and the family: A contextual
interpretation. Youth and Society, 9, 299-313.

Krohn, M. D. & Massey, J. L. (1980). Social control and delinquent
behavior: An examination of the elements of the social bond. The
Sociological Quarterly, 21: 529-543.

128

Krohn, M. D., Stern, S. B., Thomberry, T. P., & Jang, S. J. (1992). The
measurement of family process variables: The effect of adolescent and
parent perceptions of family life on delinquent behavior. Journal of
Quantitative Criminology, 8(3), 287-315.

Larzelere, R. E. & Patterson, G. R. (1990). Parental management: Mediator
of the effect of socioeconomic status on early delinquency.
Criminology, 28(2), 301-324.

Lempers, J. D., Clark-Lempers, D., & Simons, R. L. (1989). Economic
hardship, parenting, and distress in adolescence. Child Development,
60: 25-39.

Lipton, W. L. & Smith, M. D. (1983). Explaining delinquent involvement: A
consideration of suppresser effects. Journal of Research in Crime and
Delinquency, 20, 199-213.

Massey, J. L. & Krohn, M. D. (1986). A longitudinal examination of an
integrated social process model of deviant behavior. Social Forces,
65(1), 107-134.

Matsueda, R. L. & Heimer, K. (1987) Race, family structure, and
delinquency: A test of differential association and social control
theories. American Sociological Review, 52, 826- 840.

Mitchell, J ., Richard, A. D., & Terry, D. N. (1990). Neutralization and
delinquency: A comparison by sex and ethnicity. Adolescence, 98,
487-497 .

Naughton, M. J. (1988). Assessments of the home environment and early
adolescents; intentions to smoke cigarettes. Journal of Early
Adolescence, 8(2), 169-182.

Segrave, J. O. & Hastad, D. N. (1983). Evaluating structural and control
models of delinquency causation: A replication and extension. Youth
and Society, 14(4), 437-456.

Sirncha-Fagan, O. & Schwartz, J. E. (1986). Neighborhood and delinquency:
An assessment of contextual effects. Criminology, 24, 667-703.

129

Simons, R. L., Robertson, J. F ., & Downs, W. R. (1989). The nature of the
association between parental rejection and delinquent behavior.
Journal of Youth & Adolescence, 18, 297-310.

Steinberg, L., Lambom, S. D., Darling, N., Mounts, N. S., & Dombush, S.
M. (1994). Over-time changes in adjustment and competence among

adolescents from authoritative, authoritarian, indulgent, & neglectful
families. Child Development, 65, 754-770.

Thompson, W. E., Mitchell, J., & Dodder, R. A. ( 1984) An empirical test of
Hirschi's control theory of delinquency. Deviant Behavior, 5, 11-22.

Thornton, W. E. (1982). Gender traits and delinquency involvement of boys
and girls. Adolescence, 17, 749-768.

Van Voorhis, P., Cullen, F. T., Mathers, R. A., & Garner, C. C. (1988). The
impact of family structure and quality on delinquency: A comparative
assessment of structural and fimctional factors. Criminology, 26(2),
235-261.

Unpublished Theses and Dissertations:

Cashwell, C. S. (1994). Family influences on adolescent delinquent
behavior: An integrated model. Unpublished doctoral dissertation,
The University of North Carolina at Greensboro.

Cattarello, A. M. (1993). Neighborhood influences on adolescent school
bonds. Unpublished doctoral dissertation, University of Kentucky.

Esteves, A. (1995). Relationships among children’s perceptions of family
management practices, family bonding, school grades, and
absenteeism. Unpublished masters thesis, California State University,
Long Beach.

Hayrnan, J. M. (1977). Attachment to school and student involvement in
delinquency. Unpublished doctoral dissertation, Bowling Green State
University.

130

Metzler, C. W. (1990). The covariance of adolescent problem behaviors,
and the family and peer contexts in which they occur. Unpublished
doctoral dissertation, University of Oregon.

Marcos, A. C. (1985). Causal models of adolescent drug use in Arizona and
Utah. Unpublished doctoral dissertation, Brigham Young University.

Peeples, F. (1991). Working against the odds: Parents, neighborhoods, and

juvenile delinquency. Unpublished doctoral dissertation, University of
Pittsburgh.

Vazsonyi, A. T. (1993). Interpersonal and intrapersonal variables
predicting early adolescent substance use: A risk factor model.
Unpublished doctoral dissertation, The University of Arizona.

Villa, H. S. (1992). Risk factors and their relationship to substance abuse for
seventh and eleventh grade students: Implications for education.
Unpublished doctoral dissertation, The University of San Francisco.

Walls, C. H. (1994). Explaining adolescent alcohol use: An empirical test of
social control and labeling theory. Unpublished masters thesis,
Mississippi State University.

APPENDIX B

131

Appendix B
Coding Frame

Study ID:
Author:
Source:
Data Entry Decision:
Reason for Rejection:
Sample Characteristics:
Sample size:
Sampling procedure:
Age of respondents:
Gender:
Mixed gender?
Number of males:
Number of females:
Ethnicity:
Geographic area:
Country for sample:
Correlation coeflicient information:

132

BIBLIOGRAPHY

Adler, F. (1975). Sisters in crime: The rise of the new female criminals. New
York: McGraw-Hill.

Adler, F. (1980). The interaction between women’s emancipation and female
criminality: A cross-cultural perspective. In S. K. Datesman & F. R.
Scarpitti (Eds), Women, crime, and justice. New York: Oxford
University Press.

Akers, R. L., Krohn, M. D., Lanza-Kaduce, L., & Radosevich, M. ( 1979).
Social learning and deviant behavior: A specific test of a general
theory. American Sociological Review, 44, 636-655.

Alexander, K. L., Cook, M., & McDill, E. L. (1978). Curriculum tracking and
educational stratification: Some further evidence. American
Sociological Review, 43, 47-66.

Anderson, M. L. (1988). Thinking about women: Sociological perspectives
on sex and gender. New York: Macmillian Publishing Company.

Bartollas, C. (1993). Juvenile delinquency. New York: Macrrrillian
Publishing.

Becker, B. J. (1992). Using results from replicated studies to estimate linear
models. Journal of Educational Statistics, 17(4), 341-362.

Becker, B. J. & Schram, C. M. (1994). Examining explanatory models

through research synthesis. In H. Cooper & L. V. Hedges (Eds), The
handbook of research synthesis. New York: Russell Sage Foundation.

Becker, B. J. & Fahrbach, K. R. (1994). A comparison of approaches to the
synthesis of correlation matrices. Paper presented at the annual

meeting of the American Educational Research Association, New
Orleans, LA.

Bronfenbrenner, U. (1977). The ecology of human development: Experiments
by nature and design. Cambridge, MA: Harvard University Press.

133

Bronfenbrenner, U. (1986). Ecology of the family as a context for human

development: Research perspectives. Developmental Psychology, 22,
7 23-7 42.

Bronfenbrenner, U. (1989). Ecological systems theory. In R. Vasta (Ed), Six
theories of child development. Annals of child development (Vol. 6,
pp. 187-249). Greenwich, CT: JAI press.

Bronfenbrenner, U. (1993). The ecology of cognitive development: Research

models and fugitive findings. In R. H. Wozinak & K Fisher (Eds),
Scientific environment. Hillsdale, NJ: Erlbaurn.

Bronfenbrenner, U. (1994). Ecological models of human development. In T.
Husen & T. N. Postlethwaite (Eds), International encyclopedia of
education. New York: Elsevier.

Bronfenbrenner, U., Kessel, F., Kessen, W., & White, S. (1986). Toward a
critical social history of development psychology: A propaedeutic
discussion. American Psychologist, 41, 1218-1230.

Bubolz, M. M. & Sontag, M. S. (1993). Human ecology theory. In P. G.
Boss, W. J. Doherty, R. LaRoss, W. R. Schumm, & S. K. Steinmetz
(Eds), Sourcebook of family theories and methods. New York:
Plenum Press.

Cattarello, A. M. (1993). Neighborhood influences on adolescent social
bonds, peer associations, and drug use: A multi-level analysis.
Unpublished Dissertation, University of Kentucky.

Cemkovich, S. & Giordano, P. C. (1979). A comparative analysis of male
and female delinquency. Sociological Quarterly, 20: 131-145.

Cemkovich, S. & Giordano, P. C. (1987). Family relationships and
delinquency. Criminology, 25, 295-321.

Cemkovich, S. & Giordano, P. C. (1992). School bonding, race, and
delinquency. Criminology, 35: 5-29.

134

Chesney-Lind, M. & Shelden, R. G. (1992). Girls, delinquency, and juvenile
justice. Pacific Grove, CA: Brooks/Cole Publishing Company.

Cohen, J. (1986). Research on criminal careers: Individual frequency rates
and offense seriousness. In A. Blumstein, J. Cohen, J. A. Roth, & C.
A. Visher (Eds), Criminal careers and “career criminals” (Vol. 1,
pp.292-449). Washington, DC: National Academy Press.

Conger, R. D. (197 6). Social control and social learning models of delinquent
behavior: A synthesis. Criminology, 14(1), 17—40.

Cooper, H. M. (1989). Integrating research: A guide for literature reviews.
Newbury Park, CA: Sage Publications.

Datesman, S. K. & Scarpitti, F. R. (1980). The extent and nature of female
crime. In S. K. Datesman & F. R. Scarpitti (Eds), Women, crime, and
justice. New York: Oxford University Press.

Datesman, S. K., Scarpitti, F. R., Stephenson, R. M. (1975). Female
delinquency: An application of self and opportunity theories. Journal
of Research in Crime and Delinquency, 12: 107-123.

Deschenes, E. P., Rosenbaum, J ., & Fagan, J. (1990). Gender diflerences in
delinquency and drug use. Paper presented at the annual meeting of
the American Society of Criminology, Baltimore, MA.

Dryfoos, J. G. (1990). Adolescents at risk: prevalence and prevention.
New York: Oxford University Press.

Eccles, J ., Adler, T. F ., Futterman, R., Goff, S., Kaczala, C., Meece, J. L., &
Midgley, C. (1983). Expectancies, values, and academic behaviors. In
J. T. Spence (Ed.), Achievement and achievement motives:
Psychological and sociological approaches. San Francisco: Freeman.

Elliott, D. S., Ageton, S. S. & Canter, R. J. (1979) An integrated theoretical
perspective on delinquent behavior. Journal of Research in Crime and
Delinquency, 16, 3-27.

135

Elliott, D. S., Huizinga, D. & Ageton, S. S. (1985). Explaining delinquency
and drug use. Beverly Hills, CA: Sage Publications.

Elliott, D. S., Huizinga, D. & Menard, S. (1989). Multiple problem youth:
Delinquency, substance use, and mental health problems. New York,
NY: Springer-Verlag.

Farnworth, M. (1984). Family structure, family attributes, and delinquency in
a sample of low-income, minority males and females. Journal of Youth
and Adolescence, 13, 349-364.

F arrington, D. P. (1993). Have any individual, family, or neighborhood
influences on ofl‘ending been demonstrated conclusively? In D. P.
Farrington, R. J. Sampson, and P. H. Wikstrom (Eds), Integrating
individual and ecological aspects of crime (pp. 7 -3 7).
Brottsforebyggande Radet, Sweden: National Council for Crime
Prevention.

F igueira-McDonough, J. (1983). On the usefulness of Merton’s Anomie
theory: Academic failure and deviance among high school students.
Youth and Society, 14, 259-342.

Figueira-McDonough, J. (1984). Feminism and delinquency: In search of an
elusive link. British Journal of Criminology, 24, 325-342.

Figueira-McDonough, J. (1986). School context, gender, and delinquency.
Journal of Youth and Adolescence, 15, 79-98.

Fisher, R. A. (1928). Statistical methods for research workers. London:
Oliver & Boyd.

Friedman. J. & Rosenbaum, D. P. (1988). Social control theory: The salience

of components by age, gender, and type of crime. Journal of
Quantitative Criminology, 4, 363-381.

Giordano, P. C. & Cemkovich, S. A. (1979). On complicating the
relationship between liberation and democracy. Social Problems, 26,
467-481.

136

Giordano, P. C., Cemkovich, S. A., & Paug, M. D. (1986). Friendship and
delinquency. American Journal of Sociology, 91 , 1 170-1202.

Gleser, L. J. & Olkin, I. (1994). Stochastically dependent effect sizes. In C.
Harris & V. H. Larry (Eds), The handbook of research synthesis. New
York: Russell Sage Foundation.

Gold, M. (1970). Delinquent behavior in an American city. Belmont, CA:
Brooks/Cole Publishing Company.

Gold, M. & Reimer, D. J. (1975). Changing patterns of delinquent behavior
among American 13 through 16 years old: 1967-72. Crime and
Delinquency Literature, 7, 483-517.

Hays, W. L. (1973). Inferences about population means. In W. L. Hays (Ed),
Statistics. New York: CBS College Publishing.

Hedges, L. V. (1994). Fixed effects models. In C. Harris & V. H. Larry
(Eds), The handbook of research synthesis. New York: Russell Sage
Foundation.

Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis.
New York: Academic Press.

Henggeler, S. W. (1989). Delinquency in adolescence. Newbury Park, CA:
Sage Publications.

Henggeler, S. W. & Borduin, C. M. (1990). Family therapy and beyond: A
multisystemetic approach to treating the behavior problems of
children and adolescents. Pacific Grove, CA: Brooks/Cole.

Hindelang, M. J. (1971). Age, sex, and the versatility of delinquent
involvement. Social Problems, 18, 522-35.

Hindelang, M. J. (197 3). Causes of delinquency: A partial replication and
extension. Social Problems, 20, 471-487.

Hirschi, T. (1969). Causes of delinquency. Berkeley: University of California
Press.

137

Hirschi, T. (1986) Crime and family policy. Journal of Contemporary
Studies, 6, 3-16.

Huizinga, D. & Elliot, D. S. (1987) Juvenile offenders: Prevalence, offender
incidence, and arrest rates by race. Crime and Delinquency, 33, 206-
223.

Huston, A. C. (1991). Antecedents, consequences, and possible solutions for
poverty among children. In Huston, A. C. (Ed), Children in poverty.
New York: Cambridge University Press.

Jang, S. J. (1992). Sex diflerences in delinquency among African American
adolescents: A longitudinal study. Unpublished doctoral dissertation,
State university of New York at Albany.

Jensen, G. F. (1972). Parents, peers, and delinquent action: A test of the

differential association perspective. American Journal of Sociology,
78(3), 562-575.

Johnson, R. E. (1976). Delinquent behavior: The development and test of a
causal model. Unpublished Doctoral Dissertation. University of
Washington Seattle.

Kotlowitz, A. (1991). There are no children here. New York: Anchor Books.

Leonard, E. B. (1982). Women, crime, and society: A critique of criminology
theory. New York: Longrnan Inc.

Lerman, P. & Pottick, K. L. (1995). The parents ’ perspective: Delinquency,
aggression, and mental health. Switzerland: Hardwood Academic
Publishers.

Lipton, W. L. & Smith, M. D. (1983). Explaining delinquent involvement: A
consideration of suppresser effects. Journal of Research in Crime and
Delinquency, 20, 199-213.

138

Liska, A. E., Krohn, M. D., & Messner, S. F. ( 1989). Strategies and
requisites for theoretical integration in the study of deviance and crime.
In S. F. Messner, M. D. Krohn, & A. E. Liska (Eds) Theoretical
integration in the study of deviance and crime: Problems and
prospects. Albany, NY: State University of New York Press.

Loeber, R. & T. J. Dishion. (1983). Early predictors of male adolescent
delinquency: A review. Psychological Bulletin, 94, 68-99.

Loeber, R. & Stoutbamer-Loeber, M. (1986). Family factors as correlates and
predictors of juvenile conduct problems and delinquency. In M. Tonry
& N. Morris (Eds), Crime and justice: An annual review of research,

Vol 7, pp. 29-150. Chicago: The University of Chicago Press.

Marcos, A. C., Bahr, S. J ., & Johnson, R. E. (1986). Test of a
bonding/association theory of adolescent drug use. Social Forces,
65(1), 135-161.

Massey, J. L. & Krohn, M. D. (1986). A longitudinal examination of an
integrated social process model of deviant behavior. Social Forces,
65(1), 107-134.

Matsueda, R. (1982). Testing control theory and diflerential association: A

causal modeling approach. American Sociological Review, 47, 489-
504.

Matsueda, R. & Heimer, K. (1987 ). Race, family structure, and delinquency:
A test of differential association and social control theories. American
Sociological Review, 52, 826-840.

McCarthy, B. & Hagan, J. (1987). Gender, delinquency, and the great
depression: A test of power-control theory. Canadian Review of
Sociology and Anthropology, 24, 153-177.

McGarvey, B., Gabriella, W. F., Bentler, P. M., & Mednick, S. A. (1981).
Rearing social class, education, and criminality: A multiple indicator
model Journal of Abnormal Psychology, 90, 354-364.

139

Mould, E. F. (1980). Chivalry and paternalism: Disparities of treatment in the
criminal justice system. In S. K. Datesman & F. R. Scarpitti (Eds),
Women, crime, and justice. New York: Oxford University Press.

Olkin, I. & Siotani, M. (197 6). Asymptotic distribution of functions of a
correlation matrix. In S. Ideka (Ed), Essays in probability and
statistics. Tokyo: Sinko Tsusho.

Panella, D. H., Cooper, P. F ., & Henggeler, S. W. (1982). Peer relations in
adolescence. In S. W. Henggeler (Ed), Delinquency and adolescent
psychopathology: A family-ecological systems approach. Littleton,

MA: Wright-PSG.

Raudenbush, S. W. (1994). Random effects models. In C. Harris & V. H.
Larry (Eds), The handbook of research synthesis. New York: Russell
Sage Foundation.

Stefl'ensmeier, D. J. & Steffensmeier, R. H. (1980). Trends in female
delinquency: An examination of arrest, juvenile court, self-report, and
field data. Criminology, 18, 62-85.

Shaw, C. R. & McKay, H. D. ( 1942). Juvenile delinquency and urban areas.
Chicago: University of Chicago Press.

Siege], L. J ., & Senna, J. J. (1988). Juvenile delinquency. St. Paul, MN:
West Publishing Company.

Simon, R. J. (1975). Women and crime. Lexington, MA: Lexington Books.

Steinberg, L. & Silverberg, S. S. (1986). The vicissitudes of autonomy in
early adolescence. Child Development, 57, 841-851.

Sutherland, E. H. & Cressey, D. R. (197 8). Criminology, 10th ed.
Philadelphia: Lippincott.

Thomberry, T. P. (1987). Toward an interactional theory of delinquency.
Criminology, 25(4), 863-891.

140

Tolan, P. H. (1988). Socioeconomic, family, and social stress correlates of
adolescent antisocial and delinquent behavior. Journal of Abnormal
Child Psychology, 16, 317-331.

Tolan, P. H. & Lorion, R. P. (1988). Multivariate approaches to the
identification of delinquency proneness in adolescent males. American
Journal of Community Psychology, 16(4), 547-561.

Wikstrom, P. H. (1993). Preface. In D. P. Farrington, R. J. Sampson, and P.
H. Wikstrom (Eds), Integrating individual and ecological aspects of
crime. Brottsforebyggande Radet, Sweden: National Council for
Crime Prevention.

 

"I1111111111111