._. x
. 93w. .
liar" . , .
5‘ u . V ‘ . . . .
, . , ‘ ,. . .1 ha mum.
. .32.. .7 min?
govt”: .

—

I

. ..~. 11......
. leun. ,
.rn...m..,
t...-1n-..

32-h.

my“... 8

1.2,!
. 5m "

a}?
.x....uw>.....\._

 

a. . . . 1B
.‘ .uiINII y , L
tail}; . 1?qu I“. .121. .
r “Urn: CY. . 5.”sz
3233. x 4
«fl..lt¥.fia I.
vW-IvuIAUn‘a lK’gf
vxl........ L

Iamluunen 2...? in»!

. icifll.ll I, i l

\ ’39:.“ ’A.(2.Iv\
.. Il‘lt!.vhu SI: L n _
I l w a q 5
V]... “5.9% m... gnu
,quv .

{If}
If!
f
«A

a.
5...
I. ‘ A":

 

 

"MM"

‘5"

jgiaogg

LIB TE EINHVERSITY
IMKHHGANSTA
EAST LANSING, MICH 48824-1048

This is to certify that the
dissertation entitled

EXPLAINING DRUG USE AND DELINQUENCY BY
RACE AND ETHNICITY: A TEST OF DIFFERENTIAL
ASSOCIATION, SOCIAL BONDS AND SELF-CONTROL

presented by

Yan Zhang

has been accepted towards fulfillment
‘ of the requirements for the

PhD. degree in Criminal Justice

 

 

7

M/{jor Professor’s Signature

1 0/01/2004

 

Date

MSU is an Afflnnative Action/Equal Opportunity Institution

-o--A---0- -.-—-.-‘-A-

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

 

 

 

 

 

DATE DUE DATE DUE DATE DUE
—SEP—2-6 21395
{0310 2 07
JAN 1 2 2009 +

 

W216

 

 

 

 

 

 

 

 

 

 

6/01 cJClRC/DateDuopBS-pJS

 

EXPLAINING DRUG USE AND DELINQUENCY BY RACE AND ETHNICITY:
A TEST OF DIFFERENTIAL ASSOCIATION, SOCIAL BONDS
AND SELF- CONTROL

BY

Yan Zhang

A DISSERTATION
Submitted to
Michigan State University
In partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY

School of Criminal Justice

2004

ABSTRACT

EXPLAINING DRUG USE AND DELINQUENCY BY RACE AND ETHNICITY:
A TEST OF DIFFERENTIAL ASSOCIATION, SOCIAL BOND,
AND SELF-CONTROL THEORY

BY

Yan Zhang

The research presents an explanatory model that
integrates three particularly dominant criminological
theoretical perspectives, differential association, social
bond and self-control theories, to examine relationships
between drug use and delinquency and race and ethnicity
(which include non-Hispanic whites, blacks Americans, Asian
Americans, and Hispanic Americans) in the United States.
The key hypotheses are that the causal processes leading to
drug use and delinquent/criminal behavior, which are
specified in the integrated model, are invariant across
racial and ethnic groups

By using the 2001 National Household Survey of Drug
Abuse dataset, the study examines the integrated model and
tests the causal processes across race and ethnicity,
addressing the questions: do the theoretical constructs
vary across racial and ethnic groups? Are the causal

processes invariant across racial and ethnic groups? What

.fi .‘
I

 

 

are the variations in parameters across racial and ethnic
groups? And what is the relative importance of
differential association, social bonds, and self-control in
explaining drug use and delinquency for different racial
and ethnic groups.

The results support the integrated theoretical model
and indicate that this integrated model could be applied
for each racial and ethnic group. However, the overall
causal mechanisms between background variables, social
bonds, differential associations and self-control vary by
race and ethnicity as well as age. The model explains the
most variance in drug use for Whites, and the least for
Asians. On the other hand, it explains the most variances
of delinquent behavior for Asians among youths, but the
least variances in criminal behavior for Asians among
adults. The parameters of the causal linkages specified in
the model vary between the racial and ethnic groups. The
determinative process of differential associations, and
self—control also vary significantly across race and

ethnicity.

Copyright by
Yan Zhang
2004

AC KNOWL EDGMEN TS

My greatest debt is to my advisor, dissertation chair
and mentor, Dr. Merry Morash. The accomplishment of this
research would not have been possible without your solid
support and guidance. Thank you for your ongoing
encouragement. You have been the perfect mentor - pushing
me when it was most needed, giving the necessary support
and feedback without question, and helping me whenever I
need your help. It is a great blessing for me to have been
one of your students. You set the standard to which the
rest of us must strive to achieve. I wish I could be a
scholar in the future just as you are.

I would also like to give special thanks to Dr.
Christopher Maxwell for his long time support and guidance
in helping me to finish the degree. Dr. Maxwell ignited my
interest in quantitative research as a first—year master
student and has remained a main support of my academic life
ever since. Your encouragement has been beyond measure
ever since I started this program.

Special thanks go to my committee members, Dr. Mickael
Reisig, and Dr. Steven Gold who took time from their busy

schedules to support my efforts toward this degree. The

many hours of help to finish this dissertation is
appreciated more than you know.

I would also like to thank my officemate Kristy
Holtfreter, my cohort Sameer Hinduja, and Hoan Bui. Thank
you for your support and encouragement.

I would like to express my gratitude to my parents
(Xiang Tao-xiang and Zhang De—chao) my brother (Zhang Feng)
and sister (Zhang Qun) for their profound love, comfort,
and encouragement. Thank you for always being there.

Finally, I have to thank my husband Zhong Qiang. He
has made many sacrifices and provided unwavering support.

Thank you Qiang for everything you have done.

vi

TABLE OF CONTENTS

LIST OF TABLES ........................................... Xi
LIST OF FIGURES ........................................ xiii
CHAPTER I
INTRODUCTION .............................................. 1
Race and Ethnicity in the United States ................. 7
Racial Disparities in Crime and Delinquency ............ 15
Arrest Data .................................... 15
Victimization-based Estimation ................. 21
Self-reported Offending ........................ 22
Theories Explaining Racial Differences ................. 24
Biological Approaches .......................... 24
Cultural Deviance Perspectives ................. 26
Control Theories ............................... 32
Structural Perspective ......................... 34
Limited Empirical Research on Connections between Race
and Ethnicity and Crime ................................ 4O
Purpose of This Study .................................. 45
Organization of Dissertation ........................... 46
CHAPTER II
LITERATURE REVIEW ........................................ 47
Differential Association Theories ...................... 51
Empirical Research on Differential Association Theory..56
Association with Delinquent Peers .............. 56
Definitions Favorable and Unfavorable to
Delinquency .................................... 58
Family Structure ............................... 6O
Differential Association and Race and
Delinquency .................................... 61
Social Bonding Theory .................................. 62
Empirical Research on Social Bond Theory ............... 65
Attachment ..................................... 65
Commitment to Conventional Lines of Action ..... 68
Involvement .................................... 69
Belief ......................................... 7O
Religiosity .................................... 71
Social Control and Race and Delinquency ........ 71
Self—control Theory .................................... 74
Empirical Research on Self-control Theory .............. 78
Competition between Differential Association and Control
Theories ............................................... 81

Vfi

Differential Association Theory and Social Bond

Theory ......................................... 81
Self—control Theory and Social Bond Theory ..... 83
Differential Association Theory and Self-control
Theory ......................................... 85
Integration of Differential Association and Control
Theories ............................................... 85
The Integrate Theoretical Model ........................ 88
Summary ................................................ 92

CHAPTER III

RESEARCH METHODOLOGY ..................................... 95
Hypotheses ............................................. 95
The Data ............................................... 96

The National Household Survey on Drug Abuse....96
Sample ........................................ 100
Study Variables ....................................... 103
The Dependent Variables ....................... 104
Independent Variables ......................... 107
Structural Models ..................................... 121
Statistical Approach .................................. 125
Introduction .................................. 125
Latent Variables .............................. 125
Second-order Factors .......................... 126
Structural Equation Modeling (SEM) ............ 127
Mplus ......................................... 130

CHAPTER IV
ANALYSIS OF DATA ........................................ 131

Background Characteristics and Racial and Ethnic
Disparities of Drug Use and Delinquent/Criminal

Behavior .............................................. 132
Descriptive Statistics for Youths ............. 132
Descriptive Statistics for Adults ............. 137

Confirmatory Factor Analysis and Mean Structure Analysis
of Latent Variables of the Theoretical

Constructs ............................................ 143
Analysis of Latent Variables for Youths ....... 143
Analysis of Latent Variables for Adults ....... 150

Structural Equation Modeling of the Integrated

Theoretical Model ..................................... 157
Structural Equation Modeling for Youths ....... 157
Structural Equation Modeling for Adults ....... 167

Multiple Group Analysis ............................... 176
Multiple Groups Analysis for Youths ........... 176
Multiple Groups Analysis for Adults ........... 196

Summary ............................................... 204

vfii

CHAPTER V

DISCUSSION AND CONCLUSION ............................. 213
Summary and Discussion .............................. 214
Predictive Efficacy of Integrated Model ....... 214

Causal Processes by Race and Ethnicity ........ 219
Limitations of the Present Study .................... 229
Implication for the Theory and Policy ............... 232
Implication for Policy .............................. 239
Reference ............................................. 243

LI ST OF TABLES

Table 1. Crime arrests by offense charged and race,

United States, 2001 .................................... 20
Table 2. Sample size by age and race/ethnicity ........ 101
Table 3. Dependent variables and indicators ........... 105
Table 4. Independent variables ........................ 108

Table 5.1. Background characteristics and self-reported
drug use and delinquency for racial and ethnic groups—
youths ................................................ 133

Table 5.2. Background characteristics and self-reported
drug use and delinquency for racial and ethnic groups—
adults ................................................ 140

Table 6.1. Confirmatory factor analysis of latent
variables—youths.......................... ............ 142

Table 6.2. Comparison of means of latent independent and
dependent variables between racial and ethnic groups—
youths ................................................ 148

Table 6.3. Confirmatory factor analysis of latent
variables—adults ...................................... 154

Table 7.1. Unstandardized coefficients of the integrated
model—youths .......................................... 163

Table 7.2. Standardized coefficients of the integrated
model—youths .......................................... 164

Table 8.1. Unstandardized coefficients of the integrated
model—adults .......................................... 173

Table 8.2. Standardized coefficients of the integrated
model—adults .......................................... 174

Table 9.1a. Relationships between background variables
and social bond variables, unstandardized coefficients—
youths ................................................ 188

Table 9.1b. Relationships between background variables
and social bond variables, standardized coefficients—
youths ................................................ 189

Table 9.2a. Relationships between independent variables
and intervening variables, unstandardized coefficients—
youths ................................................ 190

Table 9.2b. Relationships between independent variables
and intervening variables, standardized coefficients—
youths ................................................ 191

Table 9.3a. Relationships between independent variables
and drug use and delinquency, unstandardized
coefficients—youths ................................... 192

Table 9.3b. Relationships between independent variables
and drug use and delinquency, standardized coefficients—
youths ................................................ 193

Table 9.4. A variety of models to test the significance
of groups of parameters—youths ........................ 194

Table 10.1a. Relationships between background variables
and social bond variables, unstandardized coefficients—
adults ................................................ 205

Table 10.1b. Relationships, between background variables
and social bond variables, standardized coefficients—
adults ................................................ 206

Table 10.2a. Relationships between independent variables
and intervening variables, unstandardized coefficients—
adults ................................................ 207

Table 10.2b. Relationships between independent variables
and intervening variables, coefficients
adults ................................................ 208

Table 10.3a. Relationships between independent variables
and drug use and crime, unstandardized coefficients-
adults ................................................ 209

Table 10.3b. Relationships between independent variables

and drug use and crime, standardized coefficients—
adult ................................................. 210

xi

Talbe 10.4. A variety of models to test the significance
of groups of parameters—adults ........................ 211

xfi

LIST OF FIGURES

Figure 1. United States—Race and Hispanic Origin:
1790 to 2000 ............................................. 12

Figure 2. Relationships between macro—level,
individual—level factors and crime ....................... 49

Figure 3. The integrated theoretical model ............... 91

Figure 4.1. Measurement of learned definitions
of drug use—youths ...................................... 113

Figure 4.2. Measurement of learned definitions

of drug use—adults ...................................... 113
Figure 5.1. Integrated model—youth ...................... 123
Figure 5.2. Integrated model—adults ..................... 124

Figure 6.1. Integrated model of youth drug use and
delinquency—youth ....................................... 165

Figure 6.2. Integrated model of adult drug use and
Crime-adults ............................................ 175

xfii

CHAPTER I: INTRODUCTION

Throughout the last century, minorities such as black
Americans, Hispanics and Native Americans have been more
likely than whites to be involved in the criminal justice
system. They have been more likely to commit crime and
delinquency, more likely to be in jail or prison, on
probation or parole. While scholars agree that
disproportionate representation of minorities (especially
blacks) in the justice system is partially caused by racial
bias and discrimination, many believe there do also exist
race-related differences in patterns of criminality.
Scholars have differed on the extent of how much more
involved in crime and delinquency some racial groups are
than others, and also have differed on the theoretical
causal explanations. Biological perspectives claim that
genetic attributes as intelligence, temperament, and other
individual characteristics contribute to racial differences
in the prevalence of crime (Wilson and Herrnstein 1985;
Sampson and Lauritsen 1997). Subcultural theorists believe
that a black subculture of violence could explain the
racial variation in crime (Wolfgang and Ferracuti 1967).
Social disorganization approaches, on the other hand,

emphasize community structures and cultures that produce

differential rates of crime (Sampson and Lauritsen 1997).
Strain theorists argue that a lack of economic
opportunities and success of race and ethnic minorities is
the main causes of racial differences in offending (Merton
1938; Blau and Blau 1982), and differential association
theories indicate the race and ethnic ratios in crime
result from differential associations with criminal and
anti-criminal patterns (Sutherland and Cressey 1966).
Control theorists believe differential child-rearing
practices and family supervision are the fundamental
reasons in accounting for racial or ethnic variations
(Gottfredson and Hirschi 1990). Empirical research
examining different theoretical interpretations of racial
and ethnic disparities in crime and delinquency has
produced disputable results either against or in support of
competing theoretical perspectives. There is no agreement
that any single theory has adequately addressed the reason
for racial and ethnic differences in crime and delinquency
(Hawkins 1993; Gottfredson and Hirschi 1990; Sampson and
Lauritsen, 1997).

A consideration of the literature on the debate over
causes of racial disparities in criminality reveals that
few tests of the theories mentioned above have been

designed “ad hoc” for the purpose of explaining the racial

and ethnic disparities. Rather, most theories have been
applied “post hoc” to explain racial and ethnic difference
(Sampson and Lauritsen 1997). Theorists assume that
factors causing crime and delinquency are common across
race and ethnic groups (Hirschi 1969). Using this logic,

Wilson and Herrnstein claimed:

Though racial factors may affect the crime rate, the
fundamental explanation for individual differences in
criminality ought to be based-—indeed, must be based,
if it is to be a general explanation--on factors that
are common to all societies. If racial or ethnic
identity affects the likelihood of committing a crime,
it must be because that identity co—varies with other
characteristics and experiences that affect
criminality. After examining constitutional,
familial, educational, economic, neighborhood, and
historical factors, there may or may not be anything

left to say on the subject of race (1985:29).

Hirschi made similar argument in his study of “Cause

of Delinquency”:

mthe causes of delinquency are the same among Negroes
as among whites. It follows from this assumption that
we need not study Negro boys to determine the causes

of their delinquency. If we can further interpret the
relation between academic achievement and delinquency
among white boys, we will have further interpreted the
original relation between race and delinquency (1969:

80-81).

The racial and ethnic disparities in crime are therefore a
function of the magnitude of differences of intervening
variables between racial/ethnic groups derived from each
theory.

Consistent with the assumption that there are similar
causes of crime across racial groups, most empirical
research testing theories has been restricted to the white
population. For example, in testing control theory and
differential association theory, researchers used only the
non-black male data from the Richmond Youth project (e.g.,
Costello and Vowell 1999; Hirshci 1969; Matsueda 1982),
although the original dataset include information on both
non-black and black subjects. Drug use and delinquency
researchers have also developed empirically tested models

on the basis of a database generally restricted to white,

middle-class youth from small cities and/or suburban areas
(e.g., Brook et al. 1990; Elliott et al. 1985; Jessor and
Jessor 1977; Kandel 1974; Kaplan 1975). The results are
then automatically applied to minorities groups. There is
not enough systematic evidence to evaluate the extent to
which existing mainstream theories are applicable to racial
and ethnic minorities (exception, e.g. Matsueda and Heimer
1987), the population believed to be most (or least for
Asians) at risk of engaging in crime and delinquency. The
magnitudes of the variations of intervening variables
between racial and ethnic groups are not clear; and the
capacities of theoretical models to explain crime and
delinquency across different racial and ethnic groups are
uncertain.

The few studies analyzing racial and ethnic
differences in criminality focus on white and black
individuals. Researchers and theorists have been silent
about other minority groups, such as Asian Americans and
Hispanic Americans, the two most rapidly growing minority
groups. They are either combined with other minority groups
as a whole, or just excluded from study. However,
literature shows that different racial and ethnic groups
present various patterns of involvement in criminal and

delinquent behavior. While Asian Americans generally have

the lowest crime rate, Hispanics, like African Americans,
are more likely than whites to be involved in law
violations. In order to understand the connection of race
and ethnicity to criminal behavior, it is critical to study
Asian and Hispanic Americans as well as blacks and whites,
that is, to study the full continuum for involvement of
criminal behavior.

With the aim of addressing existing issues and
contributing to the literature on empirical analysis of
relationship between race and ethnicity and criminal
behavior, this study will develop an integrated theoretical
model that incorporate three major theoretical
perspectives—differential association, social bond, and
self-control theories. This study will use the National
Household Survey of Drug Abuse 2001 dataset to test the
empirical relationships and determine the efficacy of
theoretical constructs in explaining criminal patterns
(drug use and delinquency) of non—Hispanic whites and
blacks as well as Asian Americans and Hispanic Americans in

the United States.

Race and Ethnicity in the United States

In the history of United States, race has been thought
of as a genetically distinct subpopulation of a given
species (Cornell and Hartmann 1998). For example, the
racial categories are generally assumed to represent
“natural, physical, divisions among humans that are
hereditary, reflected in morphology, and roughly but
correctly captured by terms like black, white, and Asian"
(Haney Lopez 1994z6). This three—category classification
scheme and other category schemes, has been criticized by
social scientists (Haney Lopez 1994; Adrian Piper 1992;
Martha E. Gimenez 1989; Ferrante and Brown 2001). They
point out that many people do not fit clearly into a racial
category, because no sharp dividing line distinguishes
characteristics such as black skin from white skin or curly
hair from wavy. There are also no fixed and definite
boundaries between races because people may have mixed
ancestry. Social scientists contend that the
classification rules are arbitrary, vague, contradictory
and subject to change (Cornell and Hartmann 1998; Ferrante
and Brown 2001). There exists heterogeneity among people
designated as belonging to a particular race. For example,

people classified as Asian include Chinese, Japanese,

Malayans, Mongolians, and Siberians. Therefore, social
scientists argue that race is not a biological fact but a
social creation. As Cornell and Hartmann (1998:23) remark,
“races are not established by some set of natural forces
but are products of human perception and classification.
They are social constructs.”

Social scientists define race as “a group of people
loosely bound together by historically contingent, socially
significant elements of their morphology and/or ancestry”
(Haney Lopez 1994z7). Adrian Piper (1992) states, “what
joins me to other blacks, then, and other blacks to
another, is not a set of shared physical characteristics,
for there is none that all blacks share. Rather, it is the
shared experience of being visually or cognitively
identified as black by a white racist society, and the
punitive and damaging effects of that identification" (pp.
30—31). For all social purpose, what is important is not
that people are genetically different but that they
approach one another with dissimilar perspectives. It is
in the social setting that race is important. Although
race is a social concept, and the classification of racial
groups suffer from many limitations, for the purpose of
collecting and reporting information, Federal Statistical

Directive No. 15 identified five minimum racial categories

in the United States: American Indian or Alaska Native;
Asian or Pacific Islander; Black; and Whites (Ferrante and
Brown 2001).

The concept of ethnicity in social science has a very
broad meaning. It can refer to people who share (or
believe they share) a national origin; a common ancestry; a
place of birth; distinctive and visible social traits such
as religious practice, style of dress, body ornaments, or
language; and /or socially important physical
characteristics such as skin color, hair texture, and/or
physical build (Ferrante and Brown 2001). Richard A.

\\

Schermerhorn (1978: 12) defines ethnic group as a
collectivity within a larger society having real or
putative common ancestry, memories of a shared historical
past, and a cultural focus on one or more symbolic elements
defined as the epitome of their peoplehood.”

The Federal Statistical Directive No. 15 names only
two official ethnic categories into which all people in the
United States must be placed: Hispanic origin and non—
Hispanic origin. There is great diversity in how Hispanics
define themselves racially, and there are great cultural
differences between, say, Puerto Ricans and Cubans, as

there are between racial groups (Sampson and Lauritsen

1997). Not sharing a common culture, the multitude of

groups classified as Hispanic has been criticized as a
political definition that has little meaning (e.g., Mann
1993, pp. 8-12), with many preferring the label “Latino”
instead. Similar arguments have been made about the
meaning of race categories, namely, that there is more
within-group variation (in terms of traditional cultural
experiences) than there are differences between race groups
(Sampson and Lauritsen 1997).

The minimum categories set forth in Directive No. 15
are criticized for not reflecting the diversity of the U.S.
population. In October 1997, the U.S. Office of Management
and Budget (OMB) released a notice, “Revisions to the
Standards for the Classification of Federal Data on Race
and Ethnicity”, that summarizes the results of both the
research and the public comment. It also provides the new
standard for race and ethnicity, which includes five racial
categories: American Indian or Alaska Native, Asian, black
or African American, Native Hawaiian or Other Pacific
Islander, and white; and two ethnic categories: Hispanic or
Latino and Not Hispanic or Latino. According to the OMB
noticel, American Indian or Alaska Native means “a person

having origins in any of the original peoples of North and

 

' http:C’oas.samhsa.gov/NHSDA/methods'RaceEthnicitypdf.

South America (including Central America), and who
maintains tribal affiliation or community attachment.”
Asian means “a person having origins in any of the original
peoples of the Far East, Southeast Asia, or the Indian
subcontinent including, for example, Cambodia, China,
India, Japan, Korea, Malaysia, Pakistan, the Philippine
Islands, Thailand, and Vietnam.” “A person having origins
in any of the black racial groups of Africa” is classified
as Black or African American. Native Hawaiian or Other
Pacific Islander is a person having origins in any of the
original peoples of Hawaii, Guam, Samoa, or other Pacific
Islands. White is a person having origins in any of the
original peoples of Europe, the Middle East or North
Africa. The ethnic category “Hispanic or Latino” is
defined as a person of Cuban, Mexican, Puerto Rican, Cuban,
South or Central American, or other Spanish culture or
origin, regardless of race.

The U.S. Census Bureau data indicate that the resident
population of the United States has been changed
considerably over the past 200 years. Figure 1 presents
the population changes from 1790 to 2000 by race and

Hispanic origin.

Figure 1. United States-—Race and Hispanic Origin: 1790 to 2000
1790 1790

19.30% 790% 7

   
       
    

thile
I black

Hispanic
. I Hrspnaic

  

 

1
80.70%

92.10%

1940
1940

 

  
 

       

 

 

E! w hits
I black
Hispanic
0 American I Hrspnaic
Indian origin
El Asian P. b

I ”89.71% 88.40%

 

 

 

 

  
   

    
 
 

 

  

 

 

 

 

    
     

1990
390% 1990
0 15,40%
°-8°/° thite
1 .1 ° - ‘ ' ' .
2 0 A Iblack 9,00%
DAmerican
Indian Hispanic
UAsian P. ‘ I Hlspnaic
ls. origin
I other 75.60%
80.30%
2000
2000
thite 18.40%
I black
DAmerican
Indian 12. 50% Hispanic
CIAsian P I Hrspnarc
Is. origin
I other

 

    

 

 

 

Whites made up 75.1 percent of the approximately 211
million residents of the U.S. population in 2000. This
represents a decline from 80.3 percent in 1990. Blacks
represent 12.3 percent of the 2000 population, up modestly
from 1990. American Indians comprise only 0.9 percent of
the population. However, Hispanics and Asians have been
growing fast during the past half—century. According to
U.S. Census Bureau, in 1940, the estimated Hispanic
population was only about 1.86 million, accounting for 1.4
percent of the American population. However, the Hispanic
population has increased 18 times over the last sixty
years. In 2000, the Hispanic population was more than 35
million, accounting for 12.5 percent of the American
population, and is now larger than the population of
blacks. Hispanics have become the largest minority group
in the United States. The Asian American population has
also increased rapidly. After the enactment of the
Immigration Act of 1965 that abolished discriminations
based on national origins and opened the door to
immigration from all countries, the number of Asian
immigrants to the United States increased 750 percent over
the last three decades, from 1.4 million (or 2.7 percent of
the American population) in 1970 to 11.9 million (or 4.2

percent of the American population) in 2000 (US Census

‘¢J

 

Bureau 1993, 2002). Estimates are that the Asian American
population will reach 20 million in 2020 (US Census Bureau,
2000).

The change of the population composition and the
increasing diversity in the United States call for more
attention to the minority groups. To understand American
society one should understand the heterogeneity of
different minority groups. In the current study, analysis
will concentrate on the criminal patterns of four racial
and ethnic groups: non-Hispanic white, non—Hispanic black,
non—Hispanic Asian, and Hispanic American. Three other
minority groups: non-Hispanic American Indian/Alaska
Native, non—Hispanic Hawaiian/Pacific Islander, and non—
Hispanic more than one race will not be included in the
current study because of the inadequate sample size. This
racial and ethnic classification scheme uses the combined
format and is consistent with the 1997 OMB notice of
“Revisions to the Standards for the Classification of
Federal Data on Race and Ethnicity.” It also identifies
groups that experience differential treatment and reactions
within U.S. society, and that therefore are meaningful in a
framework that considers race and ethnicity to be social

constructions.

Racial Disparities in Crime and Delinquency

Researchers and criminologists have long recognized
the racial and ethnic disparities in rates of adult crime
and juvenile delinquency in the United States (Hawkins et
al., 2000; LaFree, 1994; LaFree and Russel, 1993). Even
so, because there exist different source of data on crime—
official arrest data, self—reported crime, and national
victimization survey data etc., findings on the correlates
of criminal offending and race are not consistent. For
more than seventy years, scholars have differed on how much
more involved in crime blacks are than whites (Tonry,

1995).

Arrest Data

 

Studies that have used official data on arrests have
indicated that criminal offending, especially serious and
violent offending is more common among blacks and Native
Americans than among whites, and that offending among Asian
Americans is least common (Hawkins et al., 2000; Snyder,
1999). For example, the FBI 2001 Uniform Crime Report
(UCR) data (Table 1) indicates that whites account for 69.5
percent of all arrests, blacks, American Indians and Asians

account for 28.1 percent, 1.3 percent, and 1.1 percent of

U

all arrests respectively. Although whites are arrested for
the majority of all crimes, blacks and American Indians are
still disproportionately represented in arrests reported in
the UCR. In 2001, blacks comprised 28.1 percent of total
arrests yet constituted 12.3 percent of the total
populationz. The black arrests rate is 2.5 times that of
the arrest rate for whites. American Indian and Alaska
Natives comprised 1.3 percent of total arrests but just .9
percent of the population. The American Indian and Alaska
Native arrests rate is about 1.6 times that of whites.
Asians, however, appear to be underrepresented in arrest
statistics. Asians account for 1.1 percent of all arrests,
yet make up 3.7 percent of the population. The Asian
American arrest rate is only about a third of the white
arrests rate. The general pattern of overrepresentation of
blacks and American Indians among persons arrested has been
consistent throughout the (for the last fifty years)
recorded history of the last century, though black and
white differences in rates of offending have decreased
somewhat in recent years (Figure 2).

The relationship between race and arrest is not the

same for all crime types; there are certain offenses for

 

2 Crime arrests data are obtained from Sourcebook ofCriminal Justice Statistics 2002. pages 356-358. The
population data are obtained from U.S. Census Bureau Population estimates website,
http://eire.census.gov/popest/data/national/tables/NC-EST2003-asrh.php. Then the rates are calculated by
author.

 

which each group is over represented (Sampson and Lauritsen
1997). For instance, whites are most likely to be arrested
for driving while intoxicated, and Asians are over-
represented in arrests for illegal gambling. Native
Americans are more likely to be arrested for liquor laws
and drunkenness, and blacks are disproportionately arrested
for violent crime (Hindelang 1978; Elliott and Ageton 1980;
Sampson and Lauritsen 1997). In 2001, blacks accounted for
48.7, 34.8, 53.8 and 33.7 percent of murder, rape, robbery,
and aggravated assault arrests respectively3. In general,
blacks are approximately six times more likely to be
arrested for violent crimes than are whites.

The relationship between race and arrest also differs
by ages. In 2001, black juveniles’ arrest rates were 1.85
times that of white juveniles (36.3 per 1000 for black
juveniles, 19.6 per 1000 for white juveniles), but the
black adults' arrest rates were 2.88 times that of white
adults (88.5 per 1000 for black adults, 30.7 per 1000 for
white adults). The percentage of violent crime arrest for
black juvenile is even higher than that of black adults.
For every 100 arrested black juveniles, 7 of them are
arrested for violent crimes; and the rate for black adults

is 6.

 

’Sourcebook of Criminal Justice Statistics 2002, pages 356—358.

[7

There have been major disagreements about the validity
of the patterns shown by official statistics. Researchers
argue that the disparities in the arrest data partially
result from racial bias in police arrest and crime—
recording practices (e.g., Sutherland and Cressey, 1966;
Tonry, 1995; Sampson and Lauritsen, 1997; Wilson and
Herrnslein, 1985; Hawkins et al., 2000; Hagan and Peterson,
1995). Young black men suffer more from stereotypes of
blacks as “disreputable” and “dangerous" (Sampson and
Lauritsent 1997), which leads to police arresting young
black men more frequently than would be expected from
levels of involvement in actual criminal behavior. Thus,
almost any index of crime rates is likely to exaggerate the
rate for blacks as compared with the rate among whites
(Sutherland and Cressey 1966).

Scholars pointed out the disproportionate
representation of minorities in crime arrest could also be
the results of government policies. One example is for
drug-related arrests. Research shows that from 1965
through the early 19808, nonwhites arrest rate were
approximately twice as that of whites for drug-related
offenses (Blumstein 1993; Tonry 1995). Following the

I

federal government’s launch of the “war on drugs,’ nonwhite

arresst rates rose steadily and skyrocketed, while white

arrest rates were basically stable and increased only
slightly. By the end of the 19805, nonwhite arrest rates
of drug-related offenses were five times higher than that
of whites (Tonry 1995). An examination of the juvenile
arrest rates for drug—related offenses revealed a more
salient pattern (Blumstein 1993). From the late 19605 to
the early 19805, white arrest rates for juvenile drug
offenses were slightly higher than those for black
juveniles. White arrest rates continued to drop after
19805, but black rates soared until the late 19805 when
they were four to five times higher than white rates
(Blumstein 1993). The national drug use survey data,
however, indicated that drug use was declining among both
white and nonwhite populations since 19805 (Tonry 1995).
Therefore it is highly unlikely that these race differences
in arrests represent general substance use patterns.
Rather, these differences reflect the government’s
targeting and enforcement of specific types of drug use and
trafficking (Blumstein 1993; Tonry 1995). To check on the
extent of bias in police arrest records, victim surveys,
and self-reported data are used (Wilson and Herrnslein,

1985; Sampson and Lauritsen, 1997; Hawkins et al., 2000).

Table 1. Crime arrests by offense charged and race, United States, 2001
(percentage)

 

 

 

White Black Native Asian and
Pzzific
Island
Total 69.5 28.1 1.3 1.1
Murder and non-negligent
Manslaughter 48.4 48.7 1.3 1.6
Forcible rape 62.7 34.8 1.1 1.4
Robbery 4 5 53.8 0.6 1.1
Aggravated assault 64.0 33.7 1.1 1.2
Burglary 69.4 28.5 1.0 1.1
Larceny-theft 66.1 31.2 1.2 1.5
Motor vehicle theft 57.5 39.8 0.9 1.8
Arson 76.9 20.7 1.5 0.9
Violent crime 60.2 37.6 1.0 1.2
Property crime 66.0 31.4 1.2 1.5
Total Crime Index 64.4 33.1 1.1 1.4
Other assaults 65.4 32.2 1.3 1.1
Forgery and counterfeiting 67.8 30.1 0.6 1.5
Fraud 67.7 31.0 0.6 0.7
Embezzlement 65.8 32.0 0.5 1.6
Stolen property; buying,
Receiving, possessing 4 38 7 0. i 1
Vandalism 74.7 23.0 1. 1
Weapons; carrying,
Possessing, etc. 61.0 37.2 0.7 1.0
Prostitution and
commercialized vice 57.1 40.3 0.6 2.1
Sex offenses (except
forcible Rape and
prostitution) 73.3 24.2 1.1 1.4
Drug abuse violations 64.2 34.5 C. 0.7
Gambling 27.7 68.1 0.1 4.1
Offenses against family and
children 67.7 2 .5 1.2 1.7
Driving under the influence 87.4 1 .2 1.4 1.0
Liquor laws 86.4 10.0 2.7 0.9
Drunkenness 83.9 1 .3 2.3 0.5
Disorderly conduct 64.4 33.4 1.4 0.8
Vagrancy 61.2 35.7 2.5 0.6
All other offenses (except
traffic) 66.0 31.4 1.4 1 z
Suspicion 61.8 36.9 0.4 0
Curfew and loitering law
violations 70.5 27.1 0 9 1.5
Runaways 75.4 19.1 1.2 4.2
Source: Sourcebook of Criminal Justice Statistics 2002, pages 356—35:

Victimization-based Estimation

 

Estimates of violent crimes for different racial
groups, as reported in the 2001 National Crime Victim
Survey (NCVS) and 2001 UCR arrest data, show much the same
pattern as indicated previously. Blacks comprised 24.6
percent of the violent offenders, and whites comprised
about 64.3 percent of the violent offendersq. The black
violent crime rate is about 2.3 times of that of the white
crime rates. However, the 2001 NCVS shows that blacks
account for only 22.5 percent of rape offenders, which is
significantly lower than the percentage obtained from the
arrested data (34.8). The percentages of blacks robbery
and aggravated assault are 47.4 and 26.8 respectively.

A limitation of victim surveys is that the race of the
offender is only known when there has been interaction
between victim and offender, thus, offender race is rarely
known for burglaries and larcenies. Another problem is
that the NCVS data produce incidence rates instead of
prevalence rates, and each racial subgroup may contain
different proportions of repeat offenders (Sampson and

Lauritsen, 1997). The NCVS data is also limited because

 

4 http://www.ojp.usdoj.gov/bjs/pub/pdf/cvuso102.pdf. The population data are obtained from U.S. Census
Bureau, Census 2000 Redistricting Data (PL. 94-171) Summary File for states.

21

the race estimates are based solely on victim reports of

lone—offender crime (Sampson and Lauritsen 1997).

Self-reported Offending

 

Self-reported offending data typically measure less
serious forms of common delinquency by juveniles (Hindelang
et a1. 1979, 1981; Wilson and Herrnslein, 1985). Studies
based on self-reported data usually found little or no
differences among juveniles of different racial and ethnic
groups (see Hindelang et a1. 1979, 1981). For example, the
1997 National Longitudinal Survey of Youth indicates the
“ran away from home” rates are 11 percent and 10 percent
for white youths (12 to 16 years old) and non-white youths
respectively. About 10 percent of whites have ever carried
a handgun, and about 9 percent non-whites have ever carried
a handgun. The percentage reporting “ever stole a vehicle
for use or sale” is the same (1 percent) for both whites
and non-whites. The Monitoring the Future Study shows that
during 1997 and 1998, alcohol and drug use are more common
among white and Hispanic high school seniors than black

high school seniors. About 30 percent of black seniors

reported use of Marijuana, but 39.9 percent of whites, and
37.2 percent of Hispanics reported use of Marijuanas.

One reaction to these findings was to attribute racial
bias to official statistics. Others posited methodological
explanations. In particular, Hindelang et al. argued that
self-report studies typically measure less serious forms of
common delinquency, whereas official arrest statistics
showing race differentials refer primarily to serious index
crimes. The validity of self-reported data may also differ
by racial group, because blacks underreport certain
offenses at higher rates than do whites (Hindelang et a1.
1981; Sampson and Lauritsen 1997; Hawkins et al. 2000).

But a more recent study found no racial differences in
predictive validity based on these self—reports (Farrington
et al., 1996). One major study that used self~reported
data, the National Youth Survey, showed that serious and
violent juvenile offenders were disproportionately black
males (Elliot and Ageton, 1980; Elliott et al., 1986).

In summary, although no one denies that there is bias
in the criminal justice system and limitations of available
data sources, it is unlikely that the system bias against
racial minorities is the major reason that causes the

CiiSproportions. There do exist race-related differences in

 

 

5
h ttp : //www.ncjrs.org/html/ojjdp/nationalreport99.’chapter3 .pdf

23

 

patterns of criminality. These differences have provoked
competing theoretical interpretations and public policy
debates (Hawkins, 1995; Hawkins et a1. 2000; LaFree and

Russel, 1993).

Theories Explaining Racial Differences

Conventional theoretical models explaining the
connection between race and ethnicity and crime reflect at
least four different approaches. These are: biological (or
constitutional) approaches, the cultural deviance
perspectives, control theories, and structural theories
(Bruce, 2000; Hawkins et al., 1995; Sampson and Wilson,

1995; Wilson and Herrnstein, 1985).

Biological Approaches

 

Early biological perspectives distinguished criminals
from non—criminals by certain physical anomalies with
hereditary origins (Lombroso, 1876; Sheldon, 1936, 1940,
1942). They sought to explain high rates of crime among
African-Americans in terms of genetic inferiority or other
inherent physical deficiencies (LaFree and Russel, 1993).
For example, Brearley (1932) asserted that the

disproportionate rate of homicide among blacks is due to

“their peculiar genetically—determined temperament” as well
as “excessive emotionality” (111-16). Contemporary
biocriminal perspectives acknowledge both the importance of
biological factors and social environment. In the
biocriminal perspective, the tendency to commit crime has
both constitutional and social origins (Herrnstein and
Murray, 1994; Wilson and Herrnstein, 1985). However, a
closer inspection of these ideas reveals that the core of
the traditional argument remains intact (Bruce, 2000). For
instance, Wilson and Herrnstein (1985) suggested that
differences in black and white crime rates in the United
States could be attributed to constitutional differences of
youthfulness, intelligence and temperament for the two
groups. They claim, “to the extent these differences exist
and increase the probability of offending, some of the
racial disparity may be explained” (1985zl). Biological or
constitutional criminal perspectives generate much
controversy and have been challenged by people working
within alternative perspectives.

One controversy is that the biocriminological
perspective has a tendency to medicalize political issues.
The most disturbing example is that this theory has been
used by totalitarian regimes such as Nazi Germany (Bruce,

2000; Einstadter and Henry, 1995). As a result, many

scholars resist the biological perspective on political and
policy ground (Sampson and Lauritsen, 1997). In addition
to their political implications, biological criminal
theories have been criticized for a lack of empirical
support (see Jencks, 1992; Shoemaker, 1984; Sutherland &
Cressey, 1974). Many of the key hypotheses have yet to be
tested, making it nearly impossible to assess the empirical
validity of the arguments made (Jencks, 1992). Indeed,
even supporters of biological criminology approaches find
no evidence that biological factors play a causal role in

violent crimes in general (Mednick et al., 1988).

Cultural deviance perspectives

 

Cultural deviance theories generally assume crime is
always relative to the norms of the group defining it as
crime-—therefore, it is a product of social definitions
(Gottfredson and Hirschi 1990). Differential association,
subculture, labeling, and conflict theories all have their
foundation in the cultural deviance perspectives
(Gottfredson and Hirschi 1990). The following discussion
focuses on subcultural and differential association

explanations of racial variation in crime.

Subculture Theories

 

Subculture of violence perspectives argue that there
are certain groups in the United States that have developed
values that are conducive to crime or that approve of or
justify crime in certain circumstances (Anderson 1994;
Cloward and Ohlin 1960; Cohen 1955; Miller 1958; Wolfgan
and Eerracuti 1967). Once the values develop, they are
passed on from generation to generation, even in the
absence of the deprivation that stimulated their
development in the first place (Wolfgan and Ferracuti 1967;
Curtis 1974). The most influential of these perspectives
is the subculture of violence theory developed by Wolfgang
and Ferracuti (1967). Wolfgang and Eerracuti argued that
certain segments of society have adopted distinctively
violent subcultural values. This value system provides its
members with normative support for their violent behavior,
thereby increasing the likelihood that hostile impulses
will lead to violent action. In attempting to explain the
high rate of homicide among young African-American males in
the inner city, Wolfgang and Ferracuti (1967) speculate

that there is a subculture of violence among blacks:

A male is expected to defend the name and honor of his

mother, the virtue of womanhood and to accept no

derogation about his race, his age, or his masculinity

(1967:153).

In “The Code of the Streets,” Anderson provides a
contemporary account of the subculture of violence thesis.
He argues that there exists a “code of the street” in poor,
inner-city African-American communities. The code
pressures African American youth in the inner city to
respond to shows of disrespect with violence (Anderson,
1994).

However, little empirical evidence was found to
support the conclusion that blacks and white Americans
differ significantly in their attitudes and values toward
crime (Kornhauser, 1978; DiIulio, 1995; Cao et al., 1997;
Gottfredson and Hirschi, 1990). For instance, Ball-Rokeach
(1973) and Poland (1978) tested the hypothesis that violent
behavior results from a commitment to subcultural values
condoning violence. Both found little evidence in support
of the subculture of violence thesis. Erlanger (1974)
tested Wolfgang and Ferracuti’s contention that there is a
subculture of violence among blacks and found that there is
“an absence of major difference by race" in approval of
interpersonal violence (p. 283). An empirical study by Cao

an<i associates (1997) also rejected the hypothesis that a

unique subculture of violence exists among the general
population of African Americans in the United States.
Gottfredson and Hirschi (1990: 151) pointed out the reason
that the subcultural view of criminality lacks empirical
support is that all groups, whatever their racial or ethnic
composition, condemn the use of force or fraud and endorse
values contrary to crime. Even those who commit crime do
not endorse criminal acts.

Another limitation of the subculture theories to
explain racial differences is that they are tautology. In
subculture models, violent behaviors are used to infer the
existence of a subcultural system, which in turn is used to
explain behavior (Sampson and Lauritsen, 1997).
Structurally oriented researchers criticized that
subcultural thought assumes that members of a particular
societal subgroup create and maintain a unique subculture.
They argued that the observed cultural tendencies are in
fact manifestations of local structural conditions and
general levels of opportunity (Anderson, 1990; Wilson,
1987). As Sampson and Wilson (1995) pointed out, ignoring
the structural features of a given community has the effect
Of directing attention to alleged problems within African

American and other disadvantaged communities. The role of

subculture in creating racial disparities of criminal

behavior remains to be demonstrated.

Differential association theory

 

The premise of differential association theory is that
criminal behavior is learned by interacting with others,
especially intimate others (Sutherland and Cressey, 1966).
“A person becomes delinquent because of an excess of
definitions favorable to violation of law over definitions
unfavorable to violation of law" (Sutherland and Cressey,
1966:81). According to Sutherland, factors such as race,
social class, and broken homes influence crime because they
affect the likelihood that individuals will associate with

others who present definitions favorable to crime:

mThe race ratios in crime result from differential
associations with criminal and anti-criminal patterns.
In the United States, race may be related to
associations in either or both of two ways. First,
the inheritance of certain characteristics may
determine the social and economic level at which
members of a race must live and thus possibly throw
them into situations where certain criminal patterns

impinge upon them with great frequency and

intensity Second, confinement of a race to a given
locale may mean that the members of the race can
hardly escape the traditions which are characteristic
of the locale. The traditions may be in conflict with
the laws under which people of a certain race are
expected to live, or they may be essentially anti—
criminal. Thus, the experience of people in a racial
group may be such that there is much or little
conflict between its traditional ways of behaving and
the laws under which is the group expected to live

(Sutherland and Cressey, 1966:151-152).

Empirical analyses provide evidence supporting differential
association theories to explain crime and delinquency
(e.g., Matsueda 1982). However, little research has tested
whether differential association theory can be equally
applied to different racial and ethnic groups (an exception
is Matsueda and Heimer 1987); the capacity to explain
racial variation of crime and delinquency also is unknown.
Because differential association theory is promising, a
more detailed review of differential association theory and

its empirical status will be presented in chapter two.

Control theories

 

Control theories are distinguished from other
criminological theories by theirs assumption that
“individuals are capable of committing delinquent and
criminal acts without benefit of example, training, or
rewards over and above those inherent in the acts
themselves” (Hirschi and Gottfredson 2003: 151). Control
theories claim, “all crimes are mundane, simple, trivial,
easy acts aimed at satisfying desires of the moment, as are
many other acts of little concern to the criminal law"
(Gottfredson and Hirschi 1990: xv). Crimes are available
to everyone, and require no special motivation or training
for their commission (Hirschi and Gottfredson 2003).
Control theories assume the causes of crime are the same
everywhere and at all times, and are not products of unique
culture, peoples, setting, historical periods, or even
variable legal definitions of crime (Britt and Gottfredson
2003). Two representative control theories are Hirchi’s
social bond theory and Gottfredson and Hirschi's self-

control theory.

Social bond theory

 

Social bond theory maintains that persons conform to legal

codes because they are bonded to society. Criminal and

Lu
tJ

delinquent acts result when an individual’s bond to society
is weak or broken. Hirschi (1969) classified four elements
of the bond to conventional society: attachment, belief,
commitment, and involvement. Since control theory
presumes a general causation of criminal behavior for all
groups, Hirschi claims that the causes of crime are same
among blacks and whites. The difference of crime rates is
thus the results of differences of elements of social bond

that people in each racial group have.

Self-control theory

 

The basic premise of Gottfredson and Hirschi's self—
control theory is that juvenile delinquency, adult crime,
and “analogous” behaviors are caused by low selfecontrol,
which is formed at a very early age through inconsistent
and ineffective parenting. According to Gottfredson and
Hirschi (1990:153), differences among racial and ethnic
groups in levels of direct supervision by family, and more
importantly differences in self—control are the major
factors that account for racial or ethnic variations of
crime and deviant behavior. They claimed that the large
crime differences between racial groups in the United
States result from variation in approaches to child

rearing. The elements of child—rearing identified by

33

Gottredson and Hirschi (1990) include attachment of the
parent to the child, parental supervision, recognition of
deviant behavior, punishment of deviant acts, parental
criminality, family size, family structure, and mother‘s
employment status etc. However, to date no analysis has
been provided on how different racial groups vary in each
of the conditions of child rearing, and how these

differences are related to variation in crime rates.

Structural perspective

 

Instead of focusing on characteristics internal to a
racial or ethnic minority group or a person, social
structural perspectives center on a group’s status and
material reality in a class— and race-stratified society
(Bruce, 2000). These perspectives suggest that
characteristics of communities, cities, or societies such
as harsh economic conditions facing a population, coupled
with high levels of residential segregation, and social
disorganization and weak social control account for the
disparate crime rates. The structural theories that are
discussed here include strain theory, and macro- and

community-level perspectives.

34

Strain theory

Strain theories have sometime been classified as
social structural theories (Einstadter and Henry, 1995).
However, rather than focusing on various areas or community
environments, they locate the causal force to adapt in the
organization of the whole society (Einstadter and Henry,
1995).

Merton’s (1938) strain theory argued that the
motivation for crime frequently derives from society. He
claimed that societies that place a high relative emphasis
on cultural goals like monetary success and a low relative
emphasis on the norms or rules for goal achievement have
higher crime rates. Merton pointed out that some
individuals and groups within a society are subject to
special pressure for crime. While everyone is urged to
strive for monetary success, lower—class individuals are
frequently prevented from achieving such success through
legitimate means. Individuals experiencing such goal
blockage are under considerable strain or pressure, and are
most strongly motivated to use illegitimate means (e.g.

crime) for success:

First, such antisocial behavior is in a sense “called

forth” by certain conventional values of the culture

and by the class structure involving differential
access to the approved opportunities for legitimate,
prestige—bearing pursuit of the culture goals. The
lack of high integration between the means—and-end
elements of the cultural pattern and the particular
class structure combine to favor a heightened
frequency of antisocial conduct in such groups.
Recourse to the first of the alternative responses,
legitimate effort, is limited by the fact that actual
advance toward desired success-symbols through
conventional channels is m relatively rare and
difficult for those handicapped by little formal
education and few economic resources. The dominant
pressure of group standards of success is, therefore,
on the gradual attenuation of legitimate, but by and
large ineffective, strivings and the increasing use of
illegitimate, abut more or less effective, expedients

of vice and crime (1938).

Although Merton’s theory did not examine the relationship
of race and crime specifically, it would be consistent with
Merton's strain theory to explain the high crime rate among

blacks as the result of the conflict between the culture

goals and the lack of legitimate channel of blacks to
achieve the goals.

Strain theories are criticized because they assume
that race or ethnicity is equivalent to lower class or
deprived status (Gottfredson and Hirschi, 1990; Sampson and
Lauritsen, 1997). Such a view omits middle-class crime or
for that matter crimes of the powerful, regardless of their
class. Moreover, some racial and ethnic groups such as
Chinese, Japanese and Hispanic, who have also experienced
economic exclusion and deprivation as African Americans,
exhibit a much lower crime rate than do African Americans.
It is unknown to what extent structural or cultural
differences account for lower offending rates among those
ethnic groups, but clearly socioeconomic status and
deprivation alone are inadequate explanations (Wilson and

Herrnstein 1985).

Macro-level or community-level theories
One of the main studies focusing on macro-level
structure condition is Blau and Blau’s (1982) work on
metropolitan structure and crime. Blau and Blau (1982)
emphasized the effects of relative deprivation on black

crime. They suggested that social inequality between

37

blacks and whites, as well as economic inequality,
generally increased rates of criminal violence.

The Chicago School research of Clifford Shaw and Henry
McKay indicated that three structural factors—low economic
status, ethnic heterogeneity, and residential mobility—led
to the disruption of local community social organization,
which in turn accounted for variations in crime and
delinquency rates (Kornhauser, 1978; Sampson and Wilson,
1995). Through the study of Chicago delinquency, Shaw and
McKay (1942) observed that high rates of delinquency
persisted in certain areas over many years, in spite of

population turnover:

The important fact about rates of delinquency for
Negro boys is that they, too, vary by types of area.
They are higher than the rates for whites boys, but it
cannot be said that they are higher than rates for
white boys in comparable areas, since it is impossible
to reproduce in white communities the circumstances
under which Negro children live. Even if it were
possible to parallel the low economic status and the
inadequacy of institutions in the white community, it

would not be possible to reproduce the effects of

38

segregation and the barriers to upward mobility (1949:

641).

Shaw and McKay therefore rejected individualistic
explanations of delinquency and focused instead on the
processes by which delinquent and criminal patterns of
behavior were transmitted across generations in areas of
social disorganization and weak social controls (1942:
320). Building on Shaw and MaKay’s work, Sampson and
Wilson (1995) developed a contextual theory that
incorporated structural and cultural factors to explain the
race and crime relationship. They viewed “the race and
crime linkage from contextual lenses that highlight the
very different ecological contexts that blacks and whites
reside in-regardless of individual characteristics”
(1995:1). The basic thesis was that macro social patterns
of residential inequality give rise to the social isolation
and ecological concentration of the truly disadvantaged,
which in turn leads to structural barriers and cultural
adaptations that undermine social organization and hence
the control of crime.

Ecological studies generally are seen to have several
inherent problems and limitations that affect their

usefulness. Davidson (1981:87—89), for example, says there

39

is a fundamental methodological problem of false or
spurious results. What may be observed on the aggregate
level may not be the case for individuals. Drawing
conclusions for individuals from aggregate data is known as
the ecological fallacy.

One crucial theoretical criticism is that the
explanation is tautology. The central explanatory concept,
social disorganization, is presented as both a description
of a condition and the cause of that condition. Another
criticism is that racial composition is used to denote
material and structural disadvantage; however, this measure
does not specify why African Americans are disadvantaged at

the outset (Bruce, 2000).

Limited Empirical Research on Connections Between

Race and Ethnicity and Crime

A multitude of studies have tested and compared
theories to explain how individuals become involved in
criminal and delinquent behaviors. However, empirical
research examining competing explanations of the connection
of race and ethnicity to crime and delinquency remains
incomplete (Hawkins et al., 2000; McNulty, 2001; McNulty

and Bellair, 2003; Sampson and Lauritsen, 1997). Although

40

race and ethnicity are critical variables in many theories
of crime, previous studies of racial and ethnic differences
in the criminal justice systems centered on the processing
of offenders. A check of publications in some of the major
criminal justice and criminology journals provides evidence
of the limited research on why race is correlated with
crime. For example, Criminology, during 1989 to 2003,
published 36 articles related to race, and less than a
third of them examined theoretical models that explained
how race was related to the etiology of crime. Justice
Quarterly, during 1984 to 2003, published 25 articles
related to race, and no more than 5 of them examined the
relationship between race and criminality. The Journal of
Criminal Justice, from 1973 to 2003, published about 90
articles related to race, and only about 6 of them analyzed
the race and crime relationship. Several reasons
contribute to the limited number of studies on the
relationship of race, ethnicity, and crime.

The first reason, as Hawkins points out, is that
“public discourse about both crime and race in the United
States has always been an ideological and political mine
field” (1995). Given the history of racial discrimination
in the United States, any examination of black-white

differences in criminal behavior is likely to be

4|

politically sensitive and controversial (Wilson and
Herrnstein, 1985). Despite the abundance of empirical
data, many criminologists long avoided speaking openly on
the subject of “race and crime” for fear of being
misunderstood or labeled a racist (Tonry, 1994; Sampson and
Lauritsen, 1997). Some even argue that using empirical
data to study the race and crime question is itself
evidence of racism (MacLean and Milovanovic, 1990; from
Sampson and Lauritsen, 1997). The extreme sensitivity of
the subject has tended to discourage rather than stimulate
the development of theoretical studies of race, ethnicity,
and crime (Hawkins, 1995; Walker et al., 1996).

Second, theories about racial and ethnic differences
in violence have primarily been developed in reference to
comparisons between whites and blacks. Despite the
changing racial and ethnic diversity of U.S. society, data
on crime information for Hispanic and non-Hispanics, and
for groups such as Asian Americans and Native Americans are
lacking. However, the tremendous growth in the Hispanic
and Asian population, and different patterns of involvement
of criminal and delinquent behavior of different racial or
ethnic groups (i.e. the low crime rates of Asians, and the
relatively high crime rates for Native Americans and

Hispanics) justify consideration of these groups.

Since the UCR and NCVS data provide no information
about the offenders’ ethnicity, little is known about the
Hispanic criminality patterns. The victimization
information obtaining from the NCVS data shows that at
2002, Hispanics were victims of overall violence at about
the same rate than non—Hispanics. However, the data
obtained from national correction system indicate that
Hispanic males were about 2.6 times likely as white males
to be incarcerated. At yearend 2002 there were 1,176
sentenced Hispanic male prisoners per 100,000 Hispanic
males in the United States, compared to 450 white male
inmates per 100,000 white males6. This information might
give us a good reason to assume that Hispanics just like
African Americans and Native Americans are more likely than
whites being involved in law violations. In order to
understand the connection of race and ethnicity to criminal
behavior, it is critical to study Asian and Native American
and Hispanic groups, who are at opposite ends of the
continuum for involvement of criminal behavior.

The third issue is that most analytical work focuses
mainly on street or index crimes—especially violence (e.g.,
murder, rape, robbery, and assault) and property crime

(e.g., larceny, motor vehicle theft, and burglary). The

__

6 http://www.jointcenter.org/DB/factsheet’correctionalsyshtm

43

types of crime covered in previous racial—crime
relationship studies are not fully representative of the
landscape of criminal behavior in the United States
(Sampson and Lauritsen 1997). One example is that very
little research examines the theoretical explanation of
racial and ethnic disparities in drug abuse, although data
collected by different sources show that not only are there
large racial and ethnic differences in drug use, but also
the patterns are different from other types of crime. For
example, the National Household Survey of Drug Abuse data
shows that the rates of current illicit drug use were 7.2
percent for whites, 7.4 percent for blacks, and 6.4 percent
for Hispanics. The rate was highest among American
Indian/Alaska Natives (9.9 percent) and persons reporting
more than one race (12.6 percent). Asians had the lowest
rate (2.8 percent). The proportion of black adolescents
who use drugs is even lower than the proportion of white
adolescents who do the same (9.7 percent for black
adolescents, 11.4 percent for white adolescents). This
finding has been replicated in household surveys, school
based-surveys, and in surveys of institutionalized youth
(for example, Austin, Prendergast, and Lee, 1989; Kandel,
Single, and Kessler, 1976; Welte et al., 1999; Oetting and

Beauvais, 1990; Prendergast, et al. 1989; National

44

Institute on Drug Abuse, 1995). Overall, the body of
literature is very small, and many important topics
relevant to race, ethnicity, and crime and delinquency have

not been researched at all.

Purpose of This Study

To address existing issues and contribute to the
literature on empirical analysis of racial and ethnic
differences of criminal behavior, this study will develop
an explanatory model that integrate three particularly
dominant criminological theoretical perspectives,
differential association, social bond and self—control
theories, to examine relationships between drug use and
delinquency and race and ethnicity (which include non—
Hispanic whites, blacks Americans, Asian Americans, and
Hispanic Americans7) in the United States. By using the
2001 National Household Survey of Drug Abuse dataset, this
study will explore an integrated model that accounts for
drug use and delinquency for the general population
including all racial and ethnic groups. This study will
examine the integrated model and test the causal processes

across race and ethnicity, addressing the questions: do the

 

7 Due to the small sample size of Native Americans and Hawaii Natives in the dataset. they are not
included in this study.

theoretical constructs vary across racial and ethnic
groups? Are the causal processes invariant across racial
and ethnic groups? What are the variations in parameters
across racial and ethnic groups? And what is the relative
importance of differential association, social bonds, and
self-control in explaining drug use and delinquency for
different racial and ethnic groups.

Because the socialization process and the social
environment and experiences vary for adults and youths, the
intervening factors that cause or prevent criminalities may
differ for adults and youths. Youths’ pro— or anti-
criminal patterns are more likely to be affected by parents
and school supervision, parents' child— rearing practices,
and peer association, while adults are likely to be
influenced by employment status and quality of marriage.

To understand causal processes in crime and delinquency
both of adults and youths, this study will examine all the

questions separately for adults and youths.

Organization of Dissertation

This dissertation is divided into five chapters.
Chapter Two reviews differential association, social bond

and self-control theories, and discusses the implications

46

of different approaches for explaining the relationships
between race/ethnicity and delinquent behavior. This
chapter also demonstrates how the propositions of different
theoretical approaches are incorporated within the
integrated model. Chapter Three introduces the research
hypotheses, describes the research design and methodology,
and justifies the operationalization of the independent and
dependent variables. Chapter Four presents the finding and
results of the empirical analysis. Finally, Chapter Five
discusses implications for theory, policy and future

research.

47

CHAPTER II. LITERATURE REVIEW

This study focuses on the developing of an integrated
model that incorporate three major individual level
theoretical approaches of crime and delinquency:
differential association, social bond and self—control
theories. Besides the issues discussed in Chapter 1,
several reasons justify the selection of these three
perspectives. First, although there is increased attention
to the macro-level approach, which emphasizes the effects
of community structures and community cultures on
differential rates of crime (Sampson and Wilson 1995;
Sampson and Lauritsen 1997), studies of community level
factors cannot reveal the influences on individual
decisions to break the law. Criminality of a group is a
summery statement of the number of persons in the group who
commit crimes and the frequency with which they commit
crimes, thus the explanation of the criminal behavior of
the person is consistent with the explanation of a crime
rate (Sutherland and Cressey 1966). The relationships
between macro—level, individual-level factors and crime as
presented in Figure 2 indicates that even though there
exists a strong correlation between community level factors

and group crime rates, this correlation does not

48

necessarily reflect a causal relationship; rather, the core
of the relationships is at the individual level, when a
person is influenced to break the law. Second, the purpose
of this study is not to explain why individuals or racial
groups differ in the type of associations they have, or why
individuals vary in social bond and self-control. Instead,
this study focuses on the extent to which they differ in
associations, social bonds and self—control; whether this
variation could account for the racial and ethnic disparity
of crime; and whether the causation of crime is the same

across racial and ethnic groups.

Figure 2. Relationships between macro-levle, individual-level factors
and crime

Individual level
factors: dtfl'crcnt
zmsocnation, social

bond, and self- I criminality
/ control etc, \‘
Group

....................................... p crime rates

Individual

Community
levelfactors:
low economic
status, ethnic

49

Third, differential association theories and control
theories are the two dominant sociological/social
psychological perspectives on crime and deviance (Cullen
and Agnew 1998; Erickson et al. 2000; Massey and Krohn
1986). Two of the social control theories, social bond and
self-control theories are the most frequently discussed and
tested of all theories in criminology (Stitt and Giacopassi
1992; Li, 2004). In a recent review of research on the
correlates of juvenile delinquency, Edwards (1993) found
that social control theory was used in about 75% of the
studies reviewed. The book, “A general Theory of Crime,”
which developed self-control theory, ranks second in
citations accorded to all books in the 19905 (Cohn and
Farrington, 1999). A few previous studies examined the
theoretical competition between the differential
association and social control perspectives (e.g., Aseltine
1995; Mastsueda and Heimer 1987; Costello and Vowell 1999),
and some research joined together principles from each
approach (Erickson et al. 2000; Elliott et al. 1985; Massey
and Krohn 1986; Li 2004) to understand deviant behavior.
However, little research has examined the three competing
explanations of the race and crime relationship (the one
exception is Matsueda and Heimer 1987). This study will

contribute to the available literature by integrating

50

across the differential association, social bond, and self—
control perspectives.

This chapter first reviews the principal themes of
differential association, social bond and self-control
theories. Results of empirical research testing hypotheses
from each theory are then presented. The next part
discusses the completion and integration between the
theoretical approaches. And finally, based on the
literature review, an integrated theoretical model is

developed.

Differential Association Theories

Differential association theory was originally
developed by Sutherland in an effort to explain the
relationships between multiple factors, such as social
class, broken homes, age, race, urban or rural location,
and crime and criminality.

Sutherland viewed the nature of crime as involving
three elements: “a value which is appreciated by a group or
a part of a group which is politically important; isolation
of or cultural conflict in another part of this group so
that its members do not appreciate the value or appreciate

it less highly and consequently tend to endanger it; and a

Si

pugnacious resort to coercion decently applied by those who
appreciate the value to those who disregard the values”
(Sutherland and Cressey 1966: 15—16). When a crime is
committed, these relationships are involved. Crime is this
set of relationships when viewed from the point of view of
the group rather than of the individual.

The principle ideas of Sutherland’s differential
association theory were specified in nine propositions.
One proposition is that criminal behavior is learned in
interaction with others, especially intimate others. This
differential learning includes the techniques of committing
crime, and specific direction of motives, drives,
rationalizations, and attitudes. “A person becomes
delinquent because of an excess of definitions favorable to
violation of law over definitions unfavorable to violation
of law” (Sutherland and Cressey 1966: 81). Sutherland
pointed out that not all associations have an equal impact,
and the influence of associations varies by frequency,
duration, priority, and intensity. Definitions presented
more frequently, for a longer time of exposure, earlier in
life, and from either a more prestigious source or a more
intense relationship, receive more weight in the

differential association process (Matsueda, 1988).

52

Differential association theory is concerned with both
individual criminality and society (group) crime rates
(Sutherland and Cressey 1966). As indicated previously,
differential association at the individual level explains
the criminal behavior of a person. At the level of a
society or group, Sutherland and Cressey developed a

I

concept, “differential social organization,’ to explain the
various rates of crime. They advanced the idea that “in a
multi-group type of social organization, alternative and
inconsistent standards of conduct are possessed by various
groups, so that an individual who is a member of one group
has a high probability of learning to use legal means for
achieving success, or learning to deny the importance of
success, while an individual in another group learns to
accept the importance of success and to achieve it by
illegal means” (1966: 96). This situation is called
“differential social organization.” The extent to which
the group or society is organized in favor of crime versus
against crime determines its rate of crime. Sutherland and
Cressey (1966: 83) further pointed out that the individual
differential association explanation of personal
criminality and the society or group level differential

social organization explanation of crime rate should be

consistent with each other, “since the crime rate is a

summary statement of the number of persons in the group who
commit crimes and the frequency with which they commit
crimes.”

Taken together, the individual and group components of
differential association explain the differential
association process by which race and ethnicity influence
criminal behavior (Matsueda and Heimer 1987). According to
the differential association theory, the race and ethnicity
ratios in crime result from differential association with
criminal and anti—criminal patterns. The differential
association process should be invariant across race and
ethnic groups; however, it is affected by social and
economic levels at which members of a race or ethnic group
must live, and it is also affected by the normal conflict
between its traditional ways of behaving and the laws under
which the group is expected to live (Sutherland and Cressey
1966). Social conditions such as region, sex, age,
education, family status, and location of residence etc.
have impact on the magnitude of the differences of crime
rate across race and ethnic groups as well.

Differential association theory, as one of the most
important sociological criminal theories, has stimulated
theoretical refinements and revisions, empirical research,

and applications to programs and policy through out the

years (Matsueda 1988). Cressey (Sutherland and Cressey
1966) identified two major weaknesses in differential
association theory. The first was the difficulty of
defining the concept of definitions. And because the
concept of definitions was not precisely defined, the ratio
of the “excess of definitions” could not be
operationalized. Second, differential theory left the
learning process unspecified, providing no clue about what
in particular would be included in “all the mechanisms that
are involved in any other learning.” These two major
weaknesses caused difficulty in conducting empirical tests
of differential association theory (Glueck 1956). The
criticism that the theory is untestable was partly due to
the limitation of measurement and statistical techniques at
that time. For example, another earlier criticism of
differential association theory is that it is unclear which
came first-the association with delinquent others or the
illegal behavior itself (Glueck and Glueck 1950).
Delinquents may select as friends other youths whose values
and behaviors are similar to their own. If that is the
case, then delinquency causes delinquent friends, but
delinquent friends do not cause delinquency. This
difficulty was solved by later longitudinal research; and

findings support differential association theory (Tittle et

LI!
U!

al. 1986). In a paper that evaluated the current status of
differential association theory, Matsueda (1988) indicated
that with the advances in statistical techniques, the ratio
of definitions favorable and unfavorable to violating the
law can be measured, and differential association theory

can be tested empirically.

Empirical Research on Differential Association Theory

Tests of differential association theory have centered
on several areas. These include analyses of the main
concepts of differential association: peer relationships,
and ratio of definitions favorable and unfavorable to law
violation, and social factors that influence differential

associations.

Association with Delinquent Peers

 

Early studies of differential association theory
operationalized the concept of association as the
association with delinquent peers, and the frequency,
duration, priority and intensity of such associations (e.g.
Short 1957, 1958). Research consistently has demonstrated
that peers’ delinquency is a strong predictor of

individuals’ delinquency (Gueck and Glueck 1950; Reiss and

Rhodes 1964; Voss 1964; Krohn 1974). As early as 1950, the
Gluecks found that 98% of 500 delinquents had delinquent
friends, but only 8% of 500 nondelinquents had delinquent
friends (Glueck and Glueck 1950). Short (1957) designed a
“specific differential association score” to reflect the
cumulative effect of frequency, duration, priority, and
intensity of interaction with delinquent friends. A
significant correlation (.581 for boys, and .606 for girls)
between the “differential association score” and
delinquency was found. The correlations between delinquent
peer association and individual delinquent behavior have
been replicated in research using different measure or
modeling designs (Akers et al. 1979; Matsueda 1982; Tittle
et al. 1986; Elliott et al. 1985; Krohn et al. 1996; Warr
1993a, 1993b; Haynie 2002). For instance, Reiss (1964) and
associates measured delinquent peer associations by asking
respondents, on a four—point scale, how many of their
friends had committed a property crime in the past year.
They found that the number of friends who had committed a
property crime is significantly correlated with the
individual’s involvement of property crime. In a very
recent study Haynie (2002) used a relative measure of peer
delinquency—the proportion of delinquent friends in a

respondent’s network—to test differential association

57

theory. Haynie found that the proportion of delinquent
friends in a respondent’s network is most strongly
associated with respondents’ subsequent delinquency. These
findings again support the proposition that “delinquent
peers” is a significant entity to predict delinquent
behavior.

Although the concept of delinquent peers association
is highly correlated with the concept of associations with
definitions favorable and unfavorable to delinquency, a
criticism is that this strategy does not measure the
concept “definitions” directly. Several researchers
developed a strategy to measure the content of definitions
favorable to crime more directly. For instance, Cressey
(1952) used verbalization and rationalization to make up a
component of definitions favorable and unfavorable to
delinquency. Sykes and Matza (1957) developed a measure of
techniques of neutralization to demonstrate pro—delinquent
definitions. Later researchers have developed survey
instruments to measure learned definitions of law violation
based on these refinements (Jensen 1972; Matsueda 1982;

Tittle, Burke, and Jackson 1986; Matsueda and Heimer 1987).

Qgfiipitions Favorable and Unfavorable to Delinquency

 

The concept of definitions favorable and unfavorable

to delinquency is the core of differential association
theory. It is predicted that the ratio of definitions
favorable and unfavorable to delinquency affects the
likelihood that a person will become delinquent. Mixed
evidence has been found from previous studies. An early
study by Jensen (1972) reported that there is not an
independent effect of delinquent definitions on
delinquency. Matsueda (1982) pointed out that the Jensen’s
results might be influenced by measurement error in the
indicators of the ratio of definitions; they also might be
affected by other social conditions such as age, social
class, and broken home etc. In a study that has been
described as a “classic contribution to theory testing in
criminology” (Hagan 1992z4), Matsueda (1982) established a
structural causal model. “Definitions favorable and
unfavorable to delinquency" is treated as an unobservable
variable with multiple indicators. The theoretical
construct of the “ratio” concept in this study is
transformed to a monotonic measure to capture associations
with both procriminal and anticriminal behavior patterns.
The indicators include beliefs about police officers,
delinquency or deviant behavior, and law. The results were
consistent with differential association theory. The

effects of social conditional factors, parents’

supervision, and peer relationships, as well as social
control variables, were mediated by the ratio of
definitions favorable and unfavorable to delinquency.
Similar findings have been obtained from other studies
using measures of the “ratio of pro-crime to anti-crime
definitions” (e.g. Tittle et al. 1986; Matsueda and Heimer
1987).

Sutherland and Cressey (1966) indicated that social
economic conditions such as region, sex, age, education,
family status, and location of residence have an impact on
the ratio of definitions pro-crime to anti—crime, which in
turn influencs delinquent behavior. Low social economic
status may increase the probability of encountering many
delinquent behavior patterns; it may also affect a child’s

denial or acceptance of conventional values.

FamilyVStructure

 

Research on delinquency usually finds that female-
headed households, unemployment, illegitimacy, and
differential association influence the crime rates.
Previous studies testing differential association theory
have used broken home (a dichotomous variable indicating
whether the adolescent lived in a two—parent family) as an

indicator of family structure (Matsueda 1982; Matsueda and

60

Heimer 1987; Haynie 2002). Broken homes may promote
delinquency by increasing children’s association with
delinquent definitions, and decreasing their association
with anti-delinquent definitions (Sutherland and Cressey
1978). Matsueda and Heimer (1987) found broken homes do
have impact on delinquency. However, the broken home’s
influence was mediated by learning of definitions of
delinquency, and this impact is lager on delinquency among
backs than non-blacks. But in a multivariate analysis,
Haynie (2002) did not find significant effects of broken
homes after controlling peer delinquency, peer network, and
other control variables. Thus, existing research is

inconclusive.

Differential Association and Race and Delinquency

 

So far very few empirical studies have analyzed the
efficacy of differential association theory in explaining
the relationship between race and delinquency. One major
study was conducted by Matsueda and Heimer (1987). In this
study, Matsueda and Heimer examined differences in
parameters across populations of black and non-black youth,
and tested the efficacy of differential association versus
social control theory across race. They found that effects

of definitions on delinquency were invariant across race.

6|

Delinquent friends had a slightly larger total effect in
the model for non—blacks than for blacks. But definitions
of delinquency had relatively less effect in the model for
non—blacks than for blacks. The effects of background
variables on determinants of definitions varied by race.
Broken homes were more influential in producing delinquency
among blacks than non—blacks. Matsueda and Heimer (1987)
argued that this finding was due primarily to the larger
effects of broken homes, parental supervision, and
neighborhood trouble on the process of learning definitions
among blacks. Thus, from the standpoint of differential
association theory, the neighborhood and family
organization of blacks is most influential in the process
producing definitions of delinquency. The general findings
support the hypothesis that the differential association
causal models of delinquency are invariant across racial

groups.

Social Bonding Theory

Control theories assume a person is free to commit
crindnal acts because his or her ties to the conventional
ordem have been broken. In social control theories,

ckmriance is taken for granted. It is conformity that needs

to be explained (Hirschi 1969). Hirschi (1969) therefore
argued that social bond theory is designed to answer the
question “Why don’t we do it” instead of “Why do they do
it”, a question that most other theories try to answer.
According to social bonding theory, the motivation for
crime is constant across persons. It is not only those who
finally commit a deviant act who have the impulse to do so
but most people experience deviant impulses frequently
(Hirschi 1969). When an individual’s bond to conventional
society is weak or broken, delinquent acts will be
resulted. Hirschi (1969) identified four elements of the
social bond—attachment, commitment, involvement, and
belief.

According to Hirschi (1969), lack of attachment to
Otliers frees an individual from moral restraints, as the
irudividual has no conscience or superego. If a person is
irmsensitive to the opinion of others, then he or she is to
tflmat extent not bound by norms, and is free to deviate.
The essence of internalization of norms, conscience, or
Supexego thus lies in the attachment of the individual to
Others. A key idea in the theory is that commitment to
Conventional activities, such as getting an education,
building up a business, acquiring a reputation for virtue,

dlSsuades a person from crime and delinquency. When one

63

considers deviant behavior, one must consider the costs of
this deviant behavior and calculate the risk of losing the
investment of time and energy in conventional activities.
Involvement in conventional activities limits one’s time
and reduces the opportunity to commit deviant acts. The
element of belief is a reconceptualization of Sutherland’s
concept of definitions favorable to law violation.
Hirschi denied that there exists subculture of violence.
He argued that it is not necessary that a person construct
a system of rationalizations in order to justify commission
of acts he wants to commit, because in control theory there
is no special motivational force to account for the
neutralization. In contrast, control theory assumes the
existence of a common value system within the society or
grraup whose norms are being violated. It also assumes that
“triere is variation in the extent to which people believe
ttmey should obey the rules of society, and furthermore,
ttuat the less a person believes he should obey the rules,
the nmre likely he is to violate them” (Hirschi 1969: 26).
Hirschi (1969) argued that while the four elements of the
SOCial bond are correlated, each element could influence
delinquency independently and additively.

Hirschi’s control theory assumes that crimes are

available to everyone, and need no special motivation. The

64

causes of crime and delinquency are the same across racial
and ethnic groups. For both whites and minorities,
delinquency is determined by the independent effects of the
elements of the social bond. The racial disparities in the
crime rate are the results of differences in elements of

the bond to conventional society.

Empirical Research on Social Bond Theory

Since the publication of Hirsch’s Causes of
Delinquency, his original formulation has been tested,
challenged, and revised numerous times. Some studies have
reported findings contrary to predictions; there exist
several proposals to refine the original theory or to
integrate it with other theories (e.g., Elliott et al.
1985; Sampson and Laub 1993). Altogether, however,
empirical tests are generally supportive of social control
theory. Associations between indicators of attachment,
belief, commitment, and involvement with delinquency have

tended to be positive and significant (Siegel 1995:219).

Attachment
Hirschi's attachment construct includes affective and

behavioral ties to parents, school, and peers. However,

the social control influences of parents, school, and peers
are theoretically and empirically distinct. Researchers

usually treat them as separate constructs.

Attachment to parents

 

Attachment to parents is a central concept in
Hirschi’s control theory. Hirschi argued that the more
strongly a child is attached to his parents, the more
strongly he is bound to their expectations, and therefore
the more strongly he is bound to conformity with the legal
norms of the larger system. Hirschi included mother’s
virtual supervision, intimacy of communication with father
(mother), and affectional identification with father
(mother) as components of attachment to parents. The
bivariate analyses using the Richmond Youth Project data
showed these three components of attachment to parents are
significantly correlated with individual delinquency.

Some later studies distinguished between two different
dimensions of attachment to parents (Rankin and Wells 1990,
Costello and Vowell 1999). For instance, Rankin and Wells
(1990) distinguished attachment (indirect control due to
the affective bonds between children and parents) from
direct control (parental discipline and the administration

of punishments for deviant behaviors). Direct controls and

66

attachment were consistently related to various indicators
of delinquency. However, the expected interactions among

these variables were far less evident than prior theorizing

predicts.

Attachment to school

 

In analyzing attachment to school, Hirchi (1969)
presented a causal chain that runs from academic
incompetence to poor school performance to disliking of
school to rejection of the school’s authority and to the
commission of delinquent acts. He showed that dislike of
school significantly predict individual’s delinquency acts.
Some researchers (e.g. Massey and Krohn 1986) have argued
that Hirschi’s construct of attachment to school is not
clearly different from the commitment and involvement
concepts. They pointed out that some predictive items such
as school grades in commitment are redundant with each

other and with school attachment (Massey and Krohn 1986).

Attachment to peers

 

There are three indicators of attachment to peers in
HirSchi’s analysis: whether a person wants to be the kind
of Person his or her best friends are, whether a person

respect the best friends’ opinions about the important

67

things in life, and how many close friends have been
pricked up by police. Hirschi found that attachment to
peers does not produce attitudes and values conducive to
delinquency. There was no direct relationship between
delinquency and attachment to adolescent peers. Instead,
Hirschi argued that both delinquent acts and delinquent
companions are the results of low stakes in conformity.
Hirschi (1969:159) concluded, “Boy’s stake in conformity
affects his choice of friends rather than the other way
around." However, subsequent studies have failed to
replicate the findings (Conger 1976; Elliott et al. 1979;
Hindelang 1973; Krohn and Massey 1980). Studies have found
that attachment to peers positively relates to delinquent
involvement. Because the social control influences of
parents, school, and peers are theoretically and
empirically distinct, later studies usually treat
attachment to parents, attachment to peers and attachment
to schools as different constructs (Elliott et al. 1985;
Massey and Krohn 1986; Hindelang 1973; Costello and Vowell

1999).

Egflflflitment to Conventional Lines of Action

 

Two important components of “commitment” are

comnuitment to education and commitment to a high-status

68

occupation. Hirschi found that education expectation and
educational aspirations were both related to delinquency.
However, education expectations were more strongly related
to delinquency than educational aspirations. Achievement
orientation is also related to delinquent behavior.
Occupational expectation and occupational aspirations were
related to delinquency as well. Subsequent studies have
generally supported the hypotheses that attachment and
commitment are correlated with delinquency (Cernkovick and
Giordano 1987; Hindelang 1973; Hirschi 1969; Jensen 1972;
Junger and Marshall 1997; Krohn and Massey 1980; Rankin and

Kern 1994; Rankin and Wells 1990; Sampson and Laub 1993).

Involvement

 

Control theory predicts that involvement in
conventional activities is relevant to delinquent behavior,
because spending time in conventional activities removes
opportunities to commit delinquent acts. However, Hirschi
found negative relations between dating and working and
delinquent acts. Hirschi argued the negative relations
might be because youths who date are less likely to be
committed to higher education. Hirschi did find a
relationship between time spent on homework and delinquency

that was consistent with the hypothesis.

69

Research on the relationship between involvement in
conventional activities and delinquency has been less
supportive of social bond theory (Agnew 1993; Hirschi1969;
Kempf 1993). Because attachment to school, commitment to
education, and involvement in school activities are three
constructs that are highly correlated with each other, and
indicators of these three concepts are similar, later
studies usually used one construct to measure these
concepts (e.g. Costello and Vowell 1999; Elliott et al.

1985; Massey and Krohn 1986; Rodrigues and Weisburd 1991).

Belief

Hirschi examined three sets of beliefs-beliefs to law
and the legal system, beliefs to low—class values, and
beliefs to middle-class values—and their relationships with
delinquent acts. Hirschi (1969) found belief in the
validity of the law is negatively related to delinquency
and positively associated with attachment and commitment.
With respect to lower—class culture, Hirschi found there
are no significant differences between lower-class and
middle-class children. The academically incompetent
middle-class child was even more likely than the
academically competent lower—class child to accept the

norms, beliefs, and practices of the lower class. Evidence

70

that supports this conclusion is found in subsequent
research (e.g. Matsueda 1982; Maros et al. 1986; Krohn et

al. 1983; Costello and Vowell 1999).

Religiosity

 

Besides the four major elements of Hirschi’s social
bond, religiosity has been suggested by researchers as
another important entity that impacts the delinquent
decision (Johnson et al. 2000; Marcros et al. 1986; Evans
et al. 1995). Marcros et al. (1986) argue that religion is
a major and conventional institution that could serve to
control deviant impulses. Their empirical findings support
this argument. Johnson and associates (2000) hypothesized
that individual religiosity is protective in helping at-
risk youths such as those living in poor inner—city areas
to escape from drug use and other illegal activities.

Their research results indicated that youths’ religiosity
does have significant effect on various forms of deviance

among inner—city black male youths.

Social control and race and delinquency

 

Hirschi suggests that the models predicting causes of
crime and delinquency are the same across racial and ethnic

groups. In examining control theory, Hirschi presents

7l

empirical support based on white boys only. Hirschi does
mention that the results are similar for Black youths but
that the relations generally are weaker, which he
attributes to the lower reliability of the answers of his
Black respondents.

Several studies have analyzed how race or ethnic
background influences the impact of social bonding
variables on delinquency. Many of them concluded that,
overall, the results are very much alike for different race
and ethnic groups (e.g. Johnson 1979; Austin 1978;
Cernkovich and Giordano 1992; Eve 1978; Rodriguez and
Weisburd 1991; Wallace and Bachman 1991; Junger and
Marshall 1997). For example, Johnson (1979) found that
social control factors were, globally, the same for boys
and girls and for Asian and White respondents. Junger and
Marshall (1997) studied the interethnic generalizability of
social control theory for four ethnic groups, Surinamese,
Moroccan, Turkish, and Dutch boys in the Netherlands, and
found that the dimensions of the social bond were
interrelated in the same way among all four groups. They
also found that social bonding variables predicted
variations in a variety of types of delinquent involvement
and deviance among the four groups. The variables most

consistently related to delinquency among the four groups

were beliefs in conventional values, family supervision,
school conflict, and participation in unconventional
leisure activities.

On the other hand, a number of studies found that
social bonding variables do hold differing explanatory
power for various ethnic groups. For example, LeGrande and
Shoemaker (1989) conducted a replication study of Hirschi’s
(1969) study. They found that the association between
strength of social bond and delinquency is relatively
strong among rural and white youths and relatively weak
among urban and Black youths. Consistent with Hirschi’s
observations, they found that the scales measuring
dimensions of the social bond were less reliable for Blacks
than they were for whites. Similarly,

Cernkovich and Giordano (1992) found that social bond
variables do considerably better in accounting for the
delinquency of whites than for that of nonwhites, although
for both groups, control and supervision, identity support,
disapproval of peers, and instrumental communication are
important predictors. A study (Rodriguez and Weisburd
1991) using data from a survey of inner city Puerto Rican
American male adolescents indicated that among Puerto Rican
American adolescents, family bonding was a more important

influence on delinquency whereas peer bonding was less

73

important than was the case in the national sample of more

affluent white youths.

Self-Control Theory

The basic assumption about the nature of crime in
Gottfredson and Hirschi’s (1990) self-control theory is
consistent with the assumption of classic social control
theory. Gottfreson and Hirschi (1990: 89-90) asserted that
criminal acts as well as analogous behaviors provide
immediate, easy or simple gratification of desires, and few
or meager long—term benefits. They require little skill or
planning, and result in pain or discomfort for the victim.
While the social control theory is based on the idea that
the costs of crime depend on the individual’s current
location in or bond to society, the self-control theory
noticed that people also differ in the extent to which they
are vulnerable to the temptations of the moment. The
social and individual restraints on behavior exist
simultaneous (Gottfredson and Hirschi 1990: 87-88). In
social bond theory, Hirschi focuses on the control in a
person’s relation to society, but in self—control theory,

the focus is moved to control inside the individual.

74

Gottfredson and Hirschi indicated that the traits of
low self-control include impulsivity, insensitivity,
emphasis on the physical, risk-taking, shortsightedness,
and being nonverbal (1990:90). These factors affect
calculation of the consequences of one’s acts. For
instance, the impulsive or shortsighted person fails to
consider the negative consequences of his or her acts; the
insensitive person has fewer negative consequences to
consider; the less intelligent person also has fewer
negative consequences to consider and has less to lose.
People with the characteristics associated with low self-
control therefore will tend to engage in criminal and
analogous acts (Gottfredson 1990: 90-95). Gottfredson and
Hirschi emphasized that these components form a single,

unidimensional latent construct:

Since these traits can be identified prior to the age
of responsibility for crime, since there is a
considerable tendency for these traits to come
together in the same people, and since the traits tend
to persist through life, it seems reasonable to
consider them a stable construct useful in the

explanation of crime (1990:90-91).

\J
U:

Gottfredson and Hirschi (1990) further pointed out there
are two general sources of variation that affect the traits
of self-control. The first variation appears in the early
life of children, including low intelligence, high activity
level physical strength, and adventurousness. They
suggested that these individual differences may have an
impact on the prospects for effective socialization. The
second is the variation among caretakers in the degree to
which they recognized low self-control and its consequences
and the degree to which they are willing and able to
correct it. Therefore, Gottfredson and Hirschi (1990)
argued that the major cause of low self—control is
ineffective child-rearing.

According to Gottfredson and Hirschi (1990: 97), the
minimum conditions that are necessary for adequate child-
rearing include: monitoring the child’s behavior;
recognizing deviant behavior when it occurs; and punishing
such behavior. Two types of institutions play important
roles to produce effective or complete socialization:
family and school. The elements of the child—rearing model
within familial institutions suggested by Gottfredson and
Hirschi (1990: 98-104) contain: the attachment of the
parent to the child; parental supervision; recognition of

deviant behavior; punishment of deviant acts; parental

76

criminality; family size; the single-parent family; and the
mother who works outside the home. School is seen as an
alternative institution that provide effective
socialization to people who are not socialized sufficiently
by the family. Compared with the family, the school can
more effectively monitor behavior, recognize deviant or
disruptive behavior, discipline, and punish. Therefore
school could be an effective socializing agency. According
to Gottfredson and Hirschi, “mundane homework,” “not liking
school,” and “school performance" are indicators of lack of
socialization. Gottfredson and Hirschi (1990) also realize
that although low self—control can result in a tendency to
engage in crime and analogous acts, this tendency results
in a crime or analogous behavior only when the opportunity
to engage in the behavior is present.

Gottfredson and Hirschi (1990: 123-124) criticized
other sociological theorists for ignoring individual
correlates of crime in favor of social variables such as
urbanization, class, and culture. In explaining the racial
and ethnic differences in crime rate, they criticized that
subculture theorists simply assert that black culture
values violence more than white culture, and strain
theorists explain the differences by making nonwhite racial

status equivalent to membership in the lower class.

77

Gottfredson and Hirschi argued that race and ethnic
differences are due to differences in levels of self-
control. They claimed there are differences among racial
and ethnic groups in levels of direct supervision by
family, which may influence the crime rate. There also
exist large differences among racial groups in the elements
of child-rearing; and the differential child-rearing
practice is the major reason that accounts for the racial

differences of crime rate.

Empirical Research on Self-control Theory

Since the publication of “A General Theory of Crime,”
self-control theory has been the focus of extensive
academic attention and considerable debate. In general,
there is fairly consistent support for the theoretical
predictions—the lower a person’s self—control, the higher
his or her involvement in criminal behavior and acts
analogous to crime (Benson and Moore 1992; Evans etal.
1997; Grasmick eta. 1993; Pratt and Cullen 2000; Burton et
al. 1994; Gibbs and Giever 1995; Gibbs et al. 1998 Piquero
and Tibbetts 1996; Arneklev et al. 1993). Measures of low
Self—control have also been found to have statistically

Significant effects on crime and deviant behavior even

78

after controlling for measures of competing criminological
theories, such as strain, social bond, and differential
association theories/social learning theories (Brownfield
and Sorenson 1993; Burton et al. 1994; Burton et al. 1998;
Evans et al. 1997; Nagin and Paternoster 1993; Piquero and
Tibbetts 1996). However, empirical tests show that low
self-control cannot, as Gottfredson and Hirschi predict,
explain away the effects on crime of other sociological
factors, especially the effects of differential association
variables (Brownfield and Sorenson 1993; Burton et al.
1994; Burton et al. 1998; Evans et al. 1997; Winfree and
Bernat 1998; Partt and Cullen 2000).

Because there is only one major explanatory concept in
self-control theory, low self-control, a main issue
concerning the empirical research on self-control theory is
the operationalization of self—control. Partt and Cullen
(2000) in a study examining the empirical status of self-
control theory pointed out there are three general types of
measures of self-control: attitudinal measures, behavioral
measures, and a mixture of attitudinal and behavioral
measure. Partt and Cullen (2000)’s analysis indicates that
regardless of measurement differences, low self-control is
an important predictor of crime and of “analogous

behaviors.” Hirschi and Gottfredson (1990, 1993) preferred

79

a behavioral measure of self—control, and they argued that
“the best indicators of self—control are the acts we use
self-control to explain: criminal, delinquent, and reckless
acts” (Hirschi and Gottfredson 1993:49). However, a
research compared attitudinal and behavioral measures shows
that although both measures produce supportive evidence for
self—control theory, the behavioral measures provide no
better predication than do the cognitive measures.

Although Gottfredson and Hirschi suggested that the
traits of self-control should have equivalent factor
structures across different types of racial and ethnic
groups, little research is done to test this argument.
According to Gottredson and Hirschi (1990), the racial and
ethnic disparities of crime rates are mainly caused by
differences of child-rearing practices across racial and
ethnic groups, however, very few criminological studies
have examined self—control theory’s arguments about child—
rearing (Cochran et al. 1998; Gibbs et al., 1998; and
Polakowski 1994; Hay 2001; Unnever et al. 2003), and none
of these studies have analyzed the relations between
race/ethnicity, child—rearing and self-control, and

delinquency.

80

Competition Between Differential Association and Control

Theories

Differential Association Theory and Social Bond Theory

 

Hirschi (1969), in “Causes of Delinquency” divided
criminological theory into three main perspectives, which
he identified by the terms “control theory,” “strain

I

theory,’ and “cultural deviance theory”—a term for
differential association/learning approaches. He argued
that these three perspectives were incompatible, and that
they should be seen as rivals to be tested empirically
against one another.

In “Causes of Delinquency,” Hirschi (1969) analyzed
the self-report delinquency data collected from the
Richmond Youth Projects, and provided empirical evidence
against differential association theory. Hirschi (1969)
found that attachment to delinquent friends was negatively
related to individuals’ engaging in delinquent, which
Hirschi argued contradicts Sutherland’s differential
association theory. Another finding was that when the
number of delinquent friends is held constant, certain

indicators of attachment to parents, attachment to school,

and commitment to conventional achievement all affected

8]

delinquency. Hirschi therefore concluded that the
empirical results support social bond theory over
differential association theory. Matsueda (1982) pointed
out that Hirschi’s findings are based on a
misinterpretation and mis-operationalization of
Sutherland’s concept of definitions favorable and
unfavorable to criminal behavior. Matsueda argued that
Sutherland’s theory pertains to associations with behavior
patterns, which can be learned from nondelinquents as well
as delinquents. Measures of the number of delinquent and
nondelinquent friends should be used to investigate the
sources of learning delinquent and antidelinquent behavior
patterns. To test Sutherland's differential association
theory, Matsueda (1982) suggested that a more direct
operationalization of the core concept of delinquent and
antidelinquent definitions was needed.

Matsueda (1982) reanalyzed Hirschi’s data of Richmond
Youth Project, and tested Sutherland’s differential
association theory against Hirschi's social bond theory.
Matsueda found that indicators of definitions favorable and
unfavorable to law violation mediated all of the social
control and background variables’ effects on delinquency.
Matsueda therefore concluded that differential association

theory was supported over social bond theory.

A more recent study evaluated Matsueda’s analysis
(Costello and Vowell 1999). Costello and Vowell argued
that Matsueda’s social control model did not provide a
complete test of theory because of the omission of two of
the theory’s crucial constructs, commitment and intimacy of
communication between children and parents. Matsueda
operationalized attachment to parents solely with Hirschi's
“virtual supervision” variables. Costello and Vowell
operationalized Hirschi’s concept of belief using similar
indicators as Matsueda’s operationalization of Sutherland’s
concept of definitions favorable and unfavorable to law
violation. However, Costello and Vowell argued that belief
is one component of social bond; the location of belief in
the causal chain should be the same as other components.

In other words, belief should not mediate the effects of
other social bond variables. The findings from this
specific model showed that social bond and friends’
delinquency remained more important than did definitions.
Thus they concluded that social bond theory is supported
over differential association theory in explaining

delinquency.

Self-control Theory and Social Bond Theory

 

Self-control theory and social bond theory are both

83

control theories and both share the assumption about human
nature that people will naturally break the law. In self-
control theory, crime is viewed as rooted in individual
differences that developed early in life and have effects
across the life course (Gottfredson and Hirschi 1990), and
in social bond theory crime is rooted in social experiences
(Hirschi 1969). Gottfredson and Hirschi assert that self-
control, not social bonds, is the chief source of
resistance against criminal temptations. Child-rearing
through family and school impacts self-control, which in
turn influences the individual to be involved in criminal
behavior. Therefore the causal sequence in se1f~control
theory is from attachment to parents, attachment to school,
commitment to and involvement in education to self-control,
and then to delinquent behavior. Self-control should
mediate the effects of social bond variables. So far,
little criminological research has tested the causal model
indicated by Gottfredson and Hirschi (e.g. Cochran et al.
1998; Gibbs et al., 1998; and Polakowski 1994; Hay 2001;
Unnever et al. 2003). Unnever et al. (2003) did find that
parental monitoring not only increased self—control, but
also had direct effects on delinquency. This result
indicates that social bond variables have independent

effects on delinquent.

84

Differential Association Theory and Self-control Theory

 

Several empirical tests have shown that low self-
control cannot explain away the effects of differential
association variables (Brownfield and Sorenson 1993; Burton
et al. 1994; Burton et al. 1998; Evans et al. 1997; Winfree
and Bernat 1998; Partt and Cullen 2000). In a meta-
analysis of low self-control, Pratt and Cullen (2000) found
that studies including differential association variables
in conjunction with self—control variables explained 15.3%
more variation in crime than did studies that did not
control for differential association variables. The mean
effect size of differential association variables was still
statistically significant after controlling for self-

control.

Integration of Differential Association and Control

Theories

Even though some researchers consider the causal
models derived from differential association, social bond,
and self—control theories to be competing or alternative,
others recognize that factors deriving from different

theories could contribute to the explanation of criminal

and delinquent behavior complementarily or interactively.
Some integrated approaches have been developed to model the
casual process and explain the deviant behavior. For
instance, Elliott et al.’s integrated theory combines the
strain, social control, and differential association
(social learning) theories, and specifies an etiological
sequence. In this sequence, strain, inadequate
socialization, and social disorganization are the primary
causes of weak bonding to conventional groups, activities,
and norms. Weak conventional bonding and/or high levels of
strain lead some youths to seek out and become bonded to
peer groups that provide positive reinforcements for and
modeling of delinquent behavior. And finally, bonding to
delinquent groups, when combined with weak bonding to
conventional groups and norms, lead to a high probability
of involvement in delinquent behavior (Elliott et al.
1985:65). Using the National Youth Survey dataset, Elliott
at al. tested the integrated theoretical model, and found
that association with delinquent peers was the best
predictor of delinquency. Further, low social control
increased the likelihood of association with delinquent
peers. Social control also conditioned the effect of
delinquent peers on delinquency, such that association with

delinquent peers had a much stronger effect on delinquency

86

when social control was low. Strain increased the
likelihood that individuals would be low in social control.
However, neither the strain nor social control measures had
direct effects on delinquency. The results provided
support for the integrated theory. Elliott et al.’s theory
is among the most popular integrated theories in explaining
delinquency and drug use. Subsequent models integrating
differential association theory and social bond theory,
like Elliott et al.’s model, generally have assumed that
differential association with peers is the intervening
variable, since low social control increases the likelihood
of association with delinquent peers, which increase the
likelihood of delinquency (e.g. Massey and Krohn 1986;
Marcos et al. 1986; Erickson et al. 2000). Empirical
findings support the integrated theoretical models.
Thornberry (1987) developed an interactional theory of
delinquency by combining social bond theory and
differential association theory. Thornberry’s
interactional theory, similar to Elliott et al.’s
integrated theory, assumed that weak social bonds lead to
delinquency by increasing the likelihood of association
with delinquent peers. However, Thornberry argued that
rather than relying on unidirectional causal structure,

most variables have reciprocal causal effects on one

87

another. Thornberry also pointed out conventional theories
tend to assume uniform causal effects throughout the social
structure, and ignore the person’s structural position.
Therefore, in Thornberry’s model, delinquency itself is
involved in the pattern of reciprocal causation.
Association with delinquent peers increases the likelihood
of delinquency; delinquency could also increase the
likelihood of association with delinquent peers. Only a
few studies have examined the reciprocal relationships in
Thornberry’s interactional theory. The core arguments of
Thornberry’s theory have some support and the findings
suggest that theories of crime need to consider reciprocal

effects.

The Integrate Theoretical Model

The literature review shows that none of the theories
alone could explain the criminality satisfactorily; factors
deriving from different theories would contribute to the
explanation of criminal and delinquent behavior
complementarily or interactively. Based on the literature
reView and evidence from previous research, this study
hYPchesizes an integrated theoretical model that

lnCorporates differential association, social bond, and

88

self-control theories into a single paradigm to account for
the delinquent and criminal behavior.

A well-established theoretical model of differential
association was developed by Matsueda (1982). Matsueda
found that delinquent peers and definitions of law
violation significantly affect delinquent behavior.
Matsueda also found that effects of social bond and
background variables are mediated by definitions favorable
and unfavorable to law violation. In their integrated
model explaining delinquency and drug use, Elliott and
associates (1985) also found that bonding to delinquent
peers is the only variable the exerted direct effects on
delinquent behavior. Weak conventional bonding works
through the strong bonding to delinquent friends to produce
delinquent behavior. Following these schemes, the
integrated model hypothesizes that delinquent peers and
learned definitions of drug use mediate the effects of
social bonds.

Gottfredson and Hirschi’s (1990) self-control theory
assumed that delinquent or criminal behavior is caused by a
single factor, low self-control, which is formed through
inconsistent and ineffective child-rearing. They further
point out that the institutions of family and school play

important roles to produce complete socialization. Factors

89

within familial institutions suggested by Gottfredson and
Hirschi contain: the attachment to parents, and family
structure etc. Therefore, according to the self-control
model, elements of bond to conventional society are posited
to influence low self-control, which in turn influence the
drug use and delinquent behavior. A path between self—
control and delinquent peers is also established. It is
hypothesized that low self—control would lead to strong
association with delinquent peers.

Many theorists believe that the causal processes of
crime are the same across social demographic groups. They
argued that differences in crime rate between groups are
mainly caused by independent variables identified by the
different theories, such as differences in number of
delinquent peers, differences of definitions, and
variations in social bonds and self-control. Therefore the
background variables such as sex and age are hypothesized
to operate through the main theoretical constructs. The
independent effects of differential association and self-
control had been supported in previous research.

This integrated theoretical model is presented in
Figure 3. The integrated model incorporates the theoretical
constructs from differential association, social bond, and

self-control approaches. Weak bonding to conventional

90

social institution, social economic status, and demographic
differences are hypothesized primary causes of differential
association of crime or delinquency and low self-control.
Low self-control would further lead to strong delinquent
peer relationship. And finally, differential association
combining with low self—control leads to a high probability

of involvement in drug use and delinquent behavior.

Figure 3. The integrated theoretical model

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Differential
association of
dehnquency
Demographic and
SES variables
Delinquent
/criminal behavior
Social bonding Low self-control

 

 

 

 

 

 

 

91

Summary

The results of the literature review on differential
association and control theories provide evidence that
constructs of differential association, social bond, and
self—control theories are influential in delinquency or
deviant behavior. However, the causal structure of
delinquency is still uncertain, and little research has
tested the efficacy of theoretical models across racial and
ethnic groups. The question of whether the existing
theoretical models are sufficient to explain the racial and
ethnic disparity of crime and delinquency has not been
answered yet.

This study seeks to address several issues that remain
unresolved. First, this study focuses on developing an
explanatory model that integrates differential association,
social bond and self—control theories, to understand the
connection of race and ethnicity to drug use and
delinquency.

Second, it is evident from the review that the
research on racial and ethnic disparity of crime and
delinquent is incomplete. Existing theoretical models are
usually tested for the white population in the United

States; little is know whether these models could be

applied to explain the criminality of minority groups,
especially Native, Asian, and Hispanic Americans. This
study examines the relative explanatory power of the
theoretical model and specific causal linkages by racial
and ethnic groups of whites, blacks, Asian American, and
Hispanics in the United States.

Third, the literature review indicates that
differential association theory assumes the crime rates
differences across racial and ethnic groups are the results
of different association with criminal and anti—criminal
patterns. Social bond theory argues that the racial
disparities of crime rate are the results of differences of
elements of the bond to conventional society, while self-
control theory asserts that the differences are the results
of different child-rearing practices of different racial
and ethnic groups. But few empirical studies have been
relevant to these propositions. It is undecided that to
what extent do blacks, Asian, and Hispanic Americans differ
from whites in the ratio of definitions favorable to
unfavorable of law violation. The variations of social
bonds and self-control between different groups are unclear
too. This current study seeks to improve the understanding
Of ‘the racial and ethnic disparities of the delinquent and

crirninal patterns by answering these questions and deciding

the relative importance of each variable in explaining drug
use and in explaining the racial and ethnic disparities of

delinquent patterns.

94

CHAPTER III: RESEARCH METHODOLOGY

Hypotheses

The review of the literature on race and ethnicity and

theoretical explanations of criminality suggests several

hypotheses relevant to drug use and delinquency:

H12

H22

H33

H43

Differential association and self-control affect one’s
involvement of drug use and delinquent/criminal

behavior.

Differential association and self-control mediated the

effects of social bonds.

Differential association, self-control, and social

bonds mediated the effects of background variables.

Blacks, Asian Americans and Hispanic Americans vary
from non-Hispanic Whites on drug use and

delinquent/criminal behavior.

Blacks, Asian Americans and Hispanic Americans vary

from non—Hispanic Whites on variables reflecting

H7:

H8:

H9:

differential association, social bonds, and self-

control respectively.

The effects of background variables on social bonds,
differential associations, and self-control are

invariant by race and ethnicity respectively.

The effects of social bonds on differential
associations and self-control are invariant by race

and ethnicity.

The effects of social bonds on self-control are

invariant by race and ethnicity.

The effects of differential association on drug use
and delinquent/criminal behavior are invariant by race

and ethnicity.

The effects of self—control on drug use and

delinquent/criminal behavior are invariant by race and

ethnicity.

96

The Data

The National Household Survey on Drug Abuse

 

To test the hypotheses, a secondary data analysis was
done of the National Household Survey on Drug Abuse (NHSDA)
dataset. The NHSDA series measures the prevalence and
correlates of drug use in the United States. The surveys
are designed to provide quarterly, as well as annual,
estimates. Information is provided on the use of illicit
drugs, alcohol, and tobacco among members of United States
households aged 12 and older. Questions include age at
first use as well as lifetime, annual, and past-month usage
for the following drug classes: marijuana, cocaine (and
crack), hallucinogens, heroin, inhalants, alcohol, tobacco,
and nonmedical use of prescription drugs, including pain
relievers, tranquilizers, stimulants, and sedatives.
Respondents are also asked about personal and family income
sources and amounts. Since 1997 a new series of questions
asked only of respondents aged 12 to 17 have been
introduced. These "youth experiences" items covered a
variety of topics, such as neighborhood environment,
illegal activities, gang involvement, drug use by friends,
social support, extracurricular activities, exposure to

substance abuse prevention and education programs, and

97

perceived adult attitudes toward drug use and activities
such as schoolwork. Another set of questions is asked only
for adults (people aged 18 or older). These “social
environment” items ask perceived neighborhood’s attitudes
toward drug use, friends’ drug use and attitudes, marital
status, relationship with conventional others, criminal
behavior, and feelings about smoking/using drugs. Other
questions are asked about perceived risk of using drugs,
perceived availability of drugs, driving behavior and
personal behavior, and cigar smoking. Demographic data
include gender, race, age, ethnicity, marital status,
educational level, job status, veteran status, and current
household composition. The NHSDA dataset also includes
information on several other topics which are not relevant
to the current study.

The NHSDA uses multistage sampling designs. The sample
is stratified on multiple levels, beginning with states.
Eight states are considered large sample states and
contribute approximately 3,600 respondents per state. The
remaining states are sampled to yield 900 respondents per
state. The second level of stratification divides states

into Field Interviewer (FI) Regions. The third level of

 

8 Although the sample is not a simple random sample, it is treated as a random sample when inferring the
results to the population. The overlook of the sample design may lead to biased estimates. However. the
available techniques relating to structural equation modeling could not solve this problem yet.

98

stratification divides Fl regions into area segments
consisting of adjacent census blocks. These area segments
are used as the primary sampling units. Dwelling units in
area segments are listed in a standardized order and were
selected by systematic sampling. Field interviewers visited
each sample address to determine dwelling unit eligibility,
to list all eligible persons at the address, and to conduct
interviews. Persons were selected from the address roster
using a handheld computer. To improve the precision of
estimates, the sample allocation process targeted five age
groups: 12—17, 18-25, 26-34, 35-49, and 50 and older. The
size measures used in selecting the area segments were
coordinated with the dwelling unit and person selection
process so that a nearly self-weighting sample could be
achieved in each of the five age groups. The sample design
included approximately equal numbers of persons in the 12—
17, 18-25, and 26 and older age groups. The achieved sample
for the 2001 NHSDA was 68,929 persons. The public use file
has 55,561 records due to a sub-sampling step used in the
disclosure protection procedures. The study yielded a
weighted screening response rate of 92 percent and a
weighted interview response rate for the Computer Assisted

Interview (CAI) of 73 percent.

99

Sample

The current study will use 2001 NHSDA to test the
hypotheses. Since questions about “youth experiences” are
asked only for respondents aged 12 to 17, and questions
about “social environment” are asked primarily for
respondents aged 18 or older, the 2001 NHSDA will be
divided into two SUbngUpS*y0UthS and adults—to test the
causal models for youths and adults respectively. The 2001
youth sub-sample has 17,429 persons, and the adult sub-
sample has 38,132 persons. However, among the 17429 youths
in 2001 NHSDA dataset, 15,671 went to school during last 12
months. Since the questions concerning school
participation, academic performance, and school involvement
are important indicators of theoretical constructs of
social bonding theory and self—control theory, and these
questions were not applicable to youths who did not go to
school, youths who did not go to school during the last 12
months are therefore eliminated from the analysisg. A
further examination of the dataset found that among the
students went to school 1,025 of them failed to answer
questions concerning delinquent peers and parents’

supervision. Those who did not answer one of these

 

9 Eliminating youths who did not go to school during the last 12 months might bias the sample. Youths
who dropped out of school may have different patterns of criminality. The current study could not
generalize the research results to this part of youths.

100

questions tend to not answer other related questions. The
techniques of replacing missing values are not applicable
under this circumstance. Because the dataset is a
secondary dataset, the original dataset collecting
information cannot be obtained, no further effort could be
made in the current study. The study used listwise missing
value exclusion to eliminate the missing case. This
procedural produced 13,111 valid cases, which account for
75.2 percent of original cases in the dataset. Because
there were not enough cases for Native Americans and Mixed
racial groups, they are not included in this study. The
same technique is applied to adults and produced 31,427
adults, which account for 82.4 percent of original cases in
the dataset. The final sample size of youths and adults of

each racial and ethnic group is displayed in table 2.

Table 2: Sample size by age and race/ethnicity

 

 

 

 

 

 

 

 

 

 

Race Youth Adult
Frequency Percent Frequency Percent

non-Hispanic 9,038 be 7 22,662 71.2
white

non-Hispanic 1,618 13.4 3,338 11.0
black

non-Hispanic 375 3.1 881 3.2
Asian

Hispanic 1,589 13.3 3,668 12 0
Total 13,111 100.0 31,427 100.0

 

 

 

101

 

Differential association theory and control theories
claim that the casual model to explain crime and delinquent
behavior does not change throughout the lifetime. However,
the differential association approach admits that age and
other social conditions do have impact on the differential
association with criminal and anti-criminal patterns.
Although social control theories assert that social bonds
or self-control are mainly influenced by child-rearing in
early life, researchers find that some transitions in the
life course, such as employment and marriage, have
significant impacts on criminal decisions (Sampson and
Laub, 1992). Because theory suggests that causal factors
may be different for youths and adults, to understand the
casual process and identify the most predictive factors for
each group, the current study will test the casual models
for youths and adults separately.

Due to the information collection of NHSDA, the
measures of some core concepts are different for youths and
adults. For instance, the concept of “definitions
favorable and unfavorable to law violation” is
operationalized by three constructs, parents’ definition,
friends' definition, and the subjects’ definition for
youths; and is operationalized by two constructs,

neighborhood’s definition, and the subjects’ definition for

102

adults. The measures of youths’ social bonds are also
different from the measures of adults’ social bonds. While
commitment or involvement to marriage and employment are
assumed important for adults, the attachment to parents or
school may play more important roles for youths. The
detailed measurement differences between youths and adults
will be discussed more fully below. Even though not all
measures are available for both youth and adults, by
testing models for both groups, this study is able to draw
conclusions about whether the casual processes are

consistent for youths and adults, and identify which

factors are most important for youths and adults
separately If the results indicate that factors in

adolescent and adulthood are both significant, this will
provide information for future work that can test the full
models, in some cases with the same measures, for both

groups.

Study Variables

This study examined a casual model to explain
delinquency and illegal drug use that integrates
differential association, social bond, and self-control

theories. The literature review suggests that the main

concepts of differential association, social bond, and
self-control theories cannot be observed directly;
therefore latent measurement models and confirmatory factor

analysis were used to measure these unobservable concepts.

The Dependent Variables

 

Two dependent variables were examined in this study,
drug use and delinquent behavior. Drug use is defined as
self-report past year use of marijuana, cocaine (and
crack), heroin, hallucinogens, inhalants, and
psychotherapeuticsm. Questions in the NHSDA ask whether or
not a person used a specific type of drug during the past
year instead of the frequency or quantity of the drug use.
Each item is then a dichotomous variable, with 1 indicating
use of drug in the past year, and 0 indicating no use in
the last year. The “drug use” variable is an index of
self-reported drug use in the past year-measured as a
single latent variable with six indicators, one for each

type of drug use (see Table 3).

 

' ' Herom was eliminated from the structure equation model because m Asran group. no one reported usmg

Heroin in the past 12 months. However, it was included in descriptive analysis.

104

Table 3.

Dependent variables and indicators

 

Dependent variables

Indicators

 

 

 

Youth

Adult

Measurement

 

Past year drug use

Marijuana (mrjyr)

 

Cocaine (cocyr)

 

Hallucinogens (halyr)

 

Heroin (heryr)

 

Inhalants (inhyr)

 

Any
psychotherapeutics
(psyyrZ)

Same

as youth

O=no; l=yes

 

 

Delinquent/criminal
behavior

During the last 12
months, how many
times have you gotten
into a serious fight
at school or work?

Same

as youth

 

During the last 12
months, how many
times have you taken
part in a fight where
a group of your
friends fought
against another group

Same

as youth

 

During the past 12
months, how many
times have you
carried a handgun?

Same

as youth

 

During the last 12
months, how many
times have you sold
illegal drugs?

NA

 

During the last 12
months, how many
times have you stolen
or tried to steal
anything worth more
than $50?

NA

 

 

During the last 12
months, how many
times have you
attacked someone with
the intent to
seriously hurt him or
her?

 

NA

 

The
frequency
is recoded
as O=have
not done;
l=have
done.

 

105

 

Since the NHSDA survey is primarily designed for
estimating drug use, questions about delinquent behavior
are not as complete as those in other surveys that have
been specifically designed to study juvenile delinquency.
However, there are still some measures of deviant behavior
in the NHSDA dataset. For the youths, the delinquency is
measured as a latent dependent variable with six
indicators. For the adults, criminality is measured as a
latent variable with three indicators. The original
questions asked about the frequency of delinquent behavior.
The distribution of each item is skewed, because the
majority of subjects indicated no involvement in illegal
activities, a significant majority indicated one instance
of involvement, and very few indicated involvement more
than once. Therefore these indicators are dichotomized as
0, which means no crime, and 1, which means at least one
crime in the past 12 months (see Table 3). Previous
studies usually used an equivalent composite scale to
measure drug use or delinquency by combining all the items
into an unweighted scale (e.g., Hirschi 1969; Elliott et
al. 1985). The unweighted scale or index ignores the
measurement errors and assigns an equal predictive value to
each item. It might well be that these items are not

equally reflective of delinquency or drug use. Thus, the

106

latent variable analysis may provide a better measure of

the dependent variable than summed scales.

Independent variables

 

Based on previous research and the availability of
information from the 2001 NHSDA dataset, the current study
identified a list of latent constructs, the observed
indicators, and other directly observed control variable
relevant to differential association theory, social bond

theory and self-control theory respectively (Table 4).

107

 

 

 

 

 

 

 

 

Table 4. Independent Variables
Independent Indicators Measurement
variables Youth Adult
Delinquent/ How many students you know in grade smoke How many friends smoke l—‘none‘. 2=a
criminal cigarette cigarette few; 3=most;
peers How many students you know in grade use mj/hash How many friends use 4=a|l
mfliash
How many students you know in grade drink How many friends drink
alcohol
How many students you know get drunk weekly How many friends drunk at
least once a week
Attachment NA #frlends who you share l=none;
to friends personal issues/concerns with 2=one: 3:2 or
trfriends who you spend time 3: 4=4or 5;

with

 

#fricnds who really like and
care about you

5=more than 5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Learned Parents definition: 1=neither
Definition 110w parents would feel you smoking 1+ pack NA approve nor
cigarette per day disapprove;
How parents would feel you trying mj/hash 2=somewhat
How parents would feel you using m_j."hash monthly (“531113le
How parents would feel you drinking alcohol daily 3_:5"0n81."
Friends definition: disapprove
How your friends would feel you smoking 1+ pack NA
cigarette per day
How your friends would feel you trying mj/hash
How friends would feel you using mj/hash monthly
How friends would ft’CD’OU drinking alcohol daily
Your definition:
How you feel someone your age smoking 1+ pack How you feel adults smoking
cigarette per day 1+ pack cigarette per day
How you feel someone your age trying mjfhash How you feel adults trying
mj/hash
How you feel someone your age using mj/hash How you feel adults using
monthly mj/hash monthly
How you feel someone your age drinking alcohol How you feel adults drinking
daily alcohol daily
Neighborhood definition of
drug:
NA llow neighbor feel adults
smoking 1+ pack cigarette per
day
How neighbor feel adults
trying mj’hash
How neighbor feel adults
using mjfhash monthly
How neighbor feel adults
drinking alcohol daily
Attachment Parents check if you‘ve done homework past 12 NA 1=alwaysz 2=a

to parents

months

 

Parents helped you with homework past 12 months

 

Parents made 'ou do work-’chores ast 12 months
P

 

Parents limited your amount of TV time past 12
months

 

Parents limited your time out with friends pst 12
months

 

 

Parents let you know done a good job past 12
months

 

 

few:
3=seldom;
4=never

 

108

 

 

Parents tell you they’re proud of sth you‘d done

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Attachment How you felt overall abt going to school pst 12 NA
to school months
How ofien felt school work meaningful
How important things learned pst 12 months are
going to be to you later in life
How interesting are courses at school past 12
months
Involvement # of school based activities participate last 12 NA 0=nonez
to social months l=one; 2=two;
activities # of community based activities panic last 12 3:3 or more
months
# of faith based activities participated last 12
months
# of other activities participated past 12 months
Religiosity How many religious services in past 12 months Same as youths l=strongly
My religious beliefs are very_important disagree;
My religious beliefs influence my decisions 2=disagree;
It is important that my friends share religious 3=agree;
beliefs 4=strong|y
ggree
Low self- How often do you get a real kick out ofdoing things Same as youths 1=neverz
control that are a little dangerous 2=seldom:
How often do you like to test yourself by doing 3=sometimes.
somethinga little risky 4=always
How often do you wear a seatbelt when ride front
ass seat of a car
A factor score from a list of low self-control
behavior
Opportunity How difficult to get marijuana Same as youths l=probably
How difficult to get LSD impossible;
How difficult to get cocaine 2=very
How difficult to get crack dillicult;
How difficult to get heroin 3=fair1y
difiicult;
4=fairly easy;
5=very easy
Perceived How much do people risk harming themselves Same as youths 1=no risk;
Risk of drug physically and in other ways when: 2=slight risk;
use Risk smokingl Hacks cigarette per day 3=moderate
Risk smoking my onece a month “SK 4=great
Risk trying LSD once or twice “5k
Risk trying heroin once or twice
Risk using cocaine once a month
Risk having 5 + drinks once or twice a week
Sex / / O=l‘emale;
1=male
Age / /
Household / / Ordinal
income
Family / NA 0=both parents
structure l=one parent
Education NA / Continuous
level variable
Employment NA I l=not in labor
status force
2=employed
3=part time
employed

 

 

 

4=unemployed

 

109

 

 

 

Marital NA

status

 

/ I =widowed
2=married
3=divorced
4 =never
married

 

 

 

The core concept of differential association theory is

“definitions favorable and unfavorable to violation of the

II

law,

intimate groups.

which are learned through communication,

primarily in

The operationalization of this concept

conventionally involves “delinquent peer relationships” and

“learned definitions.”

For youths,
measured by four
subject knows in
drink

Marijuana,

adults, criminal

asked how many friends

drink alcohol,

items were measured on

3=most,

and get

and 4=all).

Delinquent peers

 

delinquent peers
items that asked
his or her grade
alcohol,

and get

peers

four value

smoke cigarettes,

drunk at least once a week.

in this study was
how many students the
who smoke cigarettes,

use

drunk weekly. For

was measured by four items that

use Marijuana,
These
scales

(l=none, 2=a few,

Because the questions about youths’

delinquent peers asked about the student population at the

youth’s grade level rather than asking about friends,

it

did not reflect the individual selection process. The

respondent might not necessarily associate with delinquent

110

 

students. Therefore, this measurement of delinquent peers

is imperfectly connected to the theoretical concept.

Learned Definitions

 

Several items in the dataset pertain to definitions
that are favorable or unfavorable to drug use. For youths,
the questions are about the viewpoints of respondents,
parents and friends. For adults, the questions are about
the viewpoints of respondents and people in the
neighborhood. For example, four questions asked the
respondents’ opinion about drug use: How do you feel about
someone your age smoking one or more pack cigarette per
day? How do you feel about someone your age trying
Marijuana? How do you feel about someone your age using
Marijuana monthly? How do you feel about someone your age
drinking alcohol daily? Since these questions represent
the individual’s exposure to definitions favorable and
unfavorable to drug use, and the belief of respondents,
parents and friends were expected to be highly correlated
with each other, the “learned definitions" was measured as
a second—order construct that accounts for the covariance
among the dimensions of viewpoints of drug use from
different sources (Figure 4.1 and 4.2). The matrix of

intercorrelations of latent variables measuring the

111

viewpoints of respondents, people in the neighborhood,
parents and friends will be factorized to a secondary

factor—learned definitions of drug use.

Figure 4.1. Measurement of learned definitions of drug
use—youths

 

You feel sb smoking

 

You feel sh trying 1le

 

Your You feel sb using MJ mon
definition

 

 

You feel sb drinking daily

 

 

Frd feel you smoking

 

1.eamed Friends‘ Frd feel you trying MJ
definition definition

 

Frd feel you using MJ mon

 

 

Frd feel you drinking daily

 

 

Parents feel you smoking

 

Parents’
definition Parents feel you trying Ml

 

 

Parents feel you usmg mj mon

 

 

Parents feel you drinking daily

 

 

 

Figure 4.2. Measurement of learned definitions of drug
use—adults

 

You feel adults smoking

 

 

You feel adults trying my

   

 

  
 

Your

- , . You feel adults using mj mon
definition

 

 
 
  

You feel adults drinking daily

 

 

  
 

Learned

definition Neighbor feel adults smoking

  
 

 

   

Neighbor’s
definition

Neighbor feel adults trying my

 

 

  

Neighbor feel adults using mj mon

 

 

Neighbor feel adults drinking daily

 

 

113

Attachment to Friends

Matsueda (1982) pointed out that the ratio of the
learned definitions depends on not only delinquent but also
non-delinquent friends. However, the 2001 NHSDA dataset
does not provide information about attachment to friends
for adolescents. For adults, attachment to friends is
measured by three items that asked about number of friends
with whom the subject shares personal issues and spends
time, and who really like and care about the subject. Each
item is on a five point scales: l=none; 2=one; 3:2 or 3;

4:4 or 5; 5=more than 5.

Attachment to Parents

 

Hirschi operationalized attachment to parents with
three elements: mother's virtual supervision, intimacy of
communication with father, and affectional identification
with father. The 2001 NHSDA dataset does not provide
information identical to Hirschi’s components. Instead,
attachment to parents was measured by seven items, which
asked questions about whether parents supervise homework,
time out with friends, and similar activities (Table 4).
These items were measured on four point scales, with
l=always, 2=sometimes, 3=seldom, and 4=never. The high

value means low attachment to parents. The variable,

114

attachment to parents, was only applicable to youths.

Attachment to School

Involvement and commitment to academic activities
according to Hirschi (1969) prevent individuals from
committing delinquent acts. School is also an alternative
institution to the family that could provide effective
socialization. The variable attachment to school in this
study was measured by four indicators assessing
respondent’s feeling about going to school, whether school
work is meaningful, how important things learned in the
past 12 months are to later life, and how interesting
courses at school are. Attachment to school is also
measured on a four point scales, l=always, 2=sometimes,
3=seldom, and 4=never. A high value on this variable
reflects low attachment to school. This variable is only

applicable to youths.

Involvement

 

Involvement includes involvement in different types of
conventional activities. However, as indicated previously,
items that reflect involvement (school performance, time
spent on homework) are redundant with items that reflect

attachment to school and commitment to education ((Massey

115

and Krohn 1986). Studies usually used one construct to
measure these concepts (e.g. Costello and Vowell 1999;
Massey and Krohn 1986). Involvement in school then will
not be a separate variable in the current study. The
involvement measure is limited to social activities. Four
items assessing number of school-based activities, number
of community-based activities, number of faith-based
activities, and number of other activities during the last
12 months were the indicators of involvement. The scale
was: O=none, l=one, 2=two, and 3:3 or more. The high value
of this variable indicates high involvement to social

activities.

Religiosity

 

Religiosity was tested in previous research (Marcros
et al. 1986) to indicate involvement in a major and
conventional institution that could serve to control
deviant impulses. In this study, religiosity was measured
by four manifest variables measuring attendance at
religious services, and asking respondents’ beliefs of
religious. A high value of this variable indicates strong

beliefs of religious.

116

Education level

The three measures of attachment to parents,
attachment to school and involvement to social activities,
are not available for adults. For the adults, education
level, employment status, and marital status establish a
strong bond to conventional society. Education level was a
directly observed continuous variable, with the lowest
value (1) indicating completion of the 5Ch grade or less,
and the highest value (11) indicating higher than a
Bachelor degree. Education level was not applied to youths
because all the youths in this study were school students,
and education level is therefore highly correlated with

adolescents’ age.

Employment status

 

Employment status was a directly observed
variable. The values were: l=not in labor force,
2=employed full time, 3=employed part time, and
4=unemployed. The original employment status variable is a
categorical variable. The bivariate analysis indicated
that those not in the labor force had the lowest
probability of drug use, and those unemployed had the
highest probability of drug use. Therefore, in this study,

employment status was treated as an ordinal variable, with

117

high value showing low employment status. Employment

status was only applicable to adults in this study.

Marital status

 

Marital status was an observed variable. The values
are: l=widowed, 2=married, 3=divorced or separated, and
4=never been married. The bivariate analysis showed that
widowed and married respondents had lower probability of
drug use, and never been married respondents had the
highest likelihood of drug use. Marital status was also
treated as an ordinal variable, with the high value
indicating the person was never married and the medium
level indicating an unstable marital status. This variable

was not applicable to youths.

Self-control

 

The study used behavioral measures of self-control.
Six items asking different types of deviant behavior are
used to construct the latent variable of self-controlla
These analogous behaviors concern risky acts, using seat

belts, using steroids, and other imprudent actions12 that

 

" Among these items. 3 questions were applied to the whole sample. However, with unknown reason, the
sample was divided into 2 parts. The other 3 questions were asked separately to the two parts of sample. A
factor score of risk acts was calculated for both parts of sample. and was used as an indicator of risk acts of
self-control. Therefore, in table 4, there are only 4 indicators of self-control.

'2 These actions include drove a car more than 100 miles/hour (if age 18 or older). were injured when tried

118

according to Gottfredson and Hirschi’s argument, are able

to reflect the notion of self-control. These behaviors are
distinguished from the dependent variables by the degree of
seriousness. The high value is associated with a low self—

control.

Opportunity

 

Another variable that is important in self-control
theory is opportunity to break the law, which was measured
as the availability of drugs in this study. Five
indicators asked how difficult it was to get different
types of drug. Each item was measured on a five point
scale. The high value means the easier to get drug, and

the big opportunity.

Perceived Risk of Drug Use

 

Research has shown that perceived risk of drug use is
significantly correlated with drug use and other health
behavior change (Bachman et al. 1998; Ma and Shive 2000;
Perry et al. 2003). Perceived risk of drug use could
result from negative information about drug use and the

consequences by observing drug users in their

 

to stop a fight or an assault (if age 18 or older). ran away from home and slept on the street (if age 12-17).
were seriously injured in a fight (if age 12-17), used steroids to become more muscular.

119

acquaintanceship circles and among public figures (Bachman
et al. 1998). Previous studies found independent effects
of perceived risk on drug use. However, it might also

operate through definitions that are not favorable to

breaking the law. Six items indicated perceived risk of
drug use. The high value means a high perceived risk of
drug use.

Age, Sex, Income, and Family Structure

 

Four background variables, age, sex, household income,
and family structure are also included in the study, though
family structure is available only for adolescents.
Sutherland and Cressey (1978) pointed out that age has
important direct and indirect effects on criminality. They
also pointed out that low socioeconomic status affects
delinquency because poverty areas are often areas of high
delinquency, children from a low SES home then have greater
probability of encountering delinquent behavior patterns.
Matsueda and Heimer (1987) found that a broken family could
increase children’s association with delinquent
definitions, and decrease their associations with anti-
delinquent definitions. A broken family could also
decrease parents’ supervision and produce ineffective

socialization. In this study age is an ordinal variable,

120

with lowest value (1) indicating 12 years old, and the
largest value (17) indicating 65 years old or older. Sex
is a dichotomous variable (O=female, 1: male). Household
income is an ordinal variable. A high value indicates the
household income is high. Family structure is a
dichotomous variable, with 1=both mother and father in

household, O=both parents not present.

Structural Models

The integrated structural models (Figure 5.1 and 5.2)
specify the causal structures of the variables reflecting
the theoretical constructs of individual theory for youths
and adults separately. Based on the model developed in
Chapter Two, learned definitions, delinquent peers and
self—control plus opportunity are expected to affect drug
use and delinquent/criminal behavior directly. According
to findings from previous study (Bachman et al.), perceived
risk is anticipated to be related to drug use and
delinquent behavior directly. Elements of social bonds
including attachment to parents, attachment to school,
involvement, and religiosity operate through definitions,
delinquent peers, and self—control. Perceived risk of drug

use might also affect learned definition, which are

indicated by feeling about drug use, his/her friends’
feeling, and parents’ feeling about drug use. it is
reasonable to hypothesize that the perceived harmful
consequence of drug use will influence one’s feeling about
drug use, that is, the learned definitions which are either
favorable or unfavorable to using drugs. Self-control
would also influence the association with delinquent peers.
The background variables of sex, age, household income and
family structure are mediated by social bonds, differential
association, and self-control. In the adult model,
elements of social bonds include education level,
employment status, marital status, attachment to friends,

and religiosity.

 

 

 
 
     
   

 

 

 

Figure 5.1. Integrated model—youth
*7
Perceived Learned
d f t
risk ofdrug ’ e m ion
Sex .
. Delinquent
Attachment I peers
to parents ’7', 1}
Age / D ‘
. rug use
Attachment
to school
Household /
income
0
Family Involvemen Delinquent
structure behavior
Religiosity ‘ Self-control
Opportunity
Background 300181 Differential Drug use and
variables bonding association and delinquency
self ~contro|

to
L0)

Figure 5.2. Integrated model—adults

 

   

Perceived

      
 
   

 

  

risk ofdrug Learned
, definition
Education ‘I/
Sex level ¥§~. Criminal
[Employment Asa??? peers
:=:l!kfl'

 

 

A Status \ ‘V ’V';‘
g6 ‘

—’ 9"‘4 ‘\
Marital status é

s::::°'d——~ V v"
2%"

 

 

 

 

 

A
Attachment ‘ . .
to friends ' / Delinquent
behavror
/ ‘
Religiosity > Self-control
Opportunity
Background Social Differential Drug use and
Va'lah'e‘ bonding association and delinquency
self-control

Statistical Approach

Introduction

I

 

To examine the empirical relationships among the
rJariables described in the integrated theoretical model, a
stzructural equation model approach was used in this study.
Triis approach allows the combination of latent variables
nusasurement models with regression analyses and the
e>eamination of the full multivariate nature of the
tlieoretical model. The structural equation model approach
prxavides a method to examine the theoretical models across
:nacial and ethnic groups, and to determine the relative
inuaortance of each of the variables. The next section
:hitroduces the latent variable and second-order factor
anualysis techniques. Then the techniques of structure

e<llJation modeling and the software Mplus are presented.

lfflgggt variables

 

Social science variables can be either observable or

Luwobservable. Examples of observable variables include
gffiider, age, and race. The attributes of these variables
(330 be observed and measured directly. Examples of

u“Observable variables are attachment to family,

definitions favorable to breaking the law, and self—

These unobservable variables are called latent

control.
variables. Latent variables have two basic
characteristics. The first is that they cannot be directly

observed; rather, they must be inferred from other

observable or “manifest” variables, which are sometimes

called indicators of the latent variables. Second, the

latent variables are responsible for systematic variation

in the observables. Latent variables can be used to

<:ombine several related variables into composite measures,

iaacluding principal components, factor scores, and

ccxnponents (Bartholomew, 1987). But more importantly,

laatent variables are theoretical constructs that are often

gxlaced in a theoretical context to make causal inferences.

Secxand—order Factors
Usually confirmatory factor analysis (CFA) is used to
detearmine the relationship between a set of observed

inCLicetor variables and a set of continuous latent

Vartiables. These latent variables are likely to be

intercorrelated. If the matrix of intercorrelations of

theEse factors is factor analyzed, a new set of factors

Called second—order factors is produced. The second-order
faCtors are fewer in number and may constitute a better

biasis for establishing theoretical constructs (Guertin and

126

Bailey 1970).

Structural Equation ModelingA(SEM)

Structural equation modeling is a statistical

methodology that is an extension of the general linear

of

et'

model (GLM) and that enables a researcher to test a s -

regression equations simultaneously. Structural equation

models can also explicitly take into account the

ineasurement error and contain latent variables (Raykov and

Ddarcoulides 2000). Therefore SEM can provide researchers

yvith a comprehensive method for the quantification and

teasting of theories. There are four major characteristics

of? SEM models (Raykov and Marcoulides 2000:1-3):
the models are usually conceived in terms of

First,

not: directly measurable, and possibly not well—defined,

thexaretical or hypothetical constructs. Second, the models

usuaally take into account potential errors of measurement
in 6311 variables. Third, the models are usually fit to

matlrices of interrelationship indices (i.e., covariance or

COIfrelation matrices) between all pairs of observed

Variables, and sometimes to variable means.

Fourth, in addition to handling measurement error, SEM

Inc><:lels also enable researchers to study both the direct and

1Widirect effects of the various variables included in a

model. Direct effects are the effects that go directly from
one variable to a second variable. Indirect effects are the
effects between two variables that are mediated by one or

more intervening variables (often referred to as a

mediating variable). The combination of direct and indirect

effects make up the total effect of the explanatory
variable on a dependent variable.

Classical general linear modeling approaches encompass
such methods as regression analysis, analysis of variance,
analysis of covariance, and a large part of multivariate
statistical methods (see Marcoulides & Hershberger, 1997).
In the classical approaches, models are fit to raw data,
eand no error of measurement in the independent variables is
éassumed. As a result, regression estimates can be
nnisleading and potentially lead to incorrect conclusions.

Four types of structural equation models will be used
.111 the current study. Confirmatory factor analysis models
Mlill be used to examine the interrelationships among
fseeveral constructs. Each construct included in the model is
1Jssually measured by its own set of observed indicators.
EStructural regression models will be used to test specific
e><planatory relationships among latent variables and
C>bserved variables. Mean structure analysis will be used

‘CO compare means of latent variables across multiple

groups, and finally the simultaneous analysis in multiple

groups will be used to compare the explanatory relationship

and parameters across multiple groups.

SEM uses the chi—square test as a measure of model

fit. The chi-square test is a simultaneous test for the

null hypothesis that the residual covariance matrix is

zero, that is, each element in this matrix is zero (Baer

2002). Another measure of model fit is the Root Mean

Residual. In a standardized model, one might regard a root

Inean residual of .03 or less as representing a good fit,

.and consider values higher than .07 to be fairly poor (Baer

22002:86). Fit indices are generally included in SEM to

eexamine the model fitness. The most commonly used fit

rneasures are the Goodness of Fit Index (GFI) and Bentler’s

Diormed Fit Index (NFI). Many researchers consider a value

C>f .90 or above to represent an adequate fit (Hayduk 1987).

A wide variety of programming is now available for the

<2<>nstruction of SEM. Most major statistical packages come

Mlith such programs. Some examples include EQS, SIMPLIS,

Zkbmgs, and the SAS CALIS procedures. The current study will

‘Jsse Mplus to conduct the analyses.

Mplus

Mplus is a statistical modeling program that can
estimate a variety of models for continuous and categorical
observed variables as well as continuous and categorical
latent variables. Besides the traditional regression
modeling, Mplus can estimate a variety of models including
confirmatory factor analysis and general structural
equation modeling. Mplus also includes multiple group
analysis including mean and threshold structures. This
capability is important in the current study, which will
compare the connection of differential association, social
bond, and self—control to drug use and delinquent behaviors
across four different racial and ethnic groups. Mplus
provides several model fit measures including chi-square
tests, root mean square error (RMSEA, cut off value is
0.06), standardized root mean square residual (SRMS, cut
off value is 0.08), and fit indices CFI and TLI (Muthen and

Muthen, 2003).

CHAPTER IV: ANALYSIS OF DATA

As indicated in Chapter One, the purpose of the
research was to improve our understanding of the connection
of drug use and delinquency to race and ethnicity. The
analysis of data addressed the objective in several ways.
First, the research describes the demographic and social
economic characteristics of the sample, and analyzes the
drug use and delinquent/criminal behavior across subgroups
of Whites, Blacks, Asian Americans and Hispanic Americans.
Second, using confirmatory factor analysis, the study
examines the latent variables reflecting the various
theoretical constructs. Then the mean structure analysis is
applied to determine the variation of latent variables
means between different racial and ethnic groups. Third,
structural equation modeling is employed to examine the
integrated theoretical model to explain drug use and
delinquent/criminal behavior. And finally, using multiple
group analysis, the key hypotheses comparing parameters
between race and ethnic groups are tested. This chapter

presents the results of these analyses.

Background Characteristics and Racial and Ethnic

Disparities of Drug Use and Delinquent/Criminal Behavior

[Deascriptive Statistics for Youths

A total of 12,678 adolescents having attended school
j.r1 the past 12 months comprised the sample of youths.

ffeakale 5.1 provides the descriptive statistics for each

rwaczial and ethnic group.

 

 

 

 

'g‘able 5.1. Background characteristics and self-rer‘lorted drug use and
delinquency during past 12 months for racial and ethnic groups
(percentage)—youths
Aalan Hispanic
i-_ 511119 5:52:15} 6 3 to firms: 1_ n 3
Background characteristics
flex
Male 50.4 48 3 49.9 50.2
Female 49.6 517 53.1 49 8
Age
12 to 13 30.1 36.3 23.2 30.1
14 to 15 34.9 33.8 -6 0 34.9
16 to 17 35.1 35.4 40.8 35 0
Household income
Less than $20,000 10.0 33.9 14.4 29.4
$20,000-S49,999 34 7 42.1 2“.3 45.9
$5 ,000-S74,999 23.9 13.1 22.7 13.6
$75,000 or more 31.4 10.9 33.6 11 1
Family structure
Both mother and father in HH 76.0 41.6 86.1 68.0
Only mother in HH 16.9 46.2 9.6 24.8
Only father in HH 4 6 2.0 1.9 2.6
Neither mother nor father in 2.6 10.2 2.4 4.7
HH
Drug use
Hiallucinogens 5 3 1.7 4 0 3 6
Heroin .3 -- -- --
EVIarijuana 16.8 13.5 8.8 15.5
Cocaine 1.9 .2 5 2.0
Irihalants 1.9 .2 5 2.0
F‘xny psychotherapeutios 8.8 6 6 5.3 7.8
Any illicit drug 22.2 19.3 13.6 21.9
[Marijuana only 8.7 10.5 4.3 9 1
I llicit drug except for 13.5 8 8 9.3 12.8
marijuana
gigguency
Sarious fight at school or 17.5 25.5 11.2 23.4
Work
Ta ken part in fight where 13.8 18.5 :1 7 19.2
group fights group
Carried a handgun 3.0 4.4 1.9 3.6
Sold illegal drugs 4.0 3.9 2.1 4.0
Stolen/tried to steal 3.9 4.9 4.5 5.0
“ah-3”: hing worth > $50
. Attacked someone with 6 9 13.6 4 3 7 9
EWCEntlon to seriously hurt
Lflem
Q; number 9038 1618 37s 1539

 

The percentages of males and females for each racial
and ethnic group are generally equivalent, except that
.females account for a little more among blacks than in
rather groups (51.7 females percent vs.48.3 percent males).
Tine age distributions for each racial and ethnic group are

s;imilar. About 30 percent of the youths are aged 12 to 13
ywears old, 35 percent of the youths are aged 14 to 15 years
<DLLd, and another 35 percent of the youths are aged 16 to 17
gzeaars old. One exception is the Asian group. In the Asian
egiroup, only 23.2 percent of adolescents are 12 to 13 years
c>l.d, and 40.8 percent of adolescents are 16 to 17.
'Plaerefore, Asians are somewhat older than youth in other
reacial and ethnic groups.

The analysis of household income indicates that Black
aIdCDlescents and Hispanic adolescents tend to live in a
l<vaer income household compared with White and ASian
AITierican adolescents. About 34 percent of Black youths and
219. 4 percent of Hispanic youths lived in household with
iTN30me less than $20,000 during the last 12 months. Only
atflsut 11 percent of Black and Hispanic youths lived in
hCNusehold with income more than $75,000. On the other

llarmh about 10 percent of White youths, and 14.4 percent of
ASian youths lived in a household with income less than

$20,000. But more than 31 percent of White youths, and

more than 33 percent of Asian youths lived in households

with income more than $75,000.
Family structure differs considerably for different

iracial and ethnic groups. Most (86 percent) Asian youths

lwived in a family with both mother and father in the

lacausehold, but more than half (58.4 percent) of the Black

ywouths lived in a broken family. About 46 percent of Black

eaciolescents, 24.8 percent of Hispanic adolescents and 16.9

gpearcent of White youths lived in a family with only the

rnc>ther in household. The percentage of Asian youths who

lgived in a family with only the mother in the household was

rteelatively smaller (9.6 percent) compared with all the

c3t:her groups. The descriptive examination hence shows tn

ZXELian youths and White youths lived in households with

ITEJAatively higher social economic status comparing with

E3léack and Hispanic youths. Youths in Asian and White

(glfcnups might have had experienced more supervision from
Phalfents and established stronger social bonds than Black

arlci Hispanic youths could have.

The study considered the use of six types of drugs:

HEillucinogens, heroin, marijuana, cocaine, inhalants and

any psychotherapeutics. White youths reported the highest

OVerall illicit drug use rate (22.2 percent), followed by

Hispanic adolescents (21.9 percent). The Black youths

reported lower overall illicit drug use rate (19.3 percent)
than Whites and Hispanics did. Asian adolescents had the
lowest drug use rate (13.6 percent) during the past 12
rnonths. Black youths reported the highest rate (10.5
Enercent) of Marijuana as the only type of drug used in the
§>ast 12 months. In contrast, Black youths reported the
.lcowest rate (8.8 percent) of other illicit drug except for
Ddearijuana use. Very few White youths reported heroin use.
rdco Blacks, Asians or Hispanics reported heroin use.
'Tlnerefore this item was eliminated in the later Structural
éecquation analysis.
'Ihable 5.1 presents the six types of juvenile delinquent
k>eahaviors examined in this research. Black youths reported
‘tlie highest overall levels of delinquency. They had been
rhostlikely to be involved in assaults. For example, 25.5
EDEzrcent of Black youths reported that they had gotten into
51 53erious fight at school or work; 18.5 percent of Black
)"3Liths reported having taken part in a fight where one
SJIKDup fights another, and 13.6 percent reported having
atitacked someone with the intention to seriously hurt them
1T1 the past 12 months. Hispanic youths had the second
highest prevalence of delinquency. Asian youths reported
the lowest prevalence of all most all types of delinquent

behavior except stealing. White youths reported the lowest

prevalence of having stolen or tried to steal anything

worth more than $50.
In summary, the demographic distributions for each

rmacial and ethnic group are similar. However, Asian youths

(arad White youths lived in households with relatively higher

satocial economic status comparing with Black and Hispanic

ywouths. More than half of White and Asian adolescents, yet

.leess than a fourth of Black and Hispanic adolescents lived

.ira household with income more than $50,000. Asian youths

«311d Black youths are very different in family structure.

bdc>st majorities of Asians lived in a family with both

rncather and father in the household, while only a third of

l3jlacks lived in an unbroken family. The drug use patterns

£1113 consistent with previous research. White youths had

tlaea highest overall illicit drug use, but Black youths had

tliea highest overall levels of delinquency. Asian youths

Thaci the lowest levels of illicit drug use as well as

de l inquency.

EE§EEEriptive Statistics for Adults

There are 30,609 adults in the sample. About 74

Emircent of the respondents are Whites, 11 percent are

Blacks, and 12 percent are Hispanics. Asians constituted

the smallest group in the sample. Only 2.9 percent of the

sampled adults are Asian Americans. The descriptive

statistics for adults are shown in Table 5.2.//The
examination of demographic and social economic
c:haracteristics for each racial and ethnic group indicated

t1hat males and females were almost the same proportions in

tine samples of White, Asian, and Hispanic groups. However,

‘tlnere was a considerable difference in the proportion of

Irmales and females for Black. About 58.3 percent of the

EBlacks in the sample were female, and only 41.7 percent of

‘tlne Blacks in the sample were male. This sex composition

C>f the sample might impact the average drug use rate and

Cirindnal behavior rate for the Black groups. Because

fremales in general are less likely to be involved in

cieaviant behaviors than are males, if there are more females

-ir1 the sample, the pooled average rate of deviant behaviors

Cinema would be expected to be lower than the rate of a

Sample with more males.
Hispanics followed by Blacks had the largest

EDEErcentage (57.5 for Hispanics, and 55.5 for Blacks) of

YCDUng adults aged between 18 and 25 years old. Hispanics

arld Asians had the lowest percentage (6.1 and 6.0) of

achults aged 50 or older. In contrast, Whites had the

largest percentage (15.1) of adults aged 50 or older, and

the lowest percentage (45.4) of young adults aged 18 to 25

years old.
The average household income of Blacks and Hispanics

evas lower than that of Whites and Asians. Less than 10

53ercent of Blacks and Hispanics lived in a household with

j.ncome of $75,000 or more. However, about 27 percent of

[asians, and 22.3 percent Whites lived in a household with

irgcome of $75,000 or more.

Asians had the highest average education level. Less

‘tlaan 6 percent of Asians had less than high school

eeciucation, 48 percent of Asians graduated from high school

car went to college (but did not obtain a Bachelors degree),

earid about 46 percent obtained Bachelor degree or higher

éeciucation. Hispanics had the lowest average education

.leavel. About 36 percent of Hispanics did not graduate from

high school, and only about 8 percent of Hispanics obtained

51 IBachelor degree or higher education. The distribution of

marital status is similar for each racial and ethnic group.

OBLIt Blacks had the largest percentage of never married
FKEOple. There were substantial differences in employment

Silatus between racial and ethnic groups. Blacks had the

highest unemployment rate (6.9 percent), and Whites had the

lowest unemployment rate (3.0 percent).

139

Table 5.2. Background characteristics and self-reported drug use and
delinquency during past 12 months for racial and ethnic groups
(percentage)*adults

 

—i

 

Asian Hispanic
{431.229 13.192139 “19.;912 smegma-vs
.Background characteristics
:Sex
Male 4‘ 5 41.7 51.6 4
Female 51.5 58.3 ".4 5
Fuse
18 to 25 45.4 55.5 52.8 57.5
25 to 34 15.3 14.9 21.1 20.0
35 to 49 24.2 19.8 20.1 16.4
50 or older 15.1 9.8 6.0 6.1
Picousehold income
Less than $20,000 20.3 36.4 24.6 33.4
$20,000-549,999 38.3 41.8 27.9 47.2
550,000-374,999 9.0 12.2 20.4 11.3
$75,000 or more 22.3 9.6 27.0 8.2
ESciucation level
Less than high school 12.0 20.4 5.9 35.9
High school and some college 63.8 68.5 48.2 55.6
Bachelor degree or higher 24.2 11.1 45.9 8.5
Marital status
Ddidowed 2.5 2.6 .9 1.1
[harried 45.8 23.2 40.7 42.4
Edvorced or separated 8.0 8.9 2.7 6.7
rdever married 43.7 65.3 55.6 49.8
Ehrqoloyment status
PQOt in the labor force 19.8 21 3 23.3 21.0
Ehnployed 59.3 57 4 54.0 61.3
Ekart time employed 17 9 14.4 19.2 1 .7
Lhaemployed 3.0 6.9 3.5 .9
D‘ru- use
Phallucinogens 5.2 2.2 5.1 3.5
Heeroin .3 .4 -- .2
Pharijuana 17.6 16.7 9.5 12.6
Ck3caine 3.6 1.9 1.4 3.7
Irnhalants 8.5 4.3 5.3 5.9
IKFQy psychotherapeutics 8.5 4.3 5.3 5.9
Andy illicit drug 21.6 20.1 13.3 17.3
hharijuana only 9.3 12.9 4.2 7.2
Illicit drug except for 12.3 7.2 9.1 10.1
{haIYijuana
E;£§gflipal behavior
3C31d illegal drugs 2.9 4.4 1.7
Stolen/tried to steal 1.7 2.8 1.6
arl‘Ything worth > $50
. Attacked someone with 2.6 6 1 1 6 3.3
Intention to seriously hurt
them
$1 number 22662 3398 341 3668

 

 

140

The self-reported drug use patterns of adults are
similar to those for youths. Whites reported the highest
(overall illicit drug use (21.6 percent) in the past 12
rnonths, and Asians had the lowest illicit drug use (13.3
goercent). Blacks reported the highest amount (12.9
gpercent) of Marijuana as the only type of drug used in the
gpast 12 months. Conversely, Blacks reported the lowest
aanmunt of use (7.2 percent), and Whites reported the
flighest amount (12.3 percent) of other illicit drug except
f13r Marijuana use in the past 12 months. Very few reported
L138 of heroin in the past 12 months, and Asians reported no
LL58 of heroin. This item was eliminated in the later
.St:ructural equation analysis for adults.

Three types of criminal behaviors—selling illegal
CirWJgs, stealing or trying to steal anything worth more than
535(3, and attacking someone with intention to seriously hurt
hirn or her—were reported for adults. Blacks reported the
highest number of instances of all three types of crime.
ASians reported the lowest crime number in the past 12
ITlonths.

To sum up, the sex and age distributions for adults

are comparable for different racial and ethnic groups,

except the proportion of females was significantly higher

l4]

than males for Black adults. Whites and Asians had more
advantaged social economic status than did Blacks and
Hispanics. Whites and Asians had relatively higher
education level, more stable employment status, and lived
.in households with higher average income. The drug use
gpatterns are consistent with youths. White adults had the
riighest overall illicit drug use; Black adults had the
flighest overall levels of criminal behavior. Asian adults
aigain had the lowest levels of drug use and criminality.
Whites and Asians were in a relatively advantaged
saocial—economic situation compared with Blacks and
Fiispanics. According to the differential association and
emacial bond theories, we would expect that Whites and
ZXEiians are more likely associated with conventional norms,
irivolved with conventional activities, and committed to
l”ligher education, and hence established stronger bond to
CZonventional society. On the other hand, the low social
eCionomic status of Blacks and Hispanics may increase the
EDIKDbability of encountering delinquent behavior patterns,
arid weaken the attachment to conventional institutions.
Ar1a1ysis of the theoretical constructs and tests of the

h"Ypotheses are presented in the following section of this

Chapter.

I42

Confirmatory Factor Analysis and Mean Structure Analysis of

Latent Variables of the Theoretical Constructs

Analysis of Latent Variables for Youths

As described in Chapter 3, the major theoretical

constructs in differential association, social bond, and

self—control theories could not be observed directly.

Confirmatory factor analysis was applied to examine the
latent variables that represent the theoretical constructs.

The results of the full measurement model for youths are

displayed in Table 6.1. Because of the big sample size,

‘the model chi—square is relatively large, and indicates

t:hat the null hypothesis that the residual covariance

nnatrix is zero should be rejected. However, the

CZCanarative fit index (CFI) of 0.91 indicates an acceptable

fj_t measure (Hayduk 1987). The root mean square error of

approximation (RMSEA) is 0.039, and the standardized root

Infian square residual (SRMR) is 0.056. Both of the model

iffi.t measures are less than the cut off values (.06 for

Rb”ISEA, and .08 for SRMR) suggested by Hu and Bentler
(1.999). Therefore the overall measurement fits fairly
Weir

After the determination of measures of latent

Variables, mean structure analysis was conducted to provide

the comparison of theoretical constructs across racial and
ethnic groups. Table 6.2 presents the mean structure
analysis results for latent variables for youths. The
results for means of drug use and delinquency variables are
consistent with the findings from the descriptive analysis.
White youths had the highest mean score for drug use in the
past 12 months. Hispanics had significantly lower mean
score for drug use than whites, but reported more drug use
than Blacks and Asians. Black youths reported a
significantly higher level of delinquency than White
jyouths, and Asians in contrast, reported significantly
jlower level of delinquency than White youths. There was
rust a significant difference between Hispanics and Whites
iri delinquent behavior.
Although delinquent peers have been showed to be one

C>f the major predictors of deviant behavior in previous
EStIJdies, no significant differences in delinquent peer
relationships were found between racial and ethnic groups.
A”? examination of the three dimensions of learned
Chéifinitions favorable and unfavorable to drug use indicated
tliéit Asian adolescents, their friends, and their parents
53141 disapproved of drug use more strongly than other racial
earud ethnic groups. Compared with White youths’ parents,

Black youths' parents were seen as paying significantly

144

less attention to their children's drug use issues, they
were more likely to neither approve nor disapprove of drug
use. Therefore, Asian youths' learned definitions were
less favorable to drug use than other groups’ definitions.
Considering Asian youths had the lowest drug use and
delinquency rates, the findings concerning group
differences in learned definitions that support drug use

support differential association theory.

145

Table 6.1. Confirmatory factor analysis of latent variables

 

youths

 

 

Constructjitems Unstandardized Standardized R
coefficients and coefficients
S.E.
Dependent latent variables:
Drug use:
Hallucinogens (halyr) 1.000 0.580 0.337
Marijuana (mrjyr) 2.143(0041) 0.713 0.508
Cocaine (cocyr) 0.431 (0.01 1) 0.409 0.167
1nha|ants(inhyr) 0.491 (0.017) 0.313 0.098
Any psychotherapeutics (psyyr2) 1.188 (0.027) 0.524 0.274
Delinquency:
# Have you gotten into a serious fight at school or 1.000 0.449 0.202
work?
# Have you taken part in a fight where a group of 0.789 (0.024) 0.391 0.153
your friends fought against another group
# Have you carried a handgun? 0.381 (0.013) 0.380 0.144
# Have you sold illegal drugs? 0.705 (0.021) 0.639 0.409
# Have you stolen or tried to steal anything worth 0.495 (0.016) 0.434 0.188
more than $50?
# Have you attacked someone with the intent to 0.800 (0.025) 0.527 0.278
seriously hurt him or her? (During last 12 months)
Independent latent variables:
Delinquent peers:
How many students you know in grade smoke 1.000 0.785 0.617
cigarette
How many students you know in grade use mj/hash 1.066 (0.012) 0.792 0.628
How many students you know in grade drink alcohol 1.230 (0.013) 0.838 0.702
How many students you know get drunk weekly 1.1 16 (0.012) 0.819 0.671
Your definition:
You feel someone your age smoking 1+ pack 1.000 0.719 0.517
Cigarette /day
You feel someone your age trying mj/hash 1.366 (0.015) 0.879 0.772
You feel someone your age using mj/hash monthly 1.464 (0.016) 0.933 0.871
You feel someone your age drinking alcohol daily 1.149 (0.013) 0.775 0.600
Friends definition:
Friends feel you smoking 1+ pack cigarette per day 1.000 0.754 0.436
Friends feel you trying mj/hash 1.234 (0.013) 0.889 0.564
Friends feel you using mj/hash monthly 1.269 (0.013) 0.936 0.704
Friends feel you drinking alcohol daily 1.032 (0.01 1) 0.775 0.486
Pa rents definition:
Parents feel you smoking 1+ pack cigarette per day 1.000 0.660 0.568
Parents feel you trying mj/hash 1.105 (0.019) 0.751 0.790
Parents feel you using mj/hash monthly 1.084 (0.017) 0.839 0.876
Parents feel you drinking alcohol daily 0.983 (0.016) 0.697 0.600
Definition (a second-order factor):
Your definition 1.000 0.866 0.751
Friends’ definition 1.100 (0.017) 0.849 0.721
Parents’ definition 0.348 (0.008) 0.516 0.266
Attachment to parents:
Parents check if you’ve done homework past 12 1.000 0.615 0.378
r“Onths
‘_\Parents helped you with homework past 12 months 1.063 (0.031) 0.648 0.419

 

146

 

Parents made you do worlo’chores past 12 months
Parents limited your amount ofTV time past 12
months
Parents limited your time out with friends pst 12
months
Parents let you know done a good job past 12 months
Parents tell you they’re proud of sth you’d done
Attachment to school:
How you felt overall abt going to school pst 12
months
How often felt school work meaningful
How important things learned pst 12 months are
going to be to you later in life
How interesting are courses at school past 12 months
Involvement:
# of school based activities participate last 12
months
# of community based activities partic last 12 months
# of faith based activities participated last 12 months
# of other activities participated past 12 months
Religiosity:
Past 12 months how many religious services
My religious beliefs are very important
My religious beliefs influence my decisions
It is important that my friends share religious beliefs
Self-control:
Freq get a real kick out of doing things a little
dangerous
Freq test yourself by doing something a little risky
Freq wear a seatbelt when ride front pass seat of a
car
A factor score from a list of low self-control
behavior
Opportunity:
How difficult to get marijuana
How difficult to get LSD
How difficult to get cocaine
How difficult to get crack
How difficult to get heroin
Perceived risk:
Risk smoking 1 + packs cigarette per day
Risk smoking mj once a month
Risk trying LSD once or twice
Risk trying heroin once or twice
Risk using cocaine once a month
wk havingS + drinks once or twice a week

0.256 (0.014)
0.705 (0.021)

0.443 (0.020)

0.849 (0.021)
0.814 (0.020)

1.000

1.109 (0.023)
1.090 (0.022)

1.167 (0.020)
1.000

1.202 (0.028)
1.984 (0.061)
0.863 (0.030)

1.000
0.756 (0.012)
0.827 (0.014)
0.543 (0.012)

1.000

0.908 (0.015)
-0.574 (0.020)

1.133 (0.027)

1.000
1.130 (0.013)
1.332 (0.016)
1.377 (0.017)
1.144 (0.015)

1.000

2.093 (0.071)
2.948 (0.095)
3.000 (0.0961)
2.789 (0.090)
1.258 (0.049)

0.196
0.404

0.251

0.613
0.584

0.580

0.679
0.684

0.694

0.418

0.489
0.745
0.405

0.534
0.838
0.899
0.609

0.518

0.498
-0.342

0.563

0.629
0.802
0.917
0.928
0.846

0.305
0.454
0.752
0.805
0.752
0.312

0.039
0.163

0.063

0.376
0.341

0.461
0.468

0.481

0.175

0.239
0.556
0.164

0.285
0.703
0.808
0.370

0.269

0.248
0.117

0.317

0.396
0.643
0.841
0.861
0.716

0.093
0.206
0.566
0.649
0.566
0.097

 

N Ote: Model chi-square=33070.9, df=1675. CF19910. TLl=.902. RMSEA=.039. SRMR=.056.

147

Table 6.2. Comparison of means of latent independent and dependent variables between

racial and ethnic groups—youths

 

 

 

Black Asian Hispanic Code
Means Means Means
(SE) (SE) (SE)
Dependent variables
Drug use -.033*** -.033*** -.014*** It is coded so that the high values =
.002 .005 .003 more drug use
Delinquency 017*" -.028*** .006 It is coded so that the high values =
.004 .006 .004 more delinquency
Independent variables
Delinquent peers .012 -.032 -.001 It is coded so that high values = more
.015 .029 .015 delinquent peers
Your definition -.015 .071 ** -.011 It is coded so that high values =
.014 .025 .014 stronger disagree with drug using
Friend definition -.010 .123**"‘ .015
.016 .025 .016
Parents definition -.045*** .043“ -.006
.01 l .013 .009
Attachment to parents 038* 222‘" 077*“ It is coded so that the high values =-
.019 .036 .019 less attached to parents
Attachment to school -.253*** -.188"* -.209*** It is coded so that the high values =
.013 .023 .014 less attached to school
Involvement 026* -.026 -.150*** It is coded so that the high values =
.013 .022 .013 more involvement in social activities
Religiosity 243*“ -.012 -.006 It is coded so that the high values =
.023 .044 .023 stronger religiosity
Self-control -.388*** -.294"‘“ -.262*’” [t is coded so that the high values =
.019 .036 .020 low self-control
Opportunity -.019 -.213*** -.024 It is coded so that the high values =
.024 .040 .023 easy access to drugs
Perceived risk -.001 -.018 .004 It is coded so that the high values =
_ .006 .01 1 .006 perceived greater risk

 

Note: the control group is whites. All the means of latent variables are set to zero for whites.
*p < .05; "p < .01; ”*p < .001

148

However, Asian youths were least attached to parents.

White youths had significantly stronger attachment to

parents than all the other racial and ethnic groups.

Contradictory to Hirschi's (1969) findings, this study

found that Black youths reported the strongest attachment

to school followed by Hispanics, and then Asians. White

youths were least attached to school in the past 12 months.
The examination of the involvement variable also found that

Black youths were most likely to be involved in social

activities. There was no significant difference in

involvement between Asian and White youths. Hispanic

jyouths were least likely to be involved in social
acrtivities. Black youths reported stronger religious

bmeliefs than White youths. No significant differences in

reeligiosity were found between White and the other groups.
'Ttle analysis of variables representing social bonding
lltueoretical constructs showed that Black youths reported

Strong attachment to school, more involvement in social

aKStLivities and strong religious believes, which does not

Sllpmmmt Hirschi’s (1969) assertion that blacks had weak

bOnds to conventional institutions.
White youths had the lowest mean value for self-

CCDntrol. Black, Asian, and Hispanic youths all had

Significantly higher mean values for self-control than did

149

White youths, and Blacks had the highest mean value for
self-control in this study. Asian youths thought it was

difficult to obtain access to drugs. The mean value of
opportunity to access drugs reported by Asians was
significantly lower than that reported by White youths.
According to Gottredson and Hirschi’s (1990) self-control
theory, we would expect that Asians have the lowest
probability to be involved in delinquent behavior, which is
evidenced in this study.

In general, the variables of theoretical constructs
vary significantly across racial and ethnic groups, except
(delinquent peers. Asian youths had strongest beliefs of
(definitions unfavorable to law violation. Whites had the
lxowest self-control, and Blacks in contrast, had the
liixghest level of self-control. Whites reported stronger
at:tachment to parents, but they were less attached to
SCZhool and religiosity than other groups. Blacks reported
StZrongest attachment to school, religiosity, as well as

irlvolvement. No significant differences were found between

‘Clle perceived risk of drug use across racial and ethnic

(groups.

Eglalysis of Latent Variables for Adults

The confirmatory factor analysis of latent variables

for adults is presented in table 6.3. The sample size for
adults in this study is 30,609. With such a big sample
size, the model chi—square is relatively large. The
measurement model chi-square is 46106.705 and degree of
freedom is 754. Based on this statistics, it is
appropriate to reject the null hypothesis that the residual
matrix is zero. However, the CPI is 0.910, which is bigger
than the cut off value of 0.90. The root mean square error
of approximation (RMSEA) is 0.044, and the standardized
root mean square residual (SRMR) is 0.055. Both of the
model fit measures are less than the cut off values. This
rneasurement model fits moderately well and is acceptable.
Table 6.4 presents the mean structure analysis results
fkar latent variables for adults. The drug use and criminal
phatterns are consistent with findings of the descriptive
araalysis, and are comparable to patterns for youths.
lfltiites reported significantly more drug use than other
riacial and ethnic groups. Blacks reported the highest
leevel of criminal behaviors, and their score was
S:ignificantly higher than the scores for other groups.
Z\Sians reported the lowest level of drug use and criminal
1C><ehavior in the past 12 months. A check of the delinquent
Ebeer relationship showed there were significant differences

iin number of delinquent peers between White adults and

other racial and ethnic groups. White adults reported
significantly more delinquent peers than other groups.
Asian adults had the least delinquent peers in the past 12
months. There also were significant differences of
definitions of drug use between racial and ethnic groups.
Asian, Hispanic, and Black adults disapproved of drug use
more strongly than did White adults. Compared with White
adults, Asian and Hispanic adults also had stronger
feelings that their neighbors would disapprove of drug use.
These results indicate that learned definitions for Asian
and Hispanic adults were less favorable to drug use than
for Whites adults.

Attachment to friends did not differ between White and
Asian adults, but Black and Hispanic adults reported
significantly smaller numbers of friends who cared for
them, or with whom they could share feeling. Black adults
had the highest religiosity of all the racial and ethnic
groups. Hispanics also had stronger beliefs about
religious than Whites. The mean analysis of latent
variables for adults again provides evidence against
Hirschi's assertion that Blacks have weaker bonds to
conventional society than do Whites.

The measurement of self-control shows that White

adults had significantly less self—control than did Blacks,

152

Asians, and Hispanics. The reported opportunity to access
drugs is greater for Blacks than for others. On the other
hand, Asian adults felt it was very difficult to get
different types of drugs. The mean value of opportunity is
the smallest for Asian adults. There were significant
differences in perceived risk of drug use between Black and
Hispanic adults and White adults. The perceived risk was
greater for Blacks and Hispanics than for Whites.

To be brief, the theoretical constructs also vary
significantly across racial and ethnic groups for adults.
Asian adults had the smallest number of delinquent peers,
and Whites had the largest number of delinquent peers.
Both Asian and Hispanic adults had stronger beliefs of
definitions unfavorable to law violation. Paralleling to
youths, White adults had significantly less self-control
than did Blacks, Asians, and Hispanics. Blacks and
Hispanics had weaker attachment to friends, yet stronger

beliefs of religiosity than did White adults.

Table 6.3. Confirrnatory factor analysis of latent variables—adults

 

Construct/items

Dgendent latent variables:
Drug use:

Hallucinogens (halyr)

Marijuana (mrjyr)

Cocaine (cocyr)

Inhalants (inhyr)

Any psychotherapeutics (psyyr2)
Crime:

# Have you sold illegal drugs?

# Have you stolen or tried to steal anything worth more
than $50?

# Have you attacked someone with the intent to
seriously hurt him or her? (During last 12 months)

Independent variables:
criminal peers:
# Friends smoke cigarette
# Friends use mj/hash
# Friends drink
# Friends drunk at least once a week
Attachment to friends:
#Friends who you share personal issues/concerns with
#Friends who you spend time with
#Friends who really like and care about you
Your definition:
You feel adults smoking 1+ pack cigarette per day
You feel adults trying mj/hash
You feel adults drinking alcohol daily
You feel adults driving after 1-2 drinks
Neighborhood definition of drug:
Neighbor feel adults smoking 1+ pack cigarette per day
Neighbor feel adults trying mj/hash
Neighbor feel adults drinking alcohol daily
Neighbor feel adults driving after 1-2 drinks
Definitions:
Your definition
Neighborhood definition
Religiosity:
Past 12 months how many religious services
My religious beliefs are very important
My religious beliefs influence my decisions
It is important that my friends share religious beliefs
Self-control:
Freq get a real kick out of doing things a little dangerous
Freq test yourself by doing something a little risky
Freq wear a seatbelt when ride front pass seat of a car
A factor score from a list of low self-control behavior
Opportunity:
How difficult to get marijuana

Unstandardized
coefficients and

SE.

1.000
1.856(00231)
0.736 (0.010)
0.309 (0.006)
1.078 (0.015)

1.000
0.415 (0.010)

0.455 (0.012)

1.000
1.314 (0.015)
1.399 (0.018)
1.470 (0.018)

1.000
1.399 (0.013)
1.094 (0.010)

1.000

1.347 (0.016)
1.349(0016)

0.883 (0.011)

1.000
1.573 (0.024)
1.451 (0.022)
1.220 (0.017)

1.000
0.427 (0.009)

1.000
0.559 (0.007)
0,642 (0.007)
0.547 (0.006)

1.000

0.886 (0.007)

-0.479 (0.012)
1.559 (0.052)

1.000

Standardized
coefficients

0.620
0.659
0.531
0.352

0.529

0.771
0.414

0.355

0.384
0.435
0.281
0.124
0.280

0.595
0.172

0.126

0.296
0.523
0.403
0.531

0.455
0.743
0.489

0.323
0.581
0.631
0.325

0.255
0.457
0.535
0.292

0.808
0.299

0.406
0.422
0.575
0.476

0.395
0.362
0.095
0.498

 

How difficult to get LSD
How difficult to get cocaine
How difficult to get crack
How difficult to get heroin
Perceived risk:
Risk smoking 1 + packs cigarette per day
Risk smoking mj once a month
Risk trying LSD once or twice
Risk trying heroin once or twice
Risk using cocaine once a month
Risk having 5 + drinks once or twice a week

1.275 (0.011)
1.434 (0.01 1)
1.550 (0.013)
1.284 (0.01 1)

1.000
2.898 (0.060)
2.768 (0.059)
1.945 (0.043 )
2.449 (0.053)
1.709 (0.034)

0.838
0.882
0.943
0.851

0.317
0.598
0.754
0.720
0.745
0.427

0.702
0.779
0.888
0.724

0.1006
0.358
0.569
0.518
0.555
0.182

Note: model chi-square=46106.705, df=754. CF1=0.910. TL1:0.897. RMSEA=0.044. SRMR=0.055.

155'

Table 6.4. Comparison of means of latent independent and dependent variables between

racial and ethnic groups—adults

 

Dependent variables
Drug use

 

Criminal behavior

lndej‘endent variables
Criminal peers

 

Your definition

Neighborhood
definition

Attachment to friends

Religiosity

Self-control

Opportunity

Perceived risk

Black
Means
(8.13.)

-.027***
.002
.01 3*“

.003

a088***
.008
.091m
.010
t009
.008
4406***
.015
.484m
.021
t271***
.012
.270***
.016
.038m

.004

Asian
hdeans
(S.E.)

a031***
.004
4014***
.004

-.141*”
.014
.234”
.017
129*"
.014
.018
.028
-.041
.040
-.269"””
.021
-.317***
.025
-.008
.007

Hispanic
Means
(SB)

-.020***
.002
.000
.003

a092***
.008
.154***
.010
.087***
.008
m506***
.016
.172***
.020
2270’*'
.012
a023
.015
.030***
.004

Code

It is coded so that the high values =
more drug use

It is coded so that the high values =
more delinquency

It is coded so that high values = more
delinquent peers ‘

It is coded so that high values =
stronger disagree with drug using

It is coded so that high values=more
friends

It is coded so that the high values =
stronger religiosity

It is coded so that the high values =
low self-control

It is coded so that the high values =
easy access to drugs

It is coded so that the high values =
perceived greater risk

Note: the control group is whites. All the means of latent variables are set to zero for whites.
"p < .05; "p < .01; *"p < .001

156

The following part of this chapter presents the
multivariate analysis of the model that integrated
differential association, social bond and self-control
theoretical constructs. First, the ability of the
integrated model to explain drug use and delinquency was
examined, and then the relative importance of variables
derived from different theoretical approaches was

estimated.

Structural Equation Modeling of the Integrated

Theoretical Model

Structural Equation Modeling for Youths

 

The integrated model for youths shown in Figure 5.1
was tested by using Mplus structural equation modeling
(Muthen and Muthen 2004). The measurement and substantive
models were joined together as a structural model.
Parameters were estimated using the maximum likelihood
estimator with conventional standard errors and chi—square
test statistic. The measures of model fit (displayed at

the end of Table 7.1) indicate, overall, the model fits

well (RMESA=0.041). Table 7.1 presents the unstandardized
parameter estimates of the model. Their standardized
counterparts appear in table 7.2. The path relations are

presented in Figure 6.1. These estimates reveal three
principal findings.

First, the model explains substantial variation in
drug use (R2=0.48) and delinquent behaviors (R2=O.40) for
youths. Measurements of theoretical constructs derived
from differential association and self-control theories
exert substantial independent effects on youths’ drug use
and delinquency.

As expected, delinquent peers has a significant
positive effect on drug use. The more the number of
delinquent peers an individual knows, the more likely
he/she uses drug. Learned definitions favorable and
unfavorable to breaking the law was coded so that a high
value means strong disparagement of drug use, and a
significant negative relationship between learned
definitions and drug use is found, which means the more
strongly one disapproves of drug use, the less likely one
uses drug. Low self-control has significantly positive
impact on drug use.

Accessibility of drug also shows some significant
effects on drug use, but the effect is not as strong as
other factors. The standardized coefficients in Table 7.2
indicate that the largest effect on drug use is exerted by

learned definitions, followed by self-control, delinquent

peers and opportunity to obtain drugs.

Learned definitions, delinquent peers, self—control,
and opportunity also have significant effects on delinquent
behavior. However the largest effect on delinquent
behavior is exerted by self-control, followed by learned
definitions, delinquent peers and opportunity. Notice that
delinquent peers has a negative relationship with
delinquent behavior, which is contradictory to differential
association theory. One possible reason is that the
Operationalization of the construct for delinquent peers
for youths in this study does not fully reflect an
individual’s decision about whom to hang out with. As
already noted, delinquent peers was measured by number of
students known in grade who smoke cigarettes, use
marijuana, drink or get drunk; this variables does not
necessarily tell whether one’s close friends have been
involved in these acts. The limitation of the
operationalization of delinquent peers for youths might
also affect findings about the relationships between social
bond variables and delinquent peers.

Second, the model also does well in explaining
variation in learned definition (85:0.322), delinquent peers
(R2=0.5142), and self—control (122:0.391). These variables

substantively mediate the effects of social bond variables.

As expected, attachment to parents, attachment to
school, involvement and religiosity all are significantly
related to one’s definition of drug use. The variable,
involvement, has half as strong an effect on learned
definitions as do the other variables (attachment to
parents, attachment to school, and religiosity). The
effect sizes for the other three variables are substantial
and similar to each other. Attachment to conventional
social institutions and involvement in social activities
are related to attitudes and values unfavorable to drug use
and delinquency. Background variables also impact the
learned definition. Sex and family structure have moderate
but significant effects on definition. Age shows
considerably strong connections to definitions. Being
male, living in a broken family without both parents in
household, and being old reduce the strength of
conventional beliefs and increase the likelihood of
approving of drug use.

Self-control also exerts a significant influence on
delinquent peers. Low self-control substantially increases
the number of delinquent friends. Background variables
exert sizeable effects on delinquent peers. Actually age
is the most important variable in predicting delinquent

txxers. Being old and living in a broken family increase

160

the number of delinquent peers. Surprisingly, males have
fewer delinquent peers. The finding that social bond
variables are not significantly related to delinquent
peers, as discussed previously, could be a result of the
operationalization of delinquent peers in this study. It
is possible that females reported more students using drugs
or getting drunk in their grade than males did, but they
themselves did not have more delinquent friends than did
males.

Consistent with self-control theory, attachment to
parents and attachment to school are the most important
factors for predicting self-control. Findings are
consistent with the proposition that the lack of parents’
supervising and dislike of school produce ineffective
socialization, and cause low self-control. Involvement and
religiosity also have significant effects on self—control.
Therefore, involvement in social and religious activities,
and religious beliefs are alternative institutions that
could impact the process of socialization. Compared with
other background variables, sex has the largest effect size
on self-control. Being male, living in low-income
household, and living in a broken family all are related to
incomplete socialization, and result in low self-control

behavior. Age was not significantly related to self-

161

control for youths. This finding indicates that the traits
of low self—control tend to persist during the adolescent

period.

._wo.onm2xm ._vo.ai<mm2~_ Swan—1:. .owwdnib

woo—1.1% $3.3“? ”93:3-..5 .252 3355.3 E 5255 .055 22:35. .596... .361 6:749. 232

 

 

 

new. 23.. gm. 3 m. mmm. 5.8. woo. med. 9:. Hz
82: :2:
.ttmo. .eteco. Azzszcagno
£2: 32.. 8::
{3.5.0: fiasco. “fawn. Eta—Guitar.
$2: 3.2:
{*iwucf ifiimo. WHUUQ “COST—:TUA—
$2: Amos; 9.25.3:
eeaoo... Seem _ .. $951.9.—
Eooc Ge: 22: .
31.0- .- coo. .1159. BREE—um
8:: 32: 3:3
:42. for 1:35. E35335
323 Clo... A2: 3023.
2.43.. time. 4.4mm; 9 E66522
3:: :3 :33 2:28
item. So. 1:..m _ .. 2 E98582
Ammo; watt
Elam. do 07: 333.6;
a2: :3 3:3 Ammo; 3:: 3.55 3::
eeevcr eemcfi slime. {:35 «ewe. No.1 :22: .- Bausbm bitum—
Ecoc $2: :2: 32: :33 Ex: :2: 2:85
3.1.28- Noe. mcor eNor 3:3. seemo. :ewci 2359.23:
:2: A83 35.. Ex: :2: :2: :2:
Beer *ELL. eeeccr 9.....th eegmo... sieve. 1.1.2. vu<
2:: 382 :33 3:3 3:: :3 53
eseON. see—N... unevor seem—.1 ..:.:LN.- 3:09. timer xum
523.5 .57.: was .obcozfim 33m :oEccou 0323.3— Equ.»_o>:_ 305m 2 2:23 332:;
:53?an Euzcéoa e653; E25382 2 E98532 9:53qu
333:; chquE— new 33E8> Eoncomon Ea 333:;
~=uc=u¢un=_

 

83.2qu 6.52; ”Eve—2 353:: 2: .8 356580 cu~_Eevcsm:: .: 632.

163

.33. mo. a $3. 8 EachwG 2a 3:2qu8 20m “202

 

 

 

e _. ma. bEztoaaO
rm. MN. vn. .obcoqum
flown

N..- E. 226530
:oéccwv

MN... vmr BEE:
3.- 50.- _~. bfioazum

3. Nov 2. 625202;

.028 9

on. 3. cm: E08592

3:83 9

mm. ooo. an: EoEsomc<

wEw mo x2.

2. 333.3

232:3

2..- 8.- 3. 8. 3. 8.- 8.- 3E5

oEooE

ma: woo. 25.. 8.- MN. I. 3.- 30:34.03

8.- mm. 2.- 2.- 2.- E. on. a?

.N. cw: 3.- 3.- 3 .- 2.. 3.- xom

5323 3: win 65:8 302% 52.5% 5623.3. E25202: 30:8 9 8:23 9 33%?
E26560 .tum EoncEEQ 3E3]— EoEgumz< EoEsoaz< 95522:.
new

832:3 9:23.25 25 832:; “Susanna mutant?

“covcoaon:_

 

SS,N_HZV m£=o> ”.2902 uBEwBE 2.50 356580 nofiuhmucmum .ms 2an

164

mac mo

   

 

 

 

 

 

$3232 3.-
\\\\\||I\|Iu|
35:8 A/
. _N. 2080295 wm BBQ—Em
2 - . .-
>\ 20:3 «fl 2. k Al 2.- 8 9&5;
a. .- ‘
8323 \ “// a. mo. 4
6023530 on.
4/1/ 2. 2. y busy? E 8.-
- « ‘I 2.-
MN- ‘ 8.- 4,088:—
X. 8‘. >" 62.8 8 A\. :. ‘
mo. 8. . Eon—:03; cm A” v...
0 »\»\ k \ a / .’
of: can
when >\ ’4
‘

 

3: mac (3 \
Comumc—MQQ \

\ 0:. ‘
a. \/ . )1
cm - 3:28 8 4

.- ow<
EoEzog< mm A1\\ 8 8.- ‘ A
m I‘Il cm.

3.- ‘

   

 

 

Ana—8w on. E coax—mam“. 8:
8a 828232 Enumfiwmmiv _u>o_ mo. 3 Emu—haw? b28338,“ =a £58508 352355 $053653.“ can 3: want 5:0» mo .358 @8832: :6 03.5

 

165

The third set of findings pertains to the
relationships between background variables and social bond
variables. Background variables generally have significant
and sizable effects on social bond variables. Age is most
strongly related to parents' supervision. Being male,
young, living in high—income household, and living in a
family with both mother and father in the household are
predictive of parents’ supervision. But being female,
young, and living in low—income household is related to
increased attachment to school. Females, young children,
and children living in a two-parents family are more
involved in social activities and religious services than
are males. Household income has the largest effect size in
the prediction of involvement. Children from high~income
household are more involved in social activities. But
children from low—income household have stronger religious
beliefs.

In summary, the findings support the integrated
theoretical model that delinquent peers and learned
definition of drug use, and self-control have independent
and significant effects on juvenile drug use and
ckalinguency. These variables also mediate the effects of

Social bond variables, and background variables

166

representing demographic and social economic
characteristics. The lack of parents' supervision,
weakened attachment to school, less involvement to social
activities and weak commitment to religiosity are
indirectly related to an increase in the likelihood of
juvenile drug use and delinquency. The findings are
consistent with the explanation that these relationships
occur because the independent variables increase the number
of delinquent peers, reduce the strength of conventional
beliefs, and lessen self—control. The following analysis
demonstrates whether the general model can be supported for

adults.

Structural Equation Modeling for Adults

The adult model differs from the youth model mainly in
variables representing social bond theoretical constructs.
The two major dimensions of Hirschi’s social bond theory:
attachment to parents and attachment to school, which are
also considered as the basic conditions of efficient child—
rearing and socialization in self—control theory, are not
available for adults in this study. Previous research
(Sampson and Laub 1993) indicated that life course events
SLu:h as marriage, and employment have significant impact on

a 13erson’s pro— or anti-criminal patterns. In this study,

167

education level, marital status, and employment status were
included as elements of social bond for adults, and were
expected to affect both individual’s definitions of drug
use and delinquent peers, as well as the traits of self-
control. Parameter estimates of the integrated model for
adults appear in Table 8.1 in unstandardized form, and
Table 8.2 in standardized form. The relations between the
variables are also showed in Figure 6.2. These estimates
reveal three principal findings.

First, the model explains substantial variation in
drug use (R?=O.44) but only limited variation in criminal
behavior (R?=O.l6) for adults. Delinquent peers, definition
favorable or unfavorable to drug use, and self-control have
significant effects on drug use. However, unlike the
findings for youths, definition of drug use only has modest
effect on adults’ decisions to use drug. The largeSt
effect on adult's drug use is exerted by criminal peers,
followed by self-control, perceived risk of drug use, and
definitions favorable and unfavorable to breaking the law.
The difficulty of accessing drugs is unrelated to drug use.

Definitions favorable or unfavorable to drug use does
n<>t appear to affect adult’s criminal behavior. Self—
cc>ntrol, criminal peers, and perceived risk of drug use

ek:ert substantial effects on adult’s criminal behavior.

I68 '

However, the largest effect on criminal behavior is exerted
by self-control, followed by criminal peers, and perceived
risk of drug use.

Because the perceived risk of drug use is controlled
in the model for adults, the fact that learned definition
has only a modest effect on drug use and does not affect
the legal behavior for adults is not so surprising.
Compared with youths, adults should be more mature and
rationale. Understanding the risk and consequences of drug
use may deter them from involvement of drug use despite
their own opinion, or the opinion of peers about drugs.

Notice that the relationship between criminal peers
and drug use and criminal behavior, as expected, are
positive for adults, which means more criminal peers
increase the probability of drug use and criminal behavior.
Different from the measurement of delinquent peers for
youths (which asked how many students in the grade smoke,
use drugs, etc.), the measurement of criminal peers for
adults was the number of friends smoking, using drug, or
getting drunk etc. Comparing this finding with that for
youths, the results suggest that the method of
Cuberationalization of the concept of “delinquent peers” has

ari important impact on how it would relate to the outcome

Va riables .

I69

Paralleling the findings for youths, the second
finding is that the model for adults also explains
substantial variation in learned definitions (R?=O.393),
criminal peers (B?=O.549), and self-control (R2=O.408).
Religiosity plays an important role in determining an
adult's definition of drug use, delinquent peers and self—
control. In fact, religiosity exerts the largest effect on
definition of drug use. People having strong religious
beliefs are more likely to attach to conventional norms,
and disapprove of drug use. Perceived risk of drug use
also is significantly related to definitions of drug use.
Attachment to friends, education level, marital status,
employment status, and background variables all affect
definitions of drug use modestly but significantly. The
close relationship with friends, commitment to education
and conventional marriage, and being employed are all
associated with increased strength of conventional beliefs.
On the other hand, being male, older, and living household
with low income is associated with reduced strength of
conventional beliefs.

The number of criminal peers for adults is mainly
EXplained by self-control and one’s religious beliefs. Low
self-control and weak religious beliefs significantly

increase the number of delinquent peers. Other social bond

170

factors and background variables are moderately related to
having criminal peers.

Study findings support the proposition that religion
is an important socializing institution. Weak religiosity
is related to low self—control for adults. Strong

attachment to friends also is connected to low self—

control. Marital status exerts substantial effect on self-
control. Staying unmarried and being unemployed is
significantly related to low self—control. Different from

the finding for youths, age exerts the largeSt effect on
self—control for adults. Young male adults are more likely
to have low self-control than others. This finding
suggests that age matters in adulthood in shaping self—
control. This finding and the evidence that marital and
employment status do affect one’s self-control contradict
Gottfredson and Hirschi’s (1990) assertion that the traits
of low self-control do not change through life course, and
ineffective child-rearing is the only cause of low self-
control (1990:90—91).

The third finding is that one's sex and age relate to
religiosity significantly. Female and old adults have
stronger religious beliefs than young males. Attachment to
friends does not differ significantly between male and

female. But age and household income exerts substantial

17]

influence on attachment to friends. Young adults living in
high—income households have closer relationship with
friends than old adults living in low—income households.

To sum up, similar to the findings for youths, the
findings for adults partially SUpport the integrated
theoretical hypothesis that delinquent peers and self—
control directly affect adults’ drug use and criminal
behavior. The learned definitions that support drug use
did not exert a strong effect on the dependent variables.
However, the perceived risk is significantly related to
adults’ drug use and criminal behavior. The effects of
religiosity and attachment to friends are mediated by the
differential and self-control variables. The findings are
consistent with the explanation that committing to high
education, being married, and being employed indirectly
decrease the possibility of drug use and criminal behavior
by reducing the number of delinquent peers and increasing
self-control. All these variables that represent different
theoretical constructs are related to background variables,
and age is especially important in predicting the variable

of self-control for adults.

172

Table 8. l. Unstandardized Coefficients of the Integrated Model: Adults (N=30,609)

 

Independent

 

variables and

lntervening Variables and Dependent Variables

 

lntervening Attachment Learned Delinquent Self- Delinquent
variables to friends Religiosity definition peers control Drug use behavior
Sex —.009 -.35*** -.O4*** ~.O92*** 31*“
(.0107) (.016) (.005) (.006) (.007)
Age -.04*** .056*** -.006*** -.02*** -.06***
(.002) (.003) (.001) (.001) (.002)
Household .l l*** -.02* -.0l4*** .005* .003
income (.005) (.007) (.002) (.003) (.004)
Education .008*** .0 l *** .02***
level (.001) (.001) (.002)
Marital status 026*" -.O43*** .084***
(.003) (.004) (.005)
Employment .008M -.007* .023***
status (.003) (.004) (.005)
Attachment to -.032"‘** .O4*** .096***
friends (.004) (.003) (.005)
Religiosity .18*** -.068*** -.12***
(.004) (.002) (.004)
Perceived risk .71*** -. 106*” -.07***
of drug (.02) (.005) (.005)
Learned -.023*** .004
definition (.002) (.003)
Criminal peers .l4*** .06***
(.005) (.005)
Self-control 38*“ .05”: .055...“
(.009) (.003) (.003)
Opportunity .000 .007ttt
(.001) (.001)
R2 .033 .044 .393 .549 .408 .439 .158

 

Note: *p<.05, **p<.Ol, ***p<.00 1. Standard errors appear in parentheses.
Model Chi-square= 66935.542, df=986. CFl=O.876, TLI=0.860. RMSEA=0.047, SRMR=0.074.

I73

Table 8.2. Standardized Coefficients ofthe Integrated Model: Adults (N=30,609)

 

Independent
variables and
lntervenin g

lntervening Variables and Dependent Variables

 

 

variables Attachment Learned Delinquent Self- Delinquent
to friends Religiosity definition peers control Drug use behavior

Sex -.005 -.15 -.10 -.05 .28

Age -.13 .14 -.I3 -.05 -.30

Household .15 -.0| .0! -.04 .007

income

Education .04 -.04 .07

level

Marital status -.IO .07 .15

Employment -.01 .02 .03

status

Attachment to -.06 .08 .I4

friends

Religiosity .47 -.21 -.25

Perceived risk .34 -.I8 -.I I

of drug

Learned -.08 .012

definition

Criminal peers .4] .l6

Self-control .57 .22 .23

Opportunity -.002 .04

 

174

wEv mo
£5232 \ll/

madam

no
H EuE>o_aEm
\ OM,- K
75:8 “n// 2 ‘
3, £3 33 xm. I '
\ \ {d/f/ 3. ~' mESZSEE 088:.
m... 1.-
.5323 o P ‘
Bogs—om 2.
f . ‘-
f
4/ . . 'K‘

            

        
     

 

S. S
:.- S. _ . 5.- . 4
% DR 3- Q K}\ E uw<
4. 28a Allilnll 3.. ‘ _mo_w__ox L\ m...
>\ . 3555 A! a. ‘4’ $
.4 .
‘/ K f
$.55
1/ 3. s 2=§§=< \
we. 5%
a... S e...

’\V\ we
2 . 30

3: misc xm: 338$;
82::an

.293 2: E 8.83% 8:
05 322253 ufichm_m5v _u>o_ no. on 385:»? 3323.38,“ =u £56508 nofiuaucflm .mucoscE—ou EH 3: ms 53% no ESE vows—mug: ”N6 25w:—

175

Multiple Group Analysis

Mplus Multiple Group Analysis provides the means for
simultaneous analysis of multiple groups. The software has
the capacity to test for equivalence of construct equations
and comparisons of parameters across groups, which would
otherwise be constructed using analysis of variance and

analysis of covariance techniques.

Multiple Groups Analysis for Youths

The results of the simultaneous analysis of the
theoretical model that integrated differential association,
social bond, and self-control theories across Whites,
Blacks, Asian Americans and Hispanics for youths are
presented in Tables 9.1a, 9.lb, 9.2a, 9.2b, 9.3a and 9.3b.
In this particular model, the parameters examining the
relationships between independent variables, intervening
variables, and dependent variables are set to be free
across racial and ethnic groups. To test the equivalence
of specific parameters across groups, a variety of
constrained13 models were examined, and the chi-square test

results are presented in Table 9.4. Six major findings

—¥

I3 . . .
The constrained model m this study means some parameters are forced to be equal between groups. Also
see the example in footnotes l2.

l76

emerge from this multiple group analysis.

First, the chi-square test comparing the fully
constrained model-parameters are set to be equal across
four racial and ethnic groups (Table 9.4, model 2)—with the
no constraints model (Table 9.4, model 1) indicates that
the parameters differ significantly across racial and
ethnic groups. The overall relationships between
demographic variables, variables representing theoretical
constructs for social bond, differential association, and
self—control theories, and drug use and delinquency are not
identical across Whites, Blacks, Asian Americans, and
Hispanics in the United States. Therefore the hypothesis
that all substantive parameters are the same for all racial
and ethnic groups is rejected (Ax?=1110.792, df=150,
p<.000). Further examinations of models that constrain
parameters between two specific groups (model 3.1
constrains parameters between Blacks and Whites”, model 3.2
constrains parameters between Asian and Whites, and model
3.3 constrains parameters between Hispanics and Whites)
provide evidence that the structural causation of drug use

and delinquency vary significantly between Black and White

youths (Ax?=853.007, df=50, p<.OOO), between Asian and

N Each parameter in the integrated model is forced to be equal between whites and blacks. For instance.
the coefficient ofdelinquent peers (on drug use) is forced to be the same for whites and blacks.

I77

Whites youths (Ax2=125.912, df=50, p<.000), and between

Hispanic and Whites youths (Afafl4l.906, df=50, p<.000).
The second finding is relevant to hypothesis 6, that
the effects of background variables on social bonding vary
by race and ethnicity. Models 4.1, 4.2 and 4.3 constrained
the effects of background variables on attachment to
parents between Blacks and Whites, Asians and Whites, and
Hispanic and Whites respectively. Chi—square tests indicate

there are significant differences between Blacks and Whites
(Ax;=4l.575, df=4, p<.000), and between Hispanics and Whites

(Ax;=20.355, df=4, p<.000). However, the overall effects of
background variables on attachment to parents are similar
for Asian youths and White youths (Ax;=3.848, df=4, p<.427).
The parameter estimation (Tables 9.1a and 9.1b) suggests an
interactive effect of race and sex on attachment to
parents. Being black and male significantly decreases the
attachment to parents. Sex exerts no significant effect on
Asian and Hispanic youths. There also are interactive
effects of race and family structure, and race and
household income. Family structure and household income
only exert significant effects on White youths. Age has
substantial effects on each group. Being old decreases the

attachment to parents.

178

Models 5.1, 5.2 and 5.3 constrained the effects of
background variables on attachment to schools in
comparisons of Blacks and Whites, Asians and Whites, and
Hispanic and Whites, respectively. Again, there are
significant differences between Blacks and Hispanic and
Whites correspondently, but no significant differences
between Asians and Whites (see Table 9.4 for detail).
Parameter estimation (Tables 9.1a and 9.1b) indicates an
interactive effect of race and household income on youths'
attachment to school. Living in a high-income household
decreases Black and Hispanic youths’ attachment to school,
but it does not affect White and Asian youths’ attachment
to school. Family structure matters only for White youths,
and it does not affect non-white minority youths in terms
of attachment to school.

Models 6.1, 6.2, and 6.3 constrained the effects of
background variables on religiosity, and models 7.1, 7.2
and 7.3 constrained the effects of background variables on
involvement in comparisons of Blacks and Whites, Asians and
Whites, and Hispanic and Whites. Significant differences
are found between Blacks and Hispanic and Whites, but not
between Asians and Whites (Table 9.4). Sex and household
income have substantial and significant connection to

youths’ involvement in social activities for Whites,

179

Asians, and Hispanics, but not for Blacks. Non-black young
females are more involved in social activities than others,
especial older black male youths living in low-income
households. Sex and family structure are significantly
related for White, Black, and Hispanic youths, but not
Asian youths. Age is the only factor that is associated
with Asian youths’ religious beliefs. Family structure
again only exerts sizable and significant effects on
involvement to social activities and religiosity for White
youths. Although some significant relationships are found
between family structure and religiosity for Blacks and
Hispanics, the effects are modest in size.

The overall findings support the hypothesis that the
effects of background variables on social bonding vary by
race and ethnicity. Specifically, the general
relationships between background variables and social
bonding are significantly different between Black and
Hispanic and White youths, but the relationships are
parallel between Asians and Whites. There exist
interactive effects of race and sex, and race and household
income on attachment to parents and attachment to school.
Sex plays a more important role in attachment to parents
among Black youths than others. Household income

Significantly affects attachment to school for Black and

ISO

Hispanic youths, but not for Whites and Asians. There also
exists an interactive effect of race and family structure
on social bonding. A broken family exerts a larger impact
on social bonding among White youths than non-white
minorities. This finding is inconsistent with Matsueda and
Heimer (1987)’s finding that broken homes have larger
impact on delinquency among blacks than nonblacks.

The third finding addresses whether effects of
background variables on differential associations vary by
race and ethnicity. Models 8.1, 8.2, and 8.3 constrained
the effects of background variables on delinquent peers,
and models 9.1, 9.2 and 9.3 constrained the effects of
background variables on learned definition in comparisons
of Blacks and Whites, Asians and Whites, and Hispanic and
Whites. The chi-square tests reveal that the effects of
background variables on delinquent peers do not vary
significantly between Blacks and White, and they do not
vary significantly between Hispanics and Whites. But the
effects differ significantly between Asians and Whites.
The parameter estimation in Tables 9.2a and 9.2b shows that
sex is substantially related to delinquent peers for
Whites, Blacks, and Hispanics. Although there is a
Statistically significant relation between sex and

GGlinquent peers among Asians, the effect size is moderate.

[81

Contradicting expectations, females generally report more
delinquent peers than do males. As discussed previously,
this might relate to the approach of opperationalization of
delinquent peers for youths in this study. Age exerts
substantially significant effects on delinquent peers for
each group. Old youths generally report more delinquent
peers than young youths. Household income and family
structure do not have significant effects or they have only
modest effects for each group.

The effects of background variables on learned
definitions vary significantly between Blacks and White,
and Asian and Whites respectively, but not between
Hispanics and Whites. None of the background variables are
significantly related to Asian youths’ definition of drug
use. For White and Hispanic youths, being male, old, and
living in broken family are related to less conventional
norms, and to a higher ratio of definitions favorable to
law violation to definitions unfavorable to law violation.
Sex exerts relatively larger effect on Blacks than on
others.

In general, the findings reveal that effects of
background variables on differential associations vary by
race and ethnicity. However, the effects do not show a

discrepancy between Hispanics and Whites. The

182

relationships between background variables and delinquent

peers and definitions vary significantly between Whites and
Asians. Whites differ significantly from Blacks only on the
relationships between background variables and definitions.

The fourth finding concerns variation by race and
ethnicity in the relationships of background variables to
self-control. Models 11.1, 11.2, and 11.3 constrained the
effects of background variables on self-control between
Blacks and Whites, Asians and Whites, and Hispanic and
Whites respectively. Chi—square tests indicate the effects
differ only between Whites and Hispanics. No significant
differences are found between Whites and Blacks, and Whites
and Asians. Generally males tend to have lower self-
control than female across racial and ethnic groups.
Household income significantly affects non-Hispanics on
self—control, but exerts no effect among Hispanics.

The fifth finding pertains to the relationships
between social bonds, differential associations and self-
control. Models 12.1, 12.2, and 12.3 constrained the
effects of social bond and self-control on delinquent peers
between Blacks and Whites, Asians and Whites, and Hispanic
and Whites correspondingly. The chi—square tests indicate
effects of social bonds and self—control on delinquent

peers do not vary across racial and ethnic groups. Models

13.1, 13.2, and 13.3 constrained the effects of social bond
and perceived risk of drug use on learned definitions
between Blacks and Whites, Asians and Whites, and Hispanic
and Whites respectively. The overall effects on learned
definitions vary significantly between Blacks and Whites,
and Hispanics and Whites but not between Asians and Whites.
While all the other variables exert similar effects on
definitions across racial and ethnic groups, involvement
presents some differences. For Whites and Asians,
involvement in social activities substantially increases
the strength of conventional beliefs. However, involvement
shows no significant effects on definitions of drug use
among Blacks and Hispanics.

Models 14.1, 14.2, and 14.3 constrained the effects of
social bonds on self-control between Blacks and Whites,
Asians and Whites, and Hispanic and Whites respectively.
The overall effects of social bonds on self—control vary
significantly by race and ethnicity. For Whites and
Hispanics, attachment to parents exerts the largest effect
on self-control, followed by attachment to school. For
Blacks and Asians, attachment to school exerts the largest
effect on self—control. Attachment to parents present the
second largest effect on self—control among Black youths.

Nevertheless, attachment to parents does not affect self-

l84

control among Asians. Religiosity and involvement in
social activities present substantially effects on self—
control among Asians.

The sixth finding addresses the question of whether
the effects of delinquent peers, learned definitions, and
self—control and opportunity on drug use and delinquency
vary by race and ethnicity. Models 15.1, 15.2, and 15.3
constrained the effects of delinquent peers, definitions,
and self—control and opportunity on drug use between Blacks
and Whites, Asians and Whites, and Hispanic and Whites
respectively. The chi-square tests demonstrate that the
overall effects on drug use vary significantly across
racial, and ethnic groups (Model 15.1, AXZ=419.498, df=4,
p<.OOO; Model 15.2, Ax2=l8.4, df=4, p<.OOl; Model 15.3,
Ax2=57.359, df=4, p<.000). This finding is the basis for
rejecting the null hypothesis that the effects of
differential associations and self-control and opportunity
on drug use are invariant by race and ethnicity. Models
16.1 to 18.3 test whether the individual variable's effect
on drug use equal across race and ethnicity. The chi-
square test results presented in Table 9.4 indicate that
only the variable of learned definitions is significantly

related to drug use differently across groups. The effects

I85

of delinquent peers and self-control on drug use do not
vary by race and ethnicity. The parameter estimation
(Tables 9.3a and 9.3b) shows that the definition favorable
and unfavorable to breaking the law exerts the largest
effect for White youths (standardized coefficients of
learned definition is —O.60), followed by Black youths (-
0.46), and then Hispanic youths (-0.42). The effect of
definitions favorable and unfavorable to breaking the law
on Asian youths is the smallest in size (—O.33). Although
the sizes of effects of delinquent peers and self-control
on drug use vary somewhat by race and ethnicity (for
example, the standardized coefficients of self—control on
drug use are 0.36 and 0.35 for Asians and Blacks
respectively, and the coefficients of self-control on drug
use are relatively smaller for Hispanics and Whites (0.26
and 0.19 correspondingly), the differences are not
significant differences.

The overall effects of delinquent peers, learned
definitions, and self-control and opportunity on
delinquency vary only between Black and White youths (Model
19.1, Axa=17.262, df=4, p<.002). No significant differences
were found between Asian and White youths (Model 19.2,

AXA=7.353, df=4, p<.ll8), and Hispanic and White youths

(Model 19.3, Ax?:4.955, df=4, p<.292). Models 20.1 to 22.3

186

test the individual variable’s effect on delinquency. The
results show that only the variable of delinquent peers is
related to delinquency differently for Whites and Asians
(Model 21.2, Ax2=4.346, df=1, p<.O37). The effects of
definitions and self—control on delinquency do not vary by
race and ethnicity. The parameter estimation (Table 9.3)
indicates that the variable of delinquent peers exerts
substantial and significant effects on delinquency for
Whites and Hispanics (the standardized coefficients are -
0.15 and -0.12), but not for Blacks and Asians.
Opportunity exerts significant effects on delinquency for
Black and White youths. The effect size of opportunity on
delinquency is modest for Hispanics, but is still
significant. Opportunity is not significant related to
delinquency for Asians. For Asians, self-control is the
most important factor predicting involvement of delinquent

behaviors.

I87

.335:er E Somme $2.5 c.5955 ._oo.va 1:. Java .1. ”.0on _. “BoZ

 

 

2e. as. as. is as. see. N8, .2. as. 38. as. ”8. :3. at. 2 _. N2. 283-”:
3:: :2 V 9...: 3:: ::: 3m: ::: 3:: E: 3:: 3:: 3:: 3:: a: :2: 3:: 8:82am
5:. o8. as. .23 8.- c5 s5. :1 _. :5. So. So. 38.- So - 2:... 2e. 3:: .- 3:28
:3: 3:: ::: 3:: ::: 3.: ::: 32: ::: 3:: 3:: so: 3:: 3:: ::: :::: 068:. c
:0 es - mo. 8.- .52. :2 .2. 3:. ice. to. 38. S. ”8.- m8. :5- 53.- 6:83:
3:: E: 3:: 32: ::: $8: ::: 3.2: 32: 3:: :2: 3:: 3:: 3:: ::: 3.2: i
3.35. oi»: - _o- 53.33.- 9.5.- ”(or 3:22- aoemor oeemo. one. oecmo. oeomo :::. 9.355 «new... :51. vu<
:2: 32: 3:: 3d: 3:: ::: 3:: 3:: 3:: £2: 3:: 3:: :5: 3:: 3:: 2::
3:0“. NM: new. .. aonzr .39. one”- car :3th «co. m-O. Qo. eeeco. hmor No. cosh—t savor xwm
a2: :22 use: 32.5 as: 5.3 fig 223 as: 8.2 seem 353 ms: 5:2 seem 223
:20 em. :25 e5 :26 e5 :26 .3
m m m m em .3
too 3:93:83: (to... cudEaucsmcz v.25 3:28:52: goo): 3.36.3523:
b.8651 E25302: .02—um o. .5858? 9:28 2 32:535..

 

.wmfnmxoflm .oSoumBEBV mfizoxlmEoBEoou 3:33:83: £032.? econ 368 new 338:; eczema—23 5252. 32228.3. ”3.9 2%...

.38 _ u 8:53: .RTEE

188

33: t... ::v: .1. ”8.x. - H202

 

189

ouauozbm
:5.- S. .8. :L _. 8.- 3. 3.- ES. :5.- 5 25. 3.8.- 5.- S- 5. 2:2.- 3:5:
oEouE :
S. 2 .- co. 2: . 2:2. :5. .8. is. 52. £5. :::. S. 3.- 2.- :5.- ..LS- 6:330:
.300.- ....ON.. NO- .3.:.. NOK $0.. team-4.- »§.oo.. ....:..:. .2. 2:..N_. 3.1-5 Esz. :89“. Elk-M. film-m. 0w<
52 .- S. :8- ...S- .3. : _ a. 8.- .tom- .8. 8. 5. ES. 3.- S. 5:.- ..8.- 5m
:2: 53.: fig: 223 :2: 523 fig: 233 ::: 522 48.: 223 :2: :22 is: 22.3
to: 3m :25 .3 :80 .3 too :5
.Qmoazux 23539,:— _oo:om c. 52:53?- 358: o. Eng—222

 

3me 85%:
.mnmucflmxx .wmfnmxuflm .ioaumBEBV mfisoxlmEoBEooQ :oEEmnfizm .mo_:mtm> ::o: :2qu ::: mo_:m:a> “Enemy—um: 5036: 32:28.9: ::6 2:3-

.momoficema E 525a Bot» c3255

:52. .5 ”avg 3. ”Eva .1202

 

 

3.: 3m 3: 3... SN. 5d mom. .2. E» :8. ”3. m3. 2.253.
$2: ::: 2:: 3::
3.35.? 5:59.. :33. tine»: _O.SCOUo.:Um
6:: E: V :2: 6:: 3: man
.32. :::. .zm, :::. :0 V3: 38.9:
:2: ::: ::: ::: 3:: 3.2: ::: :3 8:: :2: 5:: :::
toimm. 31x5 .312. e¢.0_ «5. .00, _C. 3....- _. OCA- mogu «.0. NO.- :_UEU>_O>C_
3:: 3:: ::: Ex: ::: 2:: 3:: $2: 3:: ::: 3:: Ex:
:1: - 32.- ::5 - :33 - :15, .8. :3: :::. S. S. S. 5.- 5529.3.
E: ::: 2:: E: 3:: $2: 3:: 2 2: :2: 3:: E: 3:: .028
.neom. :43! .3341 3.21%. 3.1.5”- .2...me .3::.. .32”. 9.2, o_. «.3. time: OH HCDE£063<
3:: 5:: ::: 3:: E: :2: 2:: :3 3:: 3:: 5:: 3:: 3:28
f::.. we :35; 3:30.”. 3:: u 590.- :309- .3955- met co. No. _C, Ow uCUESUNfi/NV
::: Ex: 2.: 3:: ::: A. ::: 3:: : :: :::: 39: ::: 5::
.8 - 2 5.- .3- .S. 3,- 3.- 58 :5. .m..- 3.- S - 222:3, ::::E
3:: :::: 8:: Ex: 5:: ::: ::: $2: ::: 2:: 9:: 9.2: 2:85
8. .3 - :2 - 1.3.- 5.- 8: 5. _. _ c- 3.- «o. 5. 1: 2028.5:
3:: :9: E: #2: :2: 3:: :::: 3.2: 3:5: ::: :5: 3.2: n
#00: _o.- :7 Lo. s’cmov NO... 3.1.9.3.- :Eoor 3.2.2. .qeom. «.32. .iLL. va<
::: 3:: 3:: 2:: 3:: £2: ::: $2: ::: 2:: E: E:
:::. : poomN. aaqom. .mo.- mo... .:..oo{ oNO... 9:.NN.- .05.. :3_N.- ashmm - xom
5?: 53¢. xoflm 2E3 :5: :a_m< 335 BE? .35 =£m< 385 223 9 moEata> d:—

_O.=COU..._ _0w

3: wife .5253: 353‘.

8qu Eon—.530

T.E> w=.:u_co:__

:32” 8:23: ,Rmncgé
.wmfnmxusm .oSoumBEBV misoxlmEEUEwou 3:33:33: £03.22? choEBE wan 332:; EuucoaonE c8332 85228.3: ”mmd 2an

190

.88.!H 1... 8v; 1. ”Eva _. éoz

 

 

“:::“. :::-m. 2.3—m. .aumm. _O.=:OUH=DW

8: win

:3; :3. 52. :::. go it 58.8

:::-N. «gm-N iimm. 32.2. ms. .05 No. 3:2. cc.- OCOV we. Nev EOE-wing):—
:.m__- 2.2.- 5.: - :1: - :::. .2. :E. 38 8. 8. 8. 8.- 352-33.
_oo:om

:30»! iii-m. .gumv. :::.m .33.”- 3;ri 33.x”.- .3..«&.. :03. no. no. 9:.wo. Ow «C0E50m3<
3:83

.3:- 2 :La 3.3 58.- .2.- .t: ,- $2.. 8- 8. 8 :8. 8 E25233
.8- 2. 8.- .8 - 8 8.- 8.- :8. 88. .8.- 8 - 8.- 252:: DEE
2:85

8, .2 .- :2 - 38.- 8,- 88. 8 .8.- 8 - 8. 8.- .8. 228.6:

_c. «.0.- MOL 9Mc - iiomr o0 - 9.....c_.- {oi-fl..- 3.3m? 3.3—0. 31.9%. .993“. ow<
33.2. N_. §¢§wm :LN. ’00.- no - :33..- amOI 2.5—N- 25,- .3505- 2.3—N.- xom
.3: :23 V32: 223 .82: 5.3 fig 223 E: 53- 85 223 a 322.; .9:

_o::ou.-:om

3: man go SEE-Bu 358.—

33g E25530

T3> choEuE.

.382“ 8.52:

.mnmncmiax .wmfnmxum—m .oSoumBEBv meaoxlflcuBEooQ 38235; 8032:; choEBE ucs 335:; Eovcoaonfi 50382 mafmcosfiom bmd “Em--

191

dome—22:23 E :8qu Echo Eavcflm ._oo.va 1:. Java .1. ”mon .. ”202

 

 

88. E. m2. 3.. 8m. 22. 2:. 3m. “5%.”.
32: 283 $83 283 283 283 A83 A83
.. _ o. oo. timo- ***mo. .1. _o. VOOK 5c. 33.. .0. bEztoaqo
a;
3:: $8.8 38.8 82: 283 8:: :2: 383 w
EINN. 31.2. tic” ..:.:.._N. _.:.:..mo. time. 3336. figme- _o-=:oo¢_um
Am _ o; Ammo; Clog 3309 32: 3:: 32: Soc; 3: mas
..:.:..o_.- ffwofi {troor 3....wor SIB—r 31.99- flavor 3.32.- m0 9525.30
2:: ES 2:: $2: 83 E3 :83 283 282
335.- moor Nor timc... _.:..mo. No. 3.. 5. time. 25:6:on
.22: 53 235 223 .22: :23 225 223 a 822.9 .2:
A: 338:3
5535.00 8: win choZuE—

 

.2.me 8:285: .mnmucflmg‘ ,wmfnmxugm d—ooumBEBV
m—tzoxlmEuBEooQ 23682532: .xocozcEBc 25 um: wEu van 8335.» 252532: c8335 man—£22223. ”and «Em-r

._oo.vQ :3 :ova I Hova .. ”3oz

 

*8. :2.- 1.2. s: _ _. :8. 8.- 8. :8. 322580
3.306. tiow. 22.23%. _.:.:.._©. fibN. flew. gingmm. _.:.:._Q_. _o.=:Oo..:um
8: mac W
3.1.?”- _.:._th ***NN.- 3.1. _ Nf 22.....er .3...mm.- ..:I.cv.- slice..- mo mcoECGoQ
38¢
..:..m~ - mo.. 00.- _.:.:..m_.- 3.2:. N_. 3..m_. timo. 23:9.sz
.22: :3 222m 223 22: :23 2.225 223 a $22§ .2:
A: 332:;
3:223:00 mm: min choZoE—

 

gown—u 3:592: .mnmncflfix £379.85 .oSonBEBV
fizzoxstmEoUEooQ 286528; mucozcczou can 3: win 98 33st? 2.3.539: 28389 3220:23— bmd ~33-

Table 9.4: A variety of models to test the significance of groups of parameters—youths

 

1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Model Constraints X df AX: (if p- Test implied in Ax:
value
I Allgroups free 51610829 7800 -- --
2 All groups 52721.621 7950 1 110.792 150 .000 All groups are equal
equal
31 B=w 52463.836 7850 853.007 50 .000 Coefficients are equal
32 Azw 5 1736.741 7850 125.912 50 .000 between racial and ethnic
33 1-1=w 51852735 7850 241.906 50 .000 groups
41 On BS b=w 51652.404 7804 41.575 4 .000 Effects of background
42 On BS a=w 51614677 7804 3. ~ 4 .427 variables on attachment to
43 On BS h=w 51631.184 7804 20.355 4 .000 parents are equal between
racial and ethnic groups
51 On B6 b=w 51654.945 7804 44.1 16 4 .000 Effects of background
52 On B6 a=w 51615.962 7804 5.133 4 .274 variables on attachment to
53 On B6 h=w 51634.254 7804 23.425 4 .000 schools are equal between
racial and ethnic groups
61 On B7 b=w 51645.101 7804 34.272 4 .000 Effects of background
62 On B7 a=w 51619.515 7804 8.686 4 .069 variables on religiosity
63 On B7 h=w 51624758 7804 13.929 4 .008 are equal between racial
and ethnic groups
71 On BB b=w 51660.683 7804 49.854 4 .000 Effects of background
72 On BS a=w 5161 1.023 7804 0.194 4 .996 variables on involvement
73 On BS h=w 51643.220 7804 32.391 4 .000 are equal between racial
and ethnic groups
81 Bg On Bl b=w 51614.164 7804 3.335 4 .503 Effects of background
82 Bg On Bl a=w 51625.091 7804 14.262 4 .007 variables on delinquent
83 Bg On Bl h=w 51617.661 7804 6.832 4 .145 peers are equal between
racial and ethniggroups
91 Bg On da b=w 51627.804 7804 16.975 4 .002 Effects ofbackground
92 Bg On da a=w 51620.600 7804 9.771 4 .044 variables on delinquent
93 Bg On da hzw 51613.335 7804 2.506 4 .644 peers are equal between
racial and ethnic groups
1 l 1 Bg On B9 b=w 51613.179 7804 2.35 4 .672 Effects of background
1 12 Bg On B9 a=w 51616.448 7804 5.619 4 .229 variables on self-control
I 13 Bg On B9 h=w 51624876 7804 14.047 4 .007 are equal between racial
and ethnic groups
121 On Bl b=w 51620.518 7805 9.689 5 .076 Social bond and self-
122 On Bl a=w 51614.769 7805 3.94 5 .558 control variables impact
123 On Bl h=w 51618.976 7805 8.147 5 .148 delinquent peers equally
across groups
13 On da b=w 51635.186 7805 24.357 5 .000 Social bond variables and
1‘- On da ar—w 51618.938 7805 8.109 5 .150 perceived risk impact
13 On da h=w 51620.797 7805 9.968 5 .041 definitions equally across
racial and ethnic groups
141 On B9 b=w 51631.319 7804 20.49 4 .000 Social bond variables
142 On B9 a=w 51624.494 7804 13.665 4 .008 impact self-control
143 On B9 h=w 51620.710 7804 9.881 4 .042 equally across racial and
ethnic groups
151 On f1 bew L 52030327 7804 419.498 4 .000 Delinquent peers.
152 On f1 a=w 1 51629229 7804 18.4 4 .001 definitions and self-

 

194

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

On f1 h:w 51668.188 7804 57.359 .000 control impact drug use
equally across groups

Da ON f1 b=w 51915.274 7801 304.445 .000 Learned definitions affect

Da ON f1 a=w 51627.207 7801 16.378 .000 drug use equally across

Da ON f1 h=w 51658.671 7801 47.842 .000 racial and ethnic groups

Bl ON f1 b—=w 51612.359 7801 1.53 .216 Delinquent peers affect

Bl ON f1 a=w 51610.924 7801 .095 .758 drug use equally across

Bl ON f1 h=w 51611.453 7801 .624 .430 racial and ethnic groups

B9 ON f1 b=w 51613.269 7801 2.44 .1 18 Self-control affects drug

B9 ON f1 a=w 51612.788 7801 1.959 .162 use equally across racial

B9 ON f1 h=w 51610.914 7801 .085 .771 and ethnic groups

On 12 b=w 51628.091 7804 17.262 .002 Delinquent peers,

On 12 a=w 51618.182 7804 7.353 .118 definitions and self-

On 12 h=w 51615.784 7804 4.955 .292 control affect delinquency
equally across groups

Da ON f2 b=w 5161 1.391 7801 .562 .468 Learned definitions affect

Da ON f2 a=w 51610.867 7801 .038 .845 delinquency equally

Da ON 91 h=w 51612.565 7801 across racial and ethnic
groups

Bl ON f2 b=w 51614.348 7801 3.519 .061 Delinquent peers affect

Bl ON f2 a=w 51615.175 7801 4.346 .037 delinquency equally

Bl ON f2 hew 51610.949 7801 .12 .729 across racial and ethnic
groups

B9 ON 1?. b=w 51612.770 7801 1.941 .164 Self-control affects

B9 ON a a=w 51611.938 7801 1.109 .292 delinquency equally

B9 ON 12 h=w 51611.017 7801 .188 .665 across racial and ethnic

 

 

 

 

 

 

groups

 

195

 

Multiple Groups Analysis for Adults

 

Parameter estimates of the multiple groups model for
adults appear in Tables 10.1a lO.lb, 10.2a, 10.2b, 10.3a
and 10.3b. Compatible with the model for youths, this
model examines the relationships between independent
variables, intervening variables, and dependent variables,
and the parameters are set to be free across racial and
ethnic groups. To test whether the effects of specific
variables invariant by race and ethnicity, a variety of
constrained models were examined. The chi-square test
results of different models are presented in Table 10.4.
These analyses reveal six principle findings.

First, using the overall test of invariance (Table
10.4, model 2), the hypothesis that all substantive

parameters are the same for different racial and ethnic
groups is rejected (Axe=i449.534, df=126, p<.000). Specific
cross—group comparisons show that the processes of drug use

and delinquency vary significantly between Blacks and
Whites (Axe=969.486, df=42, p<.000), and Asians and Whites
(Ax3=184.609, df=42, p<.000), as well as Hispanics and

Whites”.

 

'5 The model that constrained the equivalence of parameters between Hispanics and Whites did not reach
convergence. Even though the problem of non-convergence was not identified in this study. it indicates
that the constrained model does not fit as well as the un-constrained model.

196

Second, the tests of hypotheses that the determinants
of the processes of social bonding vary by race and
ethnicity reveal that the total effect of background
variables on attachment to friends varies significantly
between Blacks and Whites (Model 4.1, Ax1=2l.37, df=3,
p<.000). It also varies significantly between Hispanics
and Whites (Ax;=28.602, df=3, p<.000). However, no
significant difference between Asians and White is found
(Ax1=2.409, df=3, p<.492). The results also show that the
total effects of background variables on another element of
social bonding—religiosity—for Blacks, Asians, and Hispanic
are different from Whites respectively (Table 9.4, model
5.1, Ax?=7.7l, df=3, p<.052; Model 5.2, 813:9.403, df=3,
p<.024; Model 5.3, Ax1=l4.622, df=3, p<.002).

The third finding indicates whether the effects of
background variables on elements of differential
associations vary by race and ethnicity. Models 6.1, 6.2,
and 6.3 constrained the parameters between Blacks, Asians
and Hispanic and Whites correspondently. The chi—square
tests reveal that the effects of background variables on
criminal peers vary significantly between Blacks and Whites
{Ax9=27.6a, df=3, p<.000), Hispanics and Whites (813:12.445,

df=3, p<.OO6), as well as Asians and Whites (Ax?=3o.936,

197

df=3, p<.000). Sex, age and household income are unrelated
to the number of criminal peers for Black, Asian and
Hispanic adults. The effects of background variable on
criminal peers for Whites are modest in size but
statistically significant. The tests of relationships
between background variables and learned definitions across
race and ethnic groups (Table 10.4, Models 7.1, 7.2 and
7.3) show that the effects of background variables on
learned definition are invariant by race and ethnicity.
Findings about the relationships between background
variables and elements of differential association for
adults are inconsistent with findings for youths.

The fourth finding is that the effects of background
variables on self-control vary by race and ethnicity. The
chi-square tests of equivalence of parameters between
racial and ethnic groups disclose that there are
significant differences between Black and White adults,
between Asian and White adults, and between Hispanic and
White adults in terms of relationships between background
variables and self-control. For White and Asian adults,
age exerts larger effects on self-control than sex.
However, for Blacks and Hispanic, Sex exerts larger effects

than age.

198

The fifth finding concerns the relationships between
social bonds, differential association and self—control.
Models 9.1, 9.2, and 9.3 constrained the effects of social
bonds and self—control on criminal peers between Blacks,
Asians, Hispanics and Whites. The results suggest that
criminal peers have different influences on illegal
behavior depending on the racial and ethnic group. The
parameter estimation (Tables 10.2a and 10.2b) reveals that
self-control exerts the largest size of effect on criminal
peers for each group. Education level, marital status, and
employment status, attachment to friends and religiosity
all present significant effects for White adults. Low
education, staying unmarried or divorced, and being
unemployed increase the number of criminal friends for
Whites. Religiosity is substantial related to having
criminal peers for Whites. Education level, marital
status, attachment to friends and religiosity are similarly
connected to criminal friends for Blacks, however,
employment status is not significantly related to criminal
friends for Blacks. For Asian adults, marital status,
employment status, attachment to friends and religious
beliefs do not affect the number of criminal peers.
Nevertheless education level exerts much stronger effect on

criminal peers for Asians than other groups.. For

199

Hispanics, adulthood events such as education level,
marital status, and employment status do not affect the
number of criminal peers significantly. Attachment to
friends and religiosity are the two elements of social
bonding that exert effect on the number of criminal peers
for Hispanic adults.

Models 10.1, 10.2, and 10.3 constrained the effects
of social bonds and perceived risk of drug use on learned
definitions of drug use between Blacks, Asians, Hispanics
and Whites. The overall effects on learned definitions
vary significantly between Blacks and Whites, and Asian and
Whites, but not between Hispanics and Whites. For both
White and Hispanic adults, religiosity has the strongest
connection to learned definitions, followed by perceived
risk of drug use. Low education level, being unmarried or
divorced, and close relationships with friends are
associated with reduced strength of conventional beliefs
for Whites and Hispanics. Religiosity also is strongly
related to definitions for Black adults. Strong religious
beliefs, high level of perceived risk, and high level of
education, and being married are associated with increased
strength of conventional beliefs for Blacks. For Asians,
perceived risk of drug use exerts a stronger effect on

definitions than religiosity. Being married has a

200

substantial effect on Asian adults’ definitions of drug
use.

Models 11.1, 11.2, and 11.3 (Table 10.4) constrained
the effects of social bonds on self—control between Blacks
and Whites, Asians and Whites, and Hispanic and Whites
respectively. The overall effects of social bonds on self—

control vary significantly between Blacks and Whites
(Ax%=16.087, df=5, p<.OO7), and Asians and White
(Axe=lo.ls9, df=5, p<.OO6), but not between Hispanics and

Whites (Ax?=lo.les, df=5, p<.O7l). rot Whites, Blacks, and
Hispanics, religiosity again has the largest effect size.
But for Asians, religiosity exerts no significant effect on
self-control. Instead, marital status exerts the largest
size of effect for Asians. Staying single or divorced, and
having close relationships are most strongly connected to
low self-control for Asians.

The final finding about the model for adults reveals
the relationships between differential association, self-
control, and perceived risk and drug use and criminal
behaviors for different racial and ethnic groups. Models
12.1, 12.2, and 12.3 constrained the overall effects on
drug use for adults between Blacks, Asians, Hispanics, and

Whites respectively. The overall effects of differential

201

association and self-control plus perceived risk vary
between Whites and other minorities. Tests of effects of
individual variables on drug use by race and ethnicity
indicate that the variable of criminal peers is related to
drug use differently for Whites and other minority groups.
The variable, criminal peers, has a substantially larger
effect on drug use for Whites (standardized coefficient is
0.48) and Asians (0.41) than for Blacks (0.21) and
Hispanics (0.23). The effect of definitions on drug use
only varies between Hispanics and Whites. But the
parameter estimates show that the effect of definitions on
drug use is slightly larger for Blacks, Asians, and
Hispanics than for Whites. Self-control presents
significantly larger effects on drug use for Blacks
(standardized coefficient is 0.40) and Hispanic (0.42) than
for Whites (0.16) and Asians (0.18).

Models 16.1, 16.2, and 16.3 constrained the overall
effects on drug use for adults between Blacks, Asians,
Hispanics, and Whites respectively. The overall effects of
differential association and self—control and perceived
risk on criminal behavior vary significantly between Whites
and other minorities. White group has the largest size of
explained variance of criminal behavior (0.184), followed

by Black (0.165) and Hispanic (0.164) groups. The size of

explained variance of criminal behavior is fairly small for
Asian group (0.97). The relatively small size of explained
variance of criminal behavior for adults suggest that some
other crucial variables may need to be involved in order to
understand the process of adult criminal behavior. Tests
of effects of individual variables on criminal behavior
reveal criminal peers affect criminal behavior differently
between Whites and Blacks and Hispanics. The variable of
criminal peers has significantly larger effects on criminal
behavior for Whites (0.19) and Asians (0.22) than for
Blacks (0.08) and Hispanics (0.07). The effect of
definitions on criminal behavior varies between only Whites
and Blacks. The overall effect of definitions on criminal
behavior is trivial. Self-control exerts significantly
larger effects on criminal behavior for Blacks (0.33) and
Hispanics (0.30) than Whites (0.23). Self—control has no
significant effect on criminal behavior for Asians. The
parameter estimates also show that perceived risk of drug
use has significant effects on criminal behavior for

Whites, Asians, and Hispanics but not for Blacks.

Summary

As an overview of the results of the structural
equation analysis of the integrated model, it appears that
differential associations and self-control are the main
factors directly influencing both drug use and
delinquency/criminality, and these predictors mediate the
effects of social bonds as well as background variables on
the illegal behavior. The mean structure analysis and
multiple groups analysis provide results showing that the
theoretical constructs of differential association, social
bond, and self-control theories vary significantly across
racial and ethnic groups. The causal processes leading to
drug use and delinquent/criminal behavior are not invariant

between racial and ethnic groups.

KM

.33558Q E 53% 20:0 22:55

._OO.VQ n: U._o.VQ a; Mmo.vQ ... u302

 

 

.8. 88. 8... m8. .8. _ 8. 48. a8. 2283.
$8.. .88.. $8.. 38.8 e _ 3 38.. $8.. .88.. 2:85 m.
3.50.- no. annexe. _o. 11.0% mo. ***o_. nil-o. 3053.30:
.88.. 3:: $8.. 8.... 38.. : 8.. .88.. 88..
3.1.x? 4440—. 3.1.00. 3.1.00. ester-Q- tumor clamor naive... ow<
28.. A83 $.88 5:: $8.. $8.. $8.. 2 8..
stem.- ..8: on- 32.2.. 88. 8.- .8. 8.- 5m
as: 5.2 82m 253 on... can... v.85 223
mm

“Sou 38335323

8.8.83.

uuou 38235233
3:2... 0. E08583.

 

.38.? muEmamE ._wxu:m_m< $3580.35 .Nchmumozf‘C

93311956508 383.8282: 6032:; econ _m_uom ecu 832:; 959.303 503.3 3580:23— 82: 2an

 

 

.Sovg z; :38 t. 8.x. .. Heoz m
058.:

«d.- ho. :::-o. _o. 315—. mo. ..:.:..N_. ..:.:..O_. 29.330:
3.1.0.. ..:.:.._N. *......©_. fill. tic—r ..:..__.- ..:.:..o_.- :30:- ow<
..:.:..N_.- _.:..N_.- N_.. film—r :5. _o... .35. #0.- xum

.2: 5.3.. .85 223 .2: 5.3 .85 223

goo-u 35235.5 .000 33:35.5
b62331 3:05 9 26.5.8ch

 

.38.? 3.5853 Jaw-1.523.. .wom.mumv_om_m .mco.mmumo.E3v

3331353508 382353 mufimtg ucon 360m 95 33.3.3 25230.3 5923 3222.23. “n _ .o. 033-

33558.. c. .8an 20:0 Emocfim ..oo.v.. 1.... Joan. 1. ”mo/Va ... .202

 

 

8... 28. m...- ..8. 8m, 8m .8. 8... 88. am... 8.. V8. 8.8.8-..
$8.. :8. 88.. .88..

.52. :8... 3.... 2:8. .o....8.-._um

$8. $8. $8.. 38.. 8..

5.... :3... 3...... :3... 8...... x... .32....
:8. £8. .2... .8... .8. 88. .2... .88. .88.. .28.. .8... .8...

3:»...- M0.- .12....”— - 31.0..- «3.0.. :«00 1.3—N. .32.“... .gm0r N0.- 3.300... 3.1.x...- bingvd
:8. :8. .2... .8... .2... :8. :8. .8... 88.. 88.. :8. .88.. 85...
..:.:..0. 95.3. .5ng 3......0. 13.10.- 900 - N0.- i¢¢N0 - 3:90. N0. {SD-0 :32... 00 020550656.
:8. $8.. .28.. 8...... .2... 38.. .28.. .88.. .8... :8. 88.. .88..

.8. 8.. 8.... ::8. 88.- 8.... 8... .8.. 8...- 8. 8... 1:8. 88: .coEonEm
.2... $8.. 28.. .8... : 8.. 28.. .2... .88. .88.. 88.. .8. .88..
3:00. :30— .......co. .....:..0_. «3......0.- 3.2..- 310-0.- :...»..0.- 000. mmo. .390. .530. mazmum 23:02
.88.. .28.. .8... .8... .8... .28.. 88.. .88.. .8... .88.. .88.. :8.
9......m0 vmo. .0. 5.9.0. n§§N0. moor «emo. 9.5.0. 000. .aemOi 320.- :2.._0.- _D>m._ COCNUZUW
.2... :8. :8. .88.. :8. E... : 8.. .8... 8.... .2... .8... .8...
2:8, 8. 8.- 8.- 8.. m8 8.. - .8. 8,- 2...- .8- ...8 - 2:8... 22.38:
.8... .8. .8... .88.. .88. .8... :8... :8. .88.. .8... .8... .8... n

.330.- 4.2.00.- e'nfi0t .3100.- »:.N0.- :0 - pay—0.- .:.N0.- 000 N0. 000. 9.9.0.- UO<
88.. £8. .8. .8... 2.8.. 28.. 8.... 88. .1... .88.. 2.8.. .88..
3.8m” .suvm. «:ON. ..:mm. .33.. .- .23.: - 8:00.- .3590 - :0. #0.. N0.- 5:.m0t Xom

.82: 5.2 .2... 2:...» .3.-. 5.3. 82m 2...... mm... 5.5.. .82. 2...... 9 SEE; 8...

_o....oo.-:uw

um: w:.....o 2o......-.on.

83.... 855.5

A: 332:5 9.52:3...

.88.? 8.5%.: .8822

.wom.mumv_um_m .moo.~mum8£3. 33311856508 3223:8323 £282.? 9.83.82. was 833.9 80.3.2.3... 233.2. 85228.3. .mN.o.oBaH

207

.88 ... 8.x. .. 8.x. . Heoz

 

 

...8. ...8. ...8. ...8. 8.882%
om:

.....m. ...8. ...2. .....m. wan... ...»: 888..

...R - 8 - ...8- ...8- ...8. ...8. ...8. ...Nm. ..8- 8.... ....v-..- ...8- 88.9.3.
8:2...

:::. ..2 ...8. :::. ...8- .8 - 8.- ...-8.- ...8. 8. ...8. ...8. o. 825.223
.8. m8 8.- ...8. 8...- .8. 8... .8- .8.- ..8. .8. ...8. 8:8 .888...Em
...2. ...: ... _ .. ...-t. ...-8.- .....- ...8.- ...o..- .8. 8... ..8. ...8. 888.352
:8. m8. 8. ...8. ...8 :8.- ..8. ...8. .8.- ...z.- ..8- ...8.- 85. 5.80:3
...8 88, 8.- 8.- 8. c8. 8.- .8. 8.- 28.- .8.- ...8.- 2:8... 22.88:
....m - ...mm- ...-8.- ...8 - ...8.- 8.- ...-8.- ...2- 8...- :. 8. ...8- m8%
...mm ...8. ...8. .....m. .... ..- ...2- ...-8,. .... ..- :8. 8.- 8,- 38.- xom
ma: 5.8.x 42.5 2.55 ma: :w..m< xan 2.55 .92: =Em< 82m 253 A, 8388> .2:

_o==3.-:um

8: win .o 3225.32

802. 855:0

AI 338...; mEEEBE

 

.88.? 8.5%... .3888...
.wom.mumv_om_m .~o©.mmnmo._:30 mzsvmllmEoBEooU 322%:me 8.38:? 85532:. was 332...; .concogoufi c8253 maEmcozsom ”£5333-

208

$305223 E 83% 32.5 23:55 ..oova .1... :QVQ .1 HQVQ ... ”302

 

 

 

we _. So. 3.. 3:. 3. New Ev. m2. ”saga

A33 :2: s 2: $2: 2 z: 3:: 82: 62:
mm: wage

"lino—... *mo.- :0... iiwov 12,09- firmer ***mo.- ***m_... me b02095;
A83 62: $2: :2: A83 A83 A83 :2:

8o. 25.- 3o. .1; 5. 3a.- 25.- 8o. :5. beszoaao
: :3 A23 53 $2: 82: E: 62: A83
.....iao. mmo. “.212. 3.:er “......Auo. vo. 3.3.09 ......vo. .obcounzom
62: : 2: Ex: 32: A83 2 z: :53 :2:

3: wet
moov Aooo. ...NOV ...... 5. ......m9.. ......MOK ***No.- 3.... 5.. mo mcozEcoQ
G :3 889 ::: 62: 2 :3 :2: A83 82:

mo. ...ho. ......vo. 3.3.50. .....iho. :2..N_. ......*vo. ..:.:..w_. 9.0an EEC—CU
.2: 523 v.85 223 .2: =23 fism 323 a 832:2, .2:
3.8 3.8
goou .300 Al 338:;
>283:on 3: win choEoE—

 

.88.? 9:525: .~wwuca_m< .wom.mu$_on_m .moo.mmum8£3v

mzznmlmEuBEuOQ cofiEaucflmvca .xocozcczov USN 3: man can 833.? Eon—humans 50389 REESE—um Mama: anH

209

.33“. I153 .. ”3.x. ... H902

 

 

8: 330
3.1.2.- .22.- bmor tin—... ...»...N—r ......Z... 3.2.3.- .....ia—f xm: voZuuhom
05. VS- so. 33. 8.- 5.- 80.- 8o. 956250
....Ibm. 0o. .....2..mm. 3.3mm. 3.....Nv. w_. silo? ***c~. ESP—Coutum
3: man
_o.- Nmo. ..%o... .315. firm—... .I.N~.- 3.1.05- ......«vor «0 mcoEccoQ
no. ...NN. ......wo. 335—. 3.3mm. 3.3 _v. ......LN. .1ti 9.001 BEECU
.2: 52% fig 223 2: 53 62m 253 a 332? .2;
..EoU .Em uoou .Em ¢ 833:3
zucoscczoo om: wED choEBE

 

$3,? 3:535 ._mwn:m_m< .womfinmxom—m .chmmntEBV

SEEIIWEBUEQOQ nofinhmncfim £25353“. 23 3: win new 833:; acoucoaotE 503.3 33228.01 52: 2an

Table 10.4: A variety of models to test the significance of groups of parameters—adults

 

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Model Constraints X df Ax: df p- Test implied in Ax2
value

1 All groups free 72968.881 4040 -- -—

2 All groups = 744 l 8.415 4166 1449.5 34 126 .000 Parameters are
constrained to be equal
across groups

31 B=w 73938.362 4082 969.486 42 .000 Coefficients are equal
32 A=w 73153.490 4082 184.609 42 .000 between racial and ethnic
33 H=w * groups
41 Bg On a2 b=w 72990251 4043 21.37 3 .000 Effects of background
42 Bg On a2 a=w 72971.290 4043 2.409 3 .492 variables on attachment to
43 Bg On a2 h=w 72997.483 4043 28.602 3 .000 friends are equal between
racial and ethnic groups
51 Bg On a5 b=w 72976.591 4043 7.71 3 .052 Effects of background
52 Bg On a5 a=w 72978.284 4043 9.403 3 .024 variables on religiosity
53 Bg On a5h=w 72983503 4043 14.622 3 .002 are equal between racial
and ethnic groups
81 Bg On a1 b=w 72996.56] 4043 27.68 3 .000 Effects of background
82 Bg On a1 a=w 72981.326 4043 12.445 3 .006 variables on delinquent
83 Bg On a1 h=w 72999.817 4043 30.936 3 .000 peers are equal between
racial and ethnic groups
91 Bg On da b=w 72972.019 4043 3.138 3 .371 Effects of background
92 Bg On da a=w 72969427 4043 .546 3 .909 variables on delinquent
93 Bg On da h=w 72970966 4043 2.085 3 .555 peers are equal between
racial and ethnic groups
1 1 1 Bg On a6 b=w 73003.430 4043 34.549 3 .000 Effects of background
1 12 Bg On a6 a=w 72978.516 4043 9.635 3 .022 variables on self-control
l 13 Bg On a6 h=w 73026.823 4043 57.942 3 .000 are equal between racial
and ethnic groups
121 Sb On a1 b=w 72985.023 4046 16.142 6 .013 Social bond and self-
122 Sb On a1 a=w 73017.357 4046 48.476 6 .000 control variables impact
123 Sb On a1 h=w * delinquent peers equally
across groups
131 Sb On da b=w 73003.236 4046 34.549 6 .000 Social bond variables and
132 Sb On da a=w 73009.888 4046 41.007 6 .000 perceived risk impact
133 Sb On da h=w 72977.288 4046 8.407 6 .210 definitions equally across
racial and ethnic groups
141 Sb On a6 b=w 72984.968 4045 16.087 5 .007 Social bond variables
142 Sb On a6 a=w 72985.040 4045 16.159 5 .006 impact self-control
143 Sb On a6 h=w 72979.049 4045 10.168 5 .071 equally across racial and
ethnic groups
151 On f1 b=w 73593277 4045 624.396 5 .000 Delinquent peers,
152 On f1 a=w 73012049 4045 43.168 5 .000 definitions and self-
153 On f1 h=w 73128.751 4045 159.87 5 .000 control impact drug use
equally across groups
161 Da ON f1 b=w 72971.245 4041 2.364 1 .124 Learned definitions affect
162 Da ON f1 a=w 72970.908 4041 2.027 1 .155 drug use equally across
163 Da ON f1 h=w 72974.120 4041 5.239 1 .022 racial and ethnic groups
171 a1 ON f1 b=w 73249.969 4041 281.088 1 .000 Delinquent peers affect
172 a1 ON f1 a=w 72973.08] 4041 4.2 1 .040 drug use equally across

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

173 a1 ON f1 h=w 73058.797 4041 89.916 1 .000 racial and ethnic groups

181 A6 ON f1 b=w 72975.113 4041 6.232 1 .013 Self-control affects drug

182 A6 ON f1 a=w 72968.883 4041 .002 1 .964 use equally across racial

183 A6 ON f1 h=w 72997.083 4041 28.202 1 .000 and ethnic groups

191 On f2 bzw 73024.189 4045 55.308 5 .000 Delinquent peers,

192 On 12 a:w 72984.932 4045 16.051 5 .007 definitions and self-

193 On f2 h=w 72991 .558 4045 22.677 5 .000 control affect delinquency
equally across groups

201 Da ON f2 b=w 72978.835 4041 9.954 1 .002 Learned definitions affect

202 Da ON f2 a=w 72969.028 4041 .147 1 .701 delinquency equally

203 Da ON 12 h=w 72971.980 4041 3.099 1 .078 across racial and ethnic
groups

21 1 a1 ON 12 b=w 72972.871 4041 3.99 1 .046 Delinquent peers affect

212 a1 ON 12 a=w 72968.897 4041 .016 1 .899 delinquency equally

213 a1 ON f2 h=w 72975.102 4041 6.221 I .013 across racial and ethnic
groups

221 A6 ON f2 b=w 72996.285 4041 27.404 1 .000 Self-control affects

222 A6 ON 0 a=w 72970.305 4041 1.424 1 .233 delinquency equally

223 A6 ON 12 h=w 72976.585 4041 7.704 1 .006 across racial and ethnic

 

 

 

 

 

 

 

groups

 

*No convergence for these models.

lx)
is)

 

CHAPTER V: DISCUSSION AND CONCLUSION

The purpose of this study was threefold: to develop an
explanatory model that integrated differential association,
social bond, and self-control theories into a single
paradigm that accounts for drug use and illegal behavior,
to examine whether the causal processes leading to drug use
and delinquent behavior vary by race and ethnicity, and to
determine whether the causal processes are consistent for
youths and adults. This last objective led to the
identification of the relative importance of explanatory
factors for youths and adults separately. A series of
hypotheses were tested in order to determine the
relationships between race and ethnicity, differential
association, social bond, and self-control, and drug use
and delinquency. This final chapter provides an overview
of the research and discusses major findings for each of
the tested hypotheses. Limitations of the research and
suggestions for future work will be discussed in detail.
The chapter closes with a discussion of this study’s

implication for theory and policy.

Summary and Discussion

Predictive Efficacy of Integrated Model

The first objective of the research was to improve the
understanding of the causal processes leading up to drug
use and delinquent behaviors, and to provide empirical
evidence of the predictive efficacy of the differential

association, social bonds, and self-control.

Integrated MOdel for Youths

Research findings indicate that the modelexplains
substantial variation in drug use and delinquent behaViors
among youths. Delinquent peers, definitions of drug use,
and self-control, as expected, are strong predidtors of
drug use and illegal delinquency., The strongest predictor
of drug use is learned definitions favorable to drug use,
and the strongest predictor of delinquency is low self-
control. Opportunity displays modest but significant
effects on drug use and delinquency. ‘These findings
support both differential association theory and self-
control theory.

The effects of elements of social bond—attachment to

parents, attachment to school, involvement, and

religiosity—are mediated by learned definitions and self-

control. Social bonds operate indirectly through the
process of learning an excess of definitions favorable to
drug use, and through low self—control. This finding is
consistent with Metseuda’s (1982) causal modeling approach,
and it also provides evidence that supports Gottfredson and
Hirschi’s (1990) argument about the causal process in self-
control theory.

Additionally, a strong relationship is found between
self-control and delinquent peers. Low self—control is
highly related to having delinquent peers. This finding
provides support that low self-control not only directly
leads to law violation behavior, but also reinforces the
delinquent behavior by increase the involvement with
delinquent peers. Moreover, social bonds, differential
association, and self—control mediate the effects of
background variables. Sex, age, household income and
having a broken family are associated with drug use and
delinquent behavior through a process—by attenuating
parental supervision and attachment to school, reducing the
involvement in conventional activities, and weakening the
bond to religiosity, which in turn increase delinquent
companions, excess definitions favorable to law violation,
and low self-control, and ultimately, drug use and

delinquent behavior. The findings also suggest that

background variables directly impact the association with
delinquent peers, and definitions of law violation, and
self—control. However, the traits of self-control are

relatively stable during the adolescent period.

Integrated Model for Adults

The structural equation modeling for adults produced
findings analogous to those for youths. The model explains
substantial variations of drug use among adults. However,
the efficacy of the model to explain criminal behavior
among adult is moderate. Definitions of drug use, and
self-control as well as perceived risk of drug use were
strong predictors of drug use and crime. However, the
variable of criminal peers was the strongest predictor for
adults’ drug use, and self—control was still the strongest
predictor of criminal behavior. Definitions of drug use
were moderately related to drug use and have no significant
relationship to criminal behavior. The effects of
education, marriage, and employment status as well as
attachment to friends and religiosity were mediated by the
process of learning definitions favorable to drug use,
hanging out with criminal peers, and low self-control.
This result indicates that efficient socialization is not

only the consequence of child—rearing practice, but could

also be affected by adulthood transitions. Religiosity, an
alternative institution in addition to family and school,
exerted substantial effects on definitions, delinquent
peers, and self-control for adults. Consistent with the
model for youths, low self-control substantially increase
the number of criminal peers, and the effects of background
variables were mediated by social bonds, differential
association and self—control. Age, exerts significant
effects on self-control among adults. This finding is
contrary to the finding among youths. Taking together,
Gottfredson and Hirschi’s (1990) assertion that the traits
of self-control persist through life is not supported in

this study.

Summary

To sum up, the integrated structural equation model is
applicable to both youths and adults to explain the
patterns of drug use and delinquent behavior. However, the
model’s power in explaining criminal behavior among adults
is moderate to low, which indicates that either the process
of operationalization of the core theoretical constructs
needs to be improved, or some other explanatory factors
need to be considered. Although empirical support for the

general causal linkages is parallel for youths and adults,

m7

there are some crucial differences. Learned definitions
favorable to breaking the law is the strongest predictor
for a youth’s drug use, conversely, criminal peers plays
most significant role in predicting an adult's drug use.
Nevertheless, low self-control seems to be an important
predictor for both youths and adults' delinquent/criminal
behavior. The determinants of social bond vary between’
youths and adults as well. While parents’ supervision, and
attachment to school were strong predictors for a child’s
bond to conformity with the legal norms, education,
marriage, and employment status as well as religiosity were
decisive for an adult. These findings plus the evidence
that age has significant impact on adult’s self-control
suggest that the determinants of criminality change through
life. It seems that child—rearing practice, adulthood
transitions, and social demographic conditions all
influence the patterns of criminality through the effects
on differential associations, and self-control.

It was not possible to test for the degree to which
child—rearing practices explained later differences in
education level and marital status for adults. If child-
rearing practices are confounded with education and marital
status, then Gottfredson and Hirschi’s (1990) idea that

child—rearing is the basic reason for self-control may be

valid. But if child-rearing practices and education level,
and marital status are independent, then there is evidence

that challenges Gottfredson and Hirschi’s proposition.

Causal Processes by Race and Ethnicity

 

The Chapter II literature review revealed that
theorists assumed the causal processes for drug use and
other illegal activity are similar across racial and ethnic
groups, but little prior empirical research had tested the
hypothesis. The second objective of this study was to
provide empirical analyses that answered the question about
whether the causal processes leading to drug use and
delinquent/criminal behavior vary by race and ethnicity.

The findings are mixed.

Variations of Theoretical Variables

 

by Race and Ethnicity

 

Previous research shows that drug use is more common
among white and Hispanic youths than black youths (Ma and
Shive, 2000; NSDUH, 2003). But violent and property crime
are more common among blacks. This study produces the
similar results about the distribution of drug use and
delinquent/criminal behavior among racial and ethnic

groups. For both youths and adults, Whites reported the

219

highest level of drug use, followed by Hispanics, Blacks,
and finally Asians. On the other hand, Blacks reported the
highest level of delinquency and criminality, followed by
White youths, and Hispanic youths. Asian youths had the
lowest level of delinquent/criminal behavior.

The hypothesis that elements of social bond vary by
race and ethnicity is supported. For youths, Whites
reported the strongest attachment to parents followed by
Blacks, and then Hispanics, and Asian 5 reported the
weakest attachment to parents. Black children reported the
strongest attachment to school, followed by Hispanics, and
then Asians, and White children were least attached to
school. Black children reported a higher level of
religiosity than White children, but no variations of
religiosity between Whites, Asians, and Hispanics were
found.

Elements of social bond differ between youths and
adults. For adults, the indicators of social bond such as
education level, marital status, and employment status were
different between racial and ethnic groups. Attachment to
friends, and religiosity also vary significantly by race
and ethnicity. Paralleling with finding for youths, Black
adults reported higher level of religiosity than White

adults as well.

770

The hypothesis that Blacks, Asian Americans and
Hispanic Americans differ from Whites on variables
reflecting differential association is partially supported.
For youths, the number of delinquent peers was invariant by
race and ethnicity. For adults, however, the number of
criminal peers varied by race and ethnicity. Whites
reported more criminal peers than other minority groups.
For both youths and adults, definitions favorable or
unfavorable to drug use vary significantly across racial
and ethnic groups. Asians disapproved of drug use more
strongly than all the other groups. Bearing in mind that
the learned definitions of drug use is the strongest
predictor of drug use, it is not surprise that Asians
reported the lowest level of drug use. The hypothesis that
self—control varies by race and ethnicity is supported for
both youths and adults as well. Generally whites reported
less self—control than minority groups. In brief, the
hypotheses that the theoretical constructs of social bonds,
differential association, and self—control vary across

racial and ethnic groups are supported in the study.

Causal Processes leading to Drug Use and

Delinquency/Criminality by Race and Ethnicity

Causal Processes by Race and Ethnicity for Youths

Although the literature review suggests that
influences and the causal processes leading to crime and
delinquency are invariant across racial and ethnic groups,
the results of the study do not suggest this regularity.
The integrated causal model's explanatory power for drug
use and delinquent/criminal behavior varies by racial and
ethnic group. Generally, the model explains the largest
variance in drug use for Whites, followed by Blacks, then
Hispanics, and the least for Asians. However, the causal
mechanisms explain the largest variance in delinquent
behavior for Asians, followed by Whites, then Hispanics,
and the least for Blacks.

Contrary to Matsueda and Heimer’s (1987) finding that
the effect of definitions on delinquency was invariant,
this study finds that the effect of learned definitions on
drug use vary significantly across racial and ethnic
groups. The effect of learned definition on drug use is
much larger for whites than for blacks and Hispanics, and
it is almost as twice the effect for Asians. However,

consistent with Matsueda and Heimer’s (1987) finding, the

’7’)?

effect of learned definitions on delinquency was invariant
across racial and ethnic groups.

Also, the effect of delinquent peers on drug use was
invariant across racial and ethnic groups, and the effect
of delinquent peers on delinquency varied only between
Asians and Whites. As delinquent peers exerted substantial
effect on delinquency for Whites, it did not affect an
Asian's involvement in delinquent acts.

As postulated by Gottfredson and Hirschi (1990), the
effects of self-control on drug use and delinquency are
invariant across racial and ethnic groups in the study.
Gottfredson and Hirschi (1990) suggested that crime or
analogous behavior is carried out only when the opportunity
to engage in the behavior is present. This assertion is
supported among Whites and Hispanics. For Asians,
opportunity does not affect drug use or delinquency, and
for Blacks, opportunity only affect one's involvement of
delinquent behavior but not drug use.

The associations of social bonds and self—control with
delinquent peers are similar for the racial and ethnic
groups. The effects of background variables on delinquent
peers vary only between Asians and Whites. Age had a larger
effect, and sex had a smaller effect on delinquent peers

for Asians than for other groups.

223

Social bonds seem to affect learned definitions
differently for Blacks, Hispanics, and Whites. Attachment
to parents and religiosity had larger effects on learning
,definitions favorable and unfavorable to law violations for
Whites than for Blacks and Hispanics. Age exerted larger
effects on pro—drug definitions for Whites and Hispanics,
but sex exerted larger effects on definition for Blacks
than other groups.

The hypothesis that the effects of social bonds on
self—control are invariant by race and ethnicity was not
supported. For Whites and Hispanics, attachment to parents
was most strongly related to self-control, but for Blacks
and Asians, attachment to school was most strongly related
to self-control. The effects of sex on self-control were
much larger for Blacks and Whites than for Hispanics and
Asians. Household income exerted larger effects on self—
control for Asians and Blacks than for Whites and
Hispanics.

The total effects of background variables on social
control are significantly different between Blacks and
Whites, and Hispanics and Whites respectively, but not
different between Asians and Whites.

Broken family exerted very limited effects on social

bonds across racial and ethnic groups. It affected the

{Q
[U
4:.

learned definitions moderately but significantly only for
Whites and Hispanics. This finding is inconsistent with
Matsueda and Heimer's (1987) result that the total effect
of broken homes on delinquency is much larger for blacks

than nonblacks.

Causal Processes by Race and Ethnicity for Adults

The findings for adults are not fully parallel to the
findings for youths. The causal process explains
substantial variances of drug use for each group. The
order of the sizes of explained variances of drug use is
the most for Whites, second for Hispanics, then Blacks, and
the least for Asians. The ability to explain the variance
of criminal behavior, however, is very limited. The model
explains less than 20 percent of variance in criminal
behavior for Whites, Blacks, and Hispanics, and it explains
less than 10 percent of variance of criminal behavior for
Asians.

The learned definitions exert very limited effects for
each group, instead, criminal peers present substantial
effects on drug use and criminal behavior. Contrary to the
findings for youths, the effect of criminal peers on drug
use and criminal behavior varies by race and ethnicity.

The effects on drug use and criminal behavior for Whites

to
to
Ur

and Asians are about twice that for Blacks and Hispanics.
The effects of self-control on drug use and criminal
behavior varied between Blacks and Whites, and Hispanics
and Whites as well. The size of the effects of self—
control on drug use for Blacks and Hispanics were more than
twice that for Whites and Asians, but on the contrary, the
effects of self-control on criminal behavior were larger
for Blacks and Hispanics than for Whites and Asians.

Different from the findings for youth, the
associations connecting social bonds and self—control to
criminal peers vary significantly depending on the racial
and ethnic group. Self-control exerts the largest effect
on delinquent peers for each group, but the effect is much
larger for Asians than for others. The effect of education
level on delinquent peers is much larger for Asians than
for others. On the other hand, religiosity exerted
significant impact on the criminal peers for Whites and
Blacks, and moderate effect for Hispanics, but no effect
for Asians.

The effects of social bond and perceived risk on
learned definitions vary between Blacks, Asians, and
Whites. Perceived risk exerted least effect for Blacks
than for others. Religiosity again, had the largest effect

for Whites than for others. The effects of background

226

variables on learned definitions were invariant across
racial and ethnic groups.

The hypothesis that the effects of social bonds on
self-control are invariant by race and ethnicity was not
supported for adults as well. Religiosity exerted
substantial effects on self—control for Whites, Blacks, and
Hispanics, but not for Asians. The effect of marital
status on self—control was much larger for Asians than for
others.

The hypothesis that effects of background variables on
social bonds are invariant by race and ethnicity is not
supported either. Household income exerted significant
effect on attachment to friends for other groups except

Asians.

Summary

In summary, the findings of the tests across racial
and ethnic groups indicate the integrated model could be
applied for each racial and ethnic group. However, the
overall causal mechanisms between background variables,
social bonds, differential associations and self-control
seem to vary by race and ethnicity as well as age. In
terms of explanatory power, for both youths and adults, the

model explains the most variance in drug use for Whites,

t‘J
Ix)
\l

and the least for Asians. On the other hand, it explains
the most variances of delinquent behavior for Asians among
youths, but the least variances in criminal behavior for
Asians among adults. This finding suggests the need for
further research to understand the determinants of criminal
behavior for Asians youths and adults.

The parameters of the causal linkages specified in the
model also vary between the racial and ethnic groups.
Among youths, learned definitions has a much larger effect
on drug use for Whites than for other groups. Self-control
exerts the largest effect on delinquency for Asians. Among
adults, the effects of association with delinquent peers on
both drug use and criminal behavior for Whites, and Asians
are twice that for Blacks, and Hispanics. On the other
hand, the effects of self—control on both types of crime
for Blacks, and Hispanics are more than twice that for
Whites and Asians. In general, while differential
associations and self-control are both significant
determinants of criminality, differential associations have
relatively stronger effects for Whites and Asians than for
Blacks and Hispanics, and self—control is the leading
factor of crime for Black and Hispanics.

The determinative process of differential

associations, and self—control vary significantly across

 

race and ethnicity. Among youths, attachment to parents
exerts much larger effects for Whites and Hispanics, but
attachment to school has the leading effects for Blacks and
Asians. The biggest gender difference in differential
associations and self-control is among Blacks. Among
adults, religiosity as an alternative institution has
substantial effects on drug use and criminal behavior
through the process of decreasing the association with
delinquent peers, and increasing self-control for Whites,
Blacks, as well as Hispanics. But religiosity does not
affect Asians much. Instead, education and marriage are
the two important predictors of an Asian's attachment to

delinquent peers and self—control.

Limitations of the Present Study

Several limitations of the present study need to be
address. First, the data are limited by being cross—
sectional and by being collected just in 2001. The
consequence is that some theoretical causal processes could
not be tested in the current study. For example, a
critical argument of differential association theory is
whether delinquency causes delinquent peers, or delinquent

peers causes delinquency. Without knowing the

 

delinquent/criminal history of the respondent, this
controversy could not be addressed in the current study.

The second limitation concerns the measures of the
core theoretical constructs. Limited by the available
information, the latent variable of delinquent peers for
youths was operationalized as perceived number of students
in the grade who were involved in delinquent behavior.

This variable was not able to reflect an individual’s
decision about who is chosen to hang out with; therefore it
influenced the reliability of the model results.

It may be argued that the specific measures of social
bonding used in the test of the integrated model do not
adequately represent the relevant dimensions of bonding.
The current study focused upon external controls reflecting
involvement in and commitment to conventional institutions
and activities. The four measures of social bonds for
youths are attachment to parents, attachment to school,
involvement, and religiosity. In the process of
operationalization, attachment to parents focused on the
items of parents’ supervision of homework, how to spend the
spare time etc. The items of intimacy that used by Hirschi
(1969) are not available in the study.

Attachment to school mainly reflects respondent's

overall feeling about school. The academic performance and

 

self—perceived ability—two elements that were included in
Hirschi’s (1969) study—are also not available in the
current study. However, the study added religiosity as
another institution that may influence conventional
bonding. The results did indicate that religiosity is an
important dimension of social bonding. Although the
measures of social bonds may not perfectly reflect the
original theoretical constructs, this study can assess the
extent to which the variables used in the analysis captures
the variance of the total set of social bonds.

Lacking the longitudinal information also limits the
power of testing self-control theory. Gottfredson and
Hirschi (1990) argued that crimes are events that provide
immediate gratification, and the single factor of crime—
self—control is mainly the consequence of insufficient
child-rearing at the early age, therefore, longitudinal
analysis was not necessary to understand the criminality.
However, a series of empirical studies about life course
development and delinquency and crime conducted by Sampson
and Laub (1990, 1992, 1993) show that a broad range of
social phenomena (e.g., occupational attainment,
opportunity structures, marital attachment) over the life
course explain changes in crime and deviance over the life

span. Sampson and Laub suggest, there is much intra—

 

individual variability in the adult life-course that is not
reducible to levels of self—control that remain constant
within individuals. Findings of the current study also
suggest that adulthood events such as education, marriage,
and employment status are significant factors that
influence self-control. However, the question of whether
the child-rearing practices are confounded with the
adulthood activities, could not be answered without the
longitudinal analysis. Therefore, knowing the childhood
experience and also the adulthood experience are necessary

to determine the traits of self-control.

Implication for Theory

This study has examined an explanatory model of drug
use and delinquent behavior from three theoretical
perspectives by age and by race and ethnicity. Multiple
findings have been derived from the current study, which
provides several implications for theory and future study.

First, the study confirms the main theses of
differential association and self-control theories.
However, it also verifies that the main constructs of

differential association, social bond, and self-control

 

if

 

theories are not competing or alternative. Instead, they
are complementary and integrated with each other. Self-
control cannot explain away the effects of the differential
association variables on criminal behavior. Social bonds
operate through delinquent/criminal peers, learned
definitions, and self—control. The finding support the
explanation that weak conventional bonding leads youths to
become bonded to peer groups, and decreases the strength of
conventional beliefs, and decreases self-control, which in
turn provide positive reinforcements for
delinquent/criminal behavior.

The cross-sectional data limited the model’s
explanatory power to verify the relationship between
delinquent peers and delinquent/criminal behavior. This
limitation also restricts the model’s ability to understand
the process of self—control. Are the adulthood transition
activities related to child—rearing practice? Are there
independent effects of these factors? To answer these
questions, future study should use a longitudinal design
instead of a cross-sectional dataset.

Another possible modification is to establish paths
between drug use and delinquent/criminal behavior.
Thornberry (1987) argued that most variables have

reciprocal causal effects on one another. The cross-

 

sectional analysis could add paths in the model to test
whether reciprocal causal effects exist between drug use
and other delinquent/criminal behavior. However, it cannot
tell the order of the two types of deviant behavior. Does
drug use cause other criminal behavior or does criminal
behavior cause drug use? Again, to answer this question,
longitudinal analysis is necessary.

The second implication is that the current study
provides further insight into the relationships between
race, ethnicity and crime. Results from the current study
have rejected the hypotheses that variables reflecting the
core theoretical constructs are invariant across race and
ethnicity. There is evidence of significant differences in
the core theoretical variables between whites and other
racial and ethnic groups. In contrast to previous research
(Hirschi 1969), except for a stronger attachment to
parents, white youths generally reported weaker bonds to
conventional society than did youth in racial or ethnic
minority groups. White youths also reported the lowest
level of self-control. Since white youths did report the
highest level of drug use relative to groups, these
findings support the current theoretical themes. However,
if the findings of social bonds, and self-control in this

study could be generalized, the model’s power of explaining

234

 

 

 

the racial and ethnic variations of more serious crime such
as violent crimes, in which Blacks and Hispanics usually
report higher crime rates, is then questionable.

In the current study, Asians reported relatively weak
external social bond, such as attachment to parents, and
religiosity, however, Asians reported strong internal
control reflecting personal beliefs in the moral validity
of conventional norms (definitions favorable and
unfavorable to law violation). On the other hand, Blacks
presented relatively strong external social bond as
attachment to school and religiosity but weak internal
control. Considering that Asians reported the lowest level
of drug use and delinquent behavior, and Blacks reported
relatively higher level of drug use and the highest level
of delinquent behavior, this finding suggests that external
social bonds work through the beliefs or the learned
definitions of law violation to influence the decision of
involvement in criminal behavior. This again favors the
integrated theoretical approach over individual theoretical
models.

Third, this study finds that gender differences of the
core theoretical constructs are significant. In general,
males reported significantly lower level of self~control

and expressed more excessive definitions favorable to law

Id
Lu
U1

 

 

violation than female did. Furthermore, there are race,
gender and age interactive effects. For instance, the
gender differences of the number of delinquent peers was
found only among white adults. Males and females reported
the same level of learned definitions favorable and
unfavorable to drug use, and self—control among Asian
youths, while significant gender differences in these
variables are found in other racial and ethnic groups.
These issues have not been addressed in any depth in the
current study. Further research into the theoretical
model's explanations of gender differences is warranted.
Fourth, findings of this study suggest that youths and
adults differ somehow in the processes leading to drug use
and delinquent/criminal behavior. Self—control is a
relatively stable factor during the adolescent period,
however, it changes after one gets into the adulthood.
Young adults presented lower level of self-control than old
adults. This finding provides empirical evidence of
Sampson and Laub's (1993) argument that social phenomena
over the life course explain changes in crime and deviance
over the life span. Self-control is not only the results
of child—rearing practices in early childhood, it is also a

function of social events over the life course.

236

 

Of more interesting is that the comparison of causal
processes and parameters across race and ethnicity
demonstrate that although the general integrated model
could be applied across racial and ethnic groups, the
causal linkages and the relative explanatory power of
individual factor vary by race and ethnicity, especially
for adults. This empirical evidence contradicts previous
theorists’ assertion that factors causing crime and
delinquency are common across race and ethnic groups, and
existing theories are applicable to racial and ethnic
minorities equally well. Why the causal mechanisms and the
power of each factor vary between different racial and
ethnic groups?

As indicated by the descriptive analysis, Asians had
relatively advanced socioeconomic situation. They had
higher education level, and lived in a household with
higher income. The most majorities of Asian youths lived in
stable family with both parents presented. Asian adults
also had more stable employment and marital status. Blacks
and Hispanics, quite the opposite, are situated in a very
different location in society. More than a third of Blacks
lived in a household with income less than $20,000 and more
than 20 percent of Black adults had less than high school

education. More than 60 percent of Black youths lived in a

237

 

 

 

broken family. Those individual social economic
differences impact delinquency through the effects on
social bonding, differential association and self-control
variables. Viewed in a broader perspective, these results
raise questions concerning the role of social structure on
race, ethnicity, control, cultural norms, and delinquency.
In a study of relationships between race, crime and
urban inequality, Sampson and Wilson (1995) point out the
social transformation of the inner city in recent decades—
for example the deindustrialization of central cities, the
exodus of middle and upper-income black families from the
inner city, urban renewal and forced migration, and
deliberate policy decisions to concentrate minorities and
the poor in public housing—has resulted in an increased
concentration of the most disadvantaged segments of the
urban black population—especially poor, female-headed
families with children. This concentration and segregation
of poverty, heterogeneity, mutual distrust, and
institutional inability foster social isolation, which
impede communication and obstruct the quest for common
values, thereby fostering cultural diversity with respect
to nondelinquent values. The observed differences of
association with delinquent peers, and the definitions

favorable to illegal behavior in this study, therefore,

1x)
DJ
00

 

 

 

could be the results of the social structural differences
of different racial and ethnic groups. The social
structural differences could also result in the differences
in social control. As Rutter and Giller (1983:185) have
argued, socioeconomic disadvantage has potentially adverse
effects on parents, such that parental difficulties are
more likely to develop and good parenting is impeded.
Sampson and Laub (1993) also find that structural factors
strongly affect family and school social control
mechanisms. To understand why and how the social processes
leading to delinquency are different between racial and
ethnic groups, analyses of macro-social forces such as
community-level structural social disorganization and
cultural social disorganization on delinquent behavior are

necessary.

Implications for policy

The results of this theoretical test draw attention to
the critical role that learned definitions, delinquent
peers, and self—control play in the production of
delinquent behavior. It also has been evidenced that the

chain of causation starts from the personal and family

rd
Lu
\0

social economic location in society. The background
variables of individuals’ age, sex, household income, and
family structure influence one’s bonding to conventional
society as well as differential association and self—
control. And the social bonds of attachment to parents,
attachment to school, involvement in social activities plus
religiosity influence delinquency through either increases
or decrease delinquent companions, prodelinquent
definitions, and ultimately delinquent behavior. Self-
control is specifically important in the causal process.
It not only directly impacts delinquency, but also
influences delinquency through the effect on delinquent
peers. These findings have palpable implications for
delinquency prevention efforts and treatment programs.
First the integrated model of delinquency urges us to
seriously consider the role that families play in
delinquency causation. As evidenced in the study, parents’
supervision and their definitions or beliefs about illegal
behavior had significant influences on children’s
definitions favorable or unfavorable to law violation and
their association with delinquent peers. Parents'
supervision and beliefs also affect self-control
substantially. Intervention programs that focus on

enhancing the ability of parents to monitor, recognize, and

240

 

 

discipline the misbehavior of their children, as suggested
by other researchers, would be most promising in increasing
children’s self-control and reducing antisocial behavior
and delinquency (Gottfredson and Hirsch, 1990; Hirschi,
1995; Tremblay et al., 1992; Laub et al., 1995).

Findings of the study also imply that school is an
effective socializing agency against delinquent behavior.
Strong attachment to school provided sufficient
socialization, lessened prodelinquent definitions, and
enhanced children’s self-control, which in turn reduced the
involvement in delinquent behavior. As Hirschi (1969)
pointed out, positive feelings toward controlling
institutions and persons in authority are the first line of
social control (pp.127). In order to establish a strong
attachment to school, strategies should focus on improving
students’ school performance. The improvements in
performance could reduce the adversity of school, change
students’ attitudes toward school and teachers (Agnew
1995), and improve interpersonal relations between
individuals in the school system.

The third implication is that policies improving the
development of social capital in adulthood would
effectively change adult antisocial behaviors. The

analysis of adult criminality in this research indicates

 

 

that education level, employment, and marital status are
significantly related to adult criminal behavior. Sampson
and Laub (1993) suggested that when an individual has
cohesive ties in marriage or work, these ties involve
investments made by both the individuals and their spouses
and employers. The high impact and quality of these
interdependent, reciprocal investments bring about informal
social control, and deter individuals from acting in ways
that threaten such investments. The programs suggested by
Sampson and Laub (1993) include providing counseling,
education, and job training in correctional facilities and
using community-based corrections to assist the inmate in
his or her reintegration into the normal world.

Finally the findings about differences in causal
processes and parameters between racial and ethnic groups
in the research suggest that the distinct racial and ethnic
patterns of delinquency and criminality may relate to the
distinct social location and variant patterns of social
organization of the different racial and ethnic groups.
Crime prevention policy therefore should target not only
the elements of social process but also the social

structure that sets the stages for process.

 

References

Agnew, R.

1993 Why do they do it? An examination of the
intervening mechanisms between "social control"
variables and delinquency. The Journal of
Research in Crime and Delinquency 30 (3): 245.

Akers, R. L., Krohn, M. D., Lanza-Kaduce, L. and
Radosevick, M.
1979 Social learning and deviant behavior: a specific
test of a general theory. American Sociological

Review 44:635-55.

Anderson, E.
1990 Streetwise. Chicago: University of Chicago
Press.

Arneklev, B. J., Grasmick, H. G., Tittle, C. R., and
Bursick, R. J., Jr.
1993 Low self—control and imprudent behavior. Journal
of Quantitative Criminology, 9:225-247.

Aseltine, Robert H.
1995 A Reconsideration of Parental and Peer Influences
on Adolescent Deviance. Journal of Health and
Social Behavior 36: 103—121.

Austin, G. A., Prendergast, M. L., and Lee, Ho
1989 Substance abuse among Asian American youth
(Prevention Research Update No. 5, pp. 1-13).
Portland, OR: Northwest Regional Educational
Laboratory.

Bachman, J. G., Johnston, L. D., and O’Malley, P. M.
1998 Explaining recent increase in students’ marijuana

use: Impacts of perceived risks and disapproval,
1976 through 1996.

Baer, D.
2002 Latent Variable Structural Equation Models.
ICPSR Summer Program.

Benson and Moore

1992 Are white-collar and common offenders the same?
Journal of Research in Crime and Delinquency, 29:
251—272.

Blau, J. R. and Blau, P. M.

1982 The cost of inequality: Metropolitan structure
and violent crime. American Sociological Review,
47: 114-29.

Blumstein, A.

1993 Racial disproportionality of U.S. prison
populations revised. Colorado Law Review,
64:743—60.

Brearley
1932 Homicide in the United States. Chapel Hill, N.C.:

University of North Carolina Press.

Britt, C. L., and Gottfredson, M. R.
2003 Control Theories of Crime and Delinquency.
Britt, C. L. and Gottfredson and M. R. (ed.).
New Brunswick: Transaction Publishers.

Brook, J. S., Brook, D. W., Scovell Gordon, S., Whiteman,
M., and Cohen, P.
1990 The psychosocial etiology of adolescent drug use:
A family interactional approach. Genet Soc Gen
Psychol Monogr, 116 (2).

Brownfield and Sorenson
1993 Self-control and juvenile delinquency:
Theoretical issues and an empirical assessment of
selected elements of a general theory of crime.
Deviant Behavior 14(3):243—264.

Bruce, M. A.
2000 Violence among African Americans: A conceptual
assessment of potential explanations. Journal of
Contemporary Criminal Justice 16(2):171-194.

2M

Burton, V. 8., Jr., Cullen, F. T., Evans, D., and Dunaway,
R. G.
1993 Reconsidering strain: Operationalization, rival
theories, and adult criminality. Journal of
Quantitative Criminology 10:213-239.

Burton, V. 8., Jr., Cullen, F. T., Evans, D., Alarid, L. F.

and Dunaway, R. G.
1998 Gender, self-control, and crime. Journal of
Research in Crime and Delinquency 35:123-147.

Conger, R. D.

1976 Social control and social learning models of
delinquent behavior: A synthesis. Criminology,
14: 17-40.

Cressey, D. R.
1952 Application and verification of the differential
association theory. Journal of Criminal Law,
Criminology and Police Science 43:43-53.

Cullen, F. T., and Agnew, R.

1998 Criminological Theory: Past to present—Essential
readings. Los Angeles: Roxbury.
Curtis, L.
1975 Violence, Race, and Culture. Lexington, MA:
Heath.

Davidson, R. N.
1981 Crime and Environment, New York: St. Martin’s
Press p87—89.

DiIulio, J., Jr.
1995 White lies about black crime. Public Interest,
118: 30-44.

Einstadter, W. and Henry, S.
1995 Criminological Theory: An analysis of its
underlying assumptions. Philadelphia: Harcourt
Brace College Publishers

Elliot, D. S., and Ageton, S. S.
1980 Reconciling race and class differences in self-
reported and official estimates of delinquency.
American Sociological Review, 45:95-110.

25

 

 

Elliott, D. S., Ageton, S. 8., and Canter, R. J.
1979 An integrated theoretical perspective on
delinquent behavior. Journal of Research in
Crime and Delinquency, 16:3-27.

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

Erickson, K. G., Crosnoe, R., and Dornbusch, S. M.

2000 A social process model of adolescent deviance:
combining social control and differential
association perspectives. Journal of Youth and

Adolescence, 29(4):395—425.

Evans, T. D., Cullen, F. T., Burton, V. S., Dunaway, R. G.,
Jr., and Benson, M. L.

1997 The social consequences of self-control: Testing
the general theory of crime. Criminology, 35:
475-501.

Cernkovich, S. A. and Giordano, P. C.
1992 School bonding, race, and delinquency.
Criminology, 30(2):261

Gibbs, J. J. and Giever, D.

1995 Self-control and its manifestations among
university students: An empirical test of
Gottfredson and Hirschi’s general theory.
Justice Quarterly 12:231-235.

Gibbs, J. J., Giever, D., and Martin, J. S.

1998 Parental management and self—control: An
empirical test of Gottfredson and Hirschi’s
general theory. Journal of Research in Crime and
Delinquency 35:40-70.

Glueck, S.
1956 Theory and fact in criminology: A criticism of
differential association. British Journal of
Delinquency 7:92-109.

Gottfredson, M. R. and Hirschi, T.
1990 A general theory of crime. Stanford, CA:
Stanford University Press.

Grasmick
1994 Testing the core empirical implications of
Gottfredson and Hirschi’s general theory of
crime. Journal of Research in Crime and
Delinquency, 30: 5-29.

Guertin, W. H. and Bailey, J. P., Jr.
1970 Introduction to modern factor analysis. Ann
Arbor, MI: Edwards Brothers.

Hagan, J. and Peterson, R. D.
1995 Crime inequality in America: Patterns and
Consequences. In Crime and Inequality (eds.) by
Hagan and Peterson. Stanford, CA: Stanford
University Press.

Haney Lopez, Ian F.
1995 The Social Construction of Race: Some
Observations on Illusion, Fabrication, and
Choice. Harvard Civil Rights—Civil Liberties Law
Review. 29:39-53.

Hawkins, D. F.
1993 Crime and ethnicity. In The Socio—Economics of
Crime and Justice, edited by B. Forst. Armonk,
NY: M.E. Sharpe.

1995 Ethnicity, race, and crime: A review of selected
studies. In Ethnicity, Race and Crime:
Perspectives Across Time and Place, edited by D.
F. Hawkins. Albany, NY: State University of New
York Press.

Hawkins, D. F., Laub, J. H., Lauritsen, J. L., and Cothern,
L.
2000 Eace, Ethnicity, and Serious and Violent Juvenile
Offending. Juvenile Justice Bulletin. OJJDP.

Hay, C.
2001 Parenting, self-control, and delinquency: A test
of self-control theory. Criminology, 39(3):707—
736.

247

Haynie, D. L.
2002 Friendship Networks and Delinquency: The relative
nature of peer delinquency. Journal of
Quantitative Criminology. 18(2): 99—134

Hernstein, R. J. and Murray, C.
1994 The Bell Curve: Intelligence and Class Structure
in American Life. Now York: Free Press.

Hindelang, M. J.
1978 An introduction to the national crime survey.
Washinton: Department of Justice, Law Enforcement
Assistance Administration.

Hirschi, T.

1995 The family. In James Q. Wilson and Joan
Petersilia (eds.), Crime. San Francisco: ICS
Press.

Hirschi, T. and Gottfredson, M. R.

2003 Punishment of children from the perspective of
control theory. In Control Theory of Crime and
Delinquency (eds.) by Britt and Gottfredson. New
Brunswich, NJ: Transaction Publishers.

Jencks, C.

1992 Rethinking social policy: Race, poverty, and the
underclass. Cambridge, MA: Harvard University
Press.

Jessor, R., and Jessor, S.L.

1977 Problem behavior and psychosocial development: A
longitudinal study of youth. New York: Academic
Press.

Kandel, D. B.

1974 Drug and drinking behavior among youth. Annual
Review of Sociology. New York: Annual Reviews,
Inc.

Kandel, D. B., Single, E., and Kessler, R.

1976 The Epidemiology of drug use among New York
state high school students: Distribution, trends
and change in rates of use. American Journal of
Public Health, 66:43-53.

Kaplan, H. B.
1975 Self-attitude and deviant behavior. Pacific
Palisades, CA: Goodyear Publishing Co.

Kornhauser
1978 Social sources of delinquency. Chicago:
University of Chicago Press.

Krohn, M., D., Massey, J. L., Skinner, W. F., and Lauer, R.
M.
1983 Social bonding theory and adolescent cigarette
smoking: A longitudinal analysis. Journal of
Health and Social Behavior, 24: 337-49.

LaFree, G. & Russell, K. K.
1993 "The Argument for Studying Race and Crime"
Journal of Criminal Justice Education 4:273—289.

LaFree, G.

1994 "Race and Crime Trends in the United States,
1946-1990" Ethnicity, Race and Crime:
Perspectives Across Time and Place 170-193, SUNY
Press.

Laub, J. H., Sampson, R. J., Corbett, R. P., Jr., and
Smith, J. S.

1995 The public policy implications of a life-course
perspective on crime. In Hugh Barlow (eds.) Crime
and Public Policy: Putting Theory to Work.
Boulder, Colorado: Westview Press.

Li, S. D.
2004 The impacts of self-control and social bonds on
juvenile delinquency in a national sample of
midadolescents. Deviant Behavior, 25: 351-373.

Lombroso, Cesare
1876 Criminal Man.

Ma, G. X., and Shive, S.
2000 A comparative analysis of perceived risks and
substance abuse among ethnic groups. Addictive
Behaviors, 25 (3): 361-371.

MacLean, B., and Milovanovic, D.
1990 Racism, empiricism, and criminal justice (eds).
Vancouver: Collective Press.

Marcoulides, George A. and Scott L. Hershberger
1997 Multivariate Statistical Methods: A First Course.
Hillsdale, N.J.: Lawrence Erlbaum Associates.

Marcros, A. C., Bahr, S. J., and Johnson, R. E.
1986 Test of bonding/association theory of adolescent
drug use. Social Forces, 65(1): 135—161.

Gimenez, M. E.

1989 Latino/Hispanic—Who Needs a Name?: The Case
Against a Standardized Terminology.
International Journal of Health Services
19(3):567—571.

Massey, J. L., and Krohn, M. D.
1986 A longitudinal examination of an integrated
social process model of deviant behavior. Social
Forces, 65(1): 106-134.

Mastsueda and Heimer
1987 Testing control theory and differential
association: A causal modeling approach.
American Sociological Review. 47(4): 489-504.

McNulty, T. L. and Bellair, P. E.
2004 Explaining racial and ethnic differences in
serious adolescent violent behavior. Criminology
41(3):709—748.

McNulty, T. L.

2001 Assessing the race-violence relationship at the
macro level: The assumption of racial invariance
and the problem of restricted distributions.
Criminology 39(2):467-489.

Mednick S. A., Moffitt, T. E. and Stack, S. A.
1988 The causes of crime: new biological approaches.
New York: Cambridge.

Merton, R. K.
1938 Social structure and anomie. American
Sociological Review. 3(5): 672—682.

Muthen, L. K. and Muthen, B. O.
2003 Mplus User’s Guide (2” ed.) Los Angeles, CA:
Muthen & Muthen.

Nagin, D. S. and Paternoster, R.

1993 Enduring individual differences and rational
choice theories of crime. Law and Society Review
3:467—496.

National Institute on Drug Abuse
1995 Drug Use among Racial/Ethnic Minorities.
Bethesda, MD: NIH Publication.

NSDUH
2002 The 2002 National Survey on Drug Use and Health,
Table 1.27B. Rodoville, MD.: Substance Abuse
and Mental Health Services Administration

Oetting, E. R., and Beauvais, F.
1990 Adolescent drug use: Findings of national and
local surveys. Journal of Consulting Clinical
Psychology, 58(4), 385-394.

Piquero, A., and Tibbetts, S.

1996 Specifying the direct and indirect effects of low
self-control and situational factors in
offenders’ decision making: Toward a more
complete model of rational offending. Justice
Quarterly, 13:481-510.

Polakowski, M.

1994 Linking self— and social-control with deviance:
Illuminating the structure underlying a general
theory of crime and its relation to deviant
activity. Journal of Quantitative Criminology,
10: 41—78.

Pratt, T. C., and Cullen, F. T.
2000 The empirical status of Gottfredson and Hirschi’s
general theory of crime: A meta—analysis.
Criminology, 38: 931-964.

Prendergast, M. L., Austin, G. A., Maton, K., and Baker, R.
1989 Substance abuse among Black youth. Prevention
Research Update 4:1-27.

Raykov, T. M. and Marcoulides, G. A.
2000 A First Course in Structural Equation Modeling.
Mahwah, NJ: Lawrence Erlbaum.

Reiss, A. J., Jr. and Rhodes, L.
1964 An Empirical Test of Differential Association
Theory. Journal of Research in Crime and
Delinquency 1:5-18. '

Rutter and Giller
1983 Juvenile Delinquency: Trends and Perspectives.
Harmondsworth: Penguin.

Sampson, R. J. and Laub, J. H.
1990 Crime and deviance over the life course: The
salience of adult social bonds. American
Sociological Review 55:609-627.

1992 Crime and deviance in the life course. Annual
Review of Sociology 18:63-84.

1993 Crime in the Making: Pathways and Turning Points
Through Life. Cambridge: Harvard University
Press.

Sampson, R.J. and Lauritsen, J. L.
1997 "Racial and Ethnic Disparities in Crime and
Criminal Justice in the United States." In M.
Tonry (ed.) Ethnicity, Crime, and Immigration:
Comparative and Cross-National Perspectives,
Crime and Justice: A Review of Research (vol.
21). Chicago: The University of Chicago Press.

Sampson, R.J., and Wilson, W.J.
1995. Toward a theory of race, crime, and urban
inequality. In Crime and Inequality, edited by J.
Hagan and R. Peterson. Stanford, CA: Stanford
University Press, pp. 37-54.

Shaw, C. R. and McKay, H. D.
1942. Juvenile Delinquency and Urban Areas. Chicago:
University of Chicago Press.

Sheldon, W. H.
1936 Psychology and the Promethean Will. New York:
Harper.

1940 Varieties of Human Physique. New York: Harper.
1942 The Varieties of Temperament. New York: Harper.

Shoemaker, D.

1984 Theories of delinquency: An examination of
delinquent behavior. New York: Oxford University
Press.

Snyder, H. N.
1999 Juvenile Arrests 1998. Bulletin. Washington,
DC: U.S. Department of Justice, Office of Justice
Programs, Office of Juvenile Justice and
Delinquency Prevention.

Stitt, B. G. and Giacopassi, D., J.
1992 Alcohol availability and alcohol~related crime.
Criminal Justice Review 17(2):268.

Sutherland, E. H. and Cressey, D. R.
1966 Principles of Criminology. Philadelphia: J. B.
Lippincott Company.

1978 Criminology. Tenth Edition. Philadelphia:
Lippincott.

Sykes, G. and Matza, D.

1957 Techniques of neutralization: a theory of
delinquency. American Sociological Review
43:643-56.

Thornberry, T. P.
1987 Toward an interactional theory of delinquency.
Criminology 25.

Tittle, C. R., Burke, M. J. and Jackson, E.
1986 Modeling Sutherland’s theory of differential
association: Toward an empirical clarification.
Social Forces 65:405—432.

Tonry, M. H.
1995 Malign neglect—race, crime, and punishment in
America. New York: Oxford University Press.

Tremblay, R. E. et a1.
1992 Parent and child training to prevent early onset
of delinquency: The Montreal longitudinal—

experimental study. In Joan McCord and Richard
Tremblay (eds.), Preventing Antisocial Behavior:
Interventions From Birth Through Adolescence.
New York: Guilford Press.

Unnever, J. D., Cullen, F. T., and Pratt, T. C.

2003 Parental management, ADHD, and delinquent
involvement: Reassessing Gottfredson and
Hirschi’s general theory. Justice Quarterly
20(3):471.

Walker, 8., Spohn, C., and DeLone, M.

1996 The Color of Justice: Race, Ethnicity, and Crime
in America. Albany: Wadsworth Publishing
Company.

Welte, J. W., Barnes, G. M., Hoffman, J. H., and Dintcheff,
B. A.
1999 Trends in adolescent alcohol and other substance
use: Relationships to trends in peer, parent, and
school influences. Substance Use and Misuse, 34,
1427-1449.

Wilson, J. Q. and Herrnstein, R. J.
1985 Crime and Human Nature. New York: Simon and
Schuster.

Wilson, W. J.
1987 The Truly Disadvantaged. Chicago: University of
Chicago Press.

Winfree, L. T. and Bernat, F. P.

1998 Social learning, self-control, and substance
abuse by eighth grade students: A tale of two
cities. Journal of Drug Issues 28:539-558.

Wolfgang, Marvin E., & Ferracuti, F.
1967 The subculture of violence: Towards an integrated
theory in criminology. London: Tavistock
Publications.

is.)
'JI
.L.

 

 

 

   

(1111;11111111