Youth and substance abuse

Material Information

Youth and substance abuse an extension of social controlbonding theory
Vander Horst, Anthony
Place of Publication:
Denver, Colo.
University of Colorado Denver
Publication Date:
Physical Description:
viii, 78 leaves : ; 28 cm

Thesis/Dissertation Information

Master's ( Master of Arts)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Sociology, CU Denver
Degree Disciplines:
Committee Chair:
Xu, Yili
Committee Members:
Fink, Virginia
Anderson, Richard


Subjects / Keywords:
Youth -- Substance use ( lcsh )
Parental influences ( lcsh )
Parent and child ( lcsh )
Parent and child ( fast )
Parental influences ( fast )
Youth -- Substance use ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 74-78).
General Note:
Department of Sociology
Statement of Responsibility:
by Anthony Vander Horst.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
71801817 ( OCLC )
LD1193.L66 2003m H67 ( lcc )

Full Text
Anthony Vander Horst
B.S., Metropolitan State College of Denver, 1994
B.A. Metropolitan State College of Denver, 1997
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Arts

This thesis for the Master of Arts
degree by
Anthony Vander Horst
has been approved
/ / R -
Richard Anderson

Vander Horst, Anthony (M.A., Sociology)
Youth and Substance Abuse: An Extension of Social Control/Bonding Theory.
Thesis directed by Assistant Professor Yili Xu
Social control/bonding theory asserts that the higher the bond between parents and
their children in the areas of education, religion, conventional attachment, the lower
will be the drug use among those youth. In a multiple regression analysis of the
Monitoring the Future database, 1015 high-school seniors from a nationwide sample
were examined using the social control/bonding theory. Eight independent variables
were tested against the dependent variable, lifetime marijuana use. Drug-using friends
in a youths life will increase the likelihood of youth drug use. Conventional
attachment and religious attachment showed an inverse relationship regarding drug
use. The higher the reported attachment to religion and conventional attitudes, the
less likely youth report to be using drugs.
In addition to the social control/bonding variables, this research included sex
and race and region to explain youth substance abuse. The results suggest support for
differences in marijuana use between the sexs, and a significant effect for race
(dichotomous variable). Region (four categories) shows no significant effect when
youth choose to use marijuana.
Future studies in social control would benefit by including sex and gender,
multiple race classifications and composite dependent variables. Likewise, exploring
multiple drug-uses in relation to region may prove beneficial in understanding region
influence in youth drug choices.
This abstract accurately represents the content of the candidates thesis. I recommend
its publication.

1. INTRODUCTION...............................................1
Arrangement of the Thesis............................3
2. REVIEW OF SOCIAL CONTROL THEORY............................5
Literature Review....................................5
Social Control Model.................................6
Differential Association Theory............................9
Empirical Review....................................13
Systematic Review...................................16
Gender, the Chronology.......................20

A Multiple Regression Model of Social Control/Bonding
And Youth Drug Use................................... 31
Data .................................................31
Variables ............................................35
Independent Variable...........................35
Dependent Variable ........................... 38
4. HYPOTHESES.....................................................40
Hypothesis 1..........................................40
Hypothesis 2..........................................41
Hypothesis 3..........................................41
Theoretical Variables..........................50

5. DISCUSSION...................................................52
Drug-Using Peers...................................53
Educational Attachment.............................54
Parental Attachment................................54
Conventional Attachments...........................55
Religious Attachment...............................56
B. VARIABLE CONSTRUCT........................................71
Independent Variables........................71
Religious Attachment (2-8 composite range)...71
Educational Attachment (2-8 composite range).71
Parental Attachment (2-14 composite range)...71
Drug-Using Friends (3-15 composite range)....71

Conventional Attachment (3-16 composite range).72
Dependent Variables.......................72
Lifetime Marijuana (recoded)..............72
Original.......................... 72
C. FACTOR ANALYSIS...........................................73

Table A-1 Descriptive Statistics.....................64
Table A-2 Crosstabulation
LFTMMJ Marijuana use in your life Region -
4 Categories...................................65
Table A-3 ANOVA-Analysis of Variance Between -
Subjects Factors............................. 66
Table A-4 Tests of Between-Subjects Effects..........66
Table A-5 Coefficients: Number of Respondents........67
Table A-6 Multiple Comparisons-Dependent Variable:
LFTMMJ Marijuana use in your life.
Table A-7 Correlation Matrix.........................69
Table A-8 Model Summary: Number of Cases.............70
Table A-9 ANOVA......................................70
Table C-l Rotated Component Matrix...................73

The goal of this thesis is to explore the strengths and weaknesses of social
control theory in order to formulate additional, valid explanations for why youth
become involved with drugs.
For years, politicians, religious leaders, pundits from many disciplines, and
people from all walks of life have discussed reasons why youth become involved with
drugs. The focus of their discussions are drugs like alcohol, tobacco & marijuana
termed the soft drugs, and amphetamines and barbiturates (speed, cocaine heroin,
opiates) LSD, special K, GHB and other designer drugs (ecstasy, MDNA etc),
referred to as the harder drugs. The federal government created the federal
Controlled Substances Act of 1970, which created five classifications called
schedules in which drugs are controlled in varying degrees ('
From these classifications, many people have informally classified drugs as
hard and soft. Marijuana, although classified as schedule I (no medicinal
purpose) by the federal government, has been informally termed by society as a soft
drug meaning that the detrimental effects are fewer than alcohol or amphetamines.

Many social control theorists believe that Hirschis theory was
underdeveloped when explaining youth delinquency. Social theorists (In particular,
Anastasios C. Marcos, Stephen J. Bahr, and Richard E. Johnson from Brigham Young
University, 1986) argued that Hirschi did not include the influence of religion on
youth drug use. Additionally, they argue that Hirschi misinterprets peer influence by
saying that peers reduce drug-use by youth.
It will be shown that Hirschi believed that peers would reduce drug use.
However, subsequent research consistently finds that peers have a negative influence
on youth, meaning that drug using peers lead some youth into drug use.
So, it will be demonstrated that .researchers expanded social control theory to
offer better explanations as to why youth become involved in delinquent behavior,
specifically those of drug use.
Although Marcos et al. (1986) and others have found significant findings that
help explain why youth become involved with drug use, there are still several
delinquencies in their social control research because they failed to include Sex, Race,
and Region into the model.

Arrangement of the Thesis
In chapter one I construct a model that will both maintain the integrity of
social control theory while adding additional key variables (sex, region and race) to
the model for analyses. Adding additional variables should allow us to develop a
clearer understanding for why youth become involved with drugs.
In chapter two, I review the literature to discover who was responsible for
developing both social control theory and differential association theory; additionally
conduct a systematic review of research related to these theories. Moreover, I will
briefly discuss how Marcos et al. (1986), blended two theories, social control theory
and differential association theory, to remedy their critique of Hirschi. Incidentally,
this blend of two theories made a defining moment for social theory. To this day,
researchers have piggybacked this research and forged a permanent place for social
control/bonding theory. As we continue we will see some of this research.
In chapter three, I explore the more recent findings and updates of social
control theory regarding the new variables of race, gender and region. Additionally, I
explore the influence of other research regarding these three variables.

Chapter four explains the database, including the methods, sample
information, and variable construct development Marcos et al. (1986) used to expand
social control theory. In addition, chapter four explains the model construction,
variable construct and the hypothesis to be tested in this study.
Chapter five reports the results and findings and offers a discussion regarding
the implications. I will also explain in detail how hypothesis and results of this study
relate to the literature review and how future studies might be advanced.

Literature Review
Travis Hirschi (b.1935, d. ), acclaimed father of social control theory,
conducted his research and presented his conclusions in social control theory in an
attempt to explain youth delinquency. As the years in social control theory
progressed, other researchers began to test, and subsequently expanded upon
Hirschis work. Researchers attempted to apply the theory by exploring drug use as a
particular form of delinquency. The Brigham Young University study, headed by
Anastasios C. Marcos, and assisted by Stephen J Bahr, and Richard E. Johnson
recognized Hirschis theoretical likenesses when they said:
Hirschis original theory lacks both conceptual clarity and
specification of causal ordering. His attachment construct is
tremendously broad, including affective and behavioral ties to
parents, school, and peers...although it (attachment) is extremely
broad, it ignores bonds to religion, a major and eminently
conventional institution which could serve to control deviant

As a solution, Marcos et al. (1986) embarked on blending two theories, Social
Control Theory with Differential Association Theory in an attempt to mend Hirschis
theory and to offer a clearer explanation of why youth use drugs.
In the following discussion, I explore how Hirschi conceptualized social
control theory and then examine the 1986 research conducted at Brigham Young
University. Included are the Marcos et al (1986) critique of Hirschi, and the solutions
and findings for mending Hirschis theory.
Social Control Model
Travis Hirschi (1969), argues in his work, Causes of Delinquency, that
humans are inherently nonconforming and antisocial and that we are all animals and
thus naturally capable of committing criminal acts (Hirschi, 1969, p.31).
Hirschis bonding theory states that the more closely a person is tied to
conventional society in any of these ways (attachment, commitment, involvement, &
belie!), the more closely a person is tied in other ways (Hirschi 1969, p. 27).
The bonds upon which Hirschi focuses consisted of four elements or independent
variables: attachment (affective ties toward parents, school and friends); commitment
(youths aspirations for, and behavior consistent with alter attending college and
obtaining a prestigious occupation); involvement (participation in conventional

activities such as spending time on school work); and belief (respect for the moral
validity of the rules of society) (Hirschi, 1969).
Hirschis theory assumes deviance or deviant behavior is natural. Thus, it is
conformity that requires explanation. The questions for Hirschi is not Why do they
conform? but Why dont they conform? Hirschi argues that human conformity is
based on a bond that is developed between an individual and society that keeps him
or her from violating the rules (Hirschi, 1969).
Hirschis theory predicts that adolescents with higher levels of attachment,
commitment, involvement, and belief are less prone to deviate from the norms of
society. In addition to its independent effect on deviance, Hirschi sees each element
of the social bond as reinforcing the other three. Without specifying causal ordering,
Hirschi simply states that the more closely a person is tied to conventional society
in any of these ways, the more closely he is likely to be tied in other ways (1969, p,
27). Moreover, a weakly bonded person is simply free, but not forced or driven, to
commit delinquent acts, as there is no place in pure control theory for motivations to
deviate, such as frustration or peer pressure.
Hirschi believed that peers could lead to a reduction in the probability of
delinquent behavior (Hirschi, 1969), while replication studies found exactly the

opposite (Hindelang, 1973). Other studies found both effects, that is to say that both
delinquent as well as conforming peers influence behavior (Conger 1976; Polk 1971).
Anastasios C. Marcos, Stephen J. Bahr, & Richard E. Johnson conducted
research based on social control theory and published a paper entitled Test of a
Bonding/Association Theory of Adolescent Drug Use (Marcos et al., 1986). They
believed that Hirschi had two major delinquencies in his theory. The first major
problem, in their estimation, was that Hirschi failed to include religion into the model.
Second, they disagreed with Hirschis argument that peer influence would encourage
youth to reduce drug use. Rather, they found that peers negatively influenced drug
According to the Marcos et al. (1986) findings and other research, drug using
peers are consistently found to be the number one influence for why youth become
involved with substance abuse (Marcos, Bahr and Johnson 1986; Aseltine, 1995).
And when parents know their children(s) peers, youth are less likely to smoke
cigarettes (Krohn, Massey & Zielinski, 1988). Clearly, these findings are contrary to
Hirschis postulates regarding peer influence.

These conclusions are the primary reason why Marcos and his team to argue
that Hirschis social control theory lacks both conceptual clarity and specification of
causal ordering... and that the social control influences of parents, school, and peers
are simply too theoretically and empirically distinct to be treated as one construct
(Marcos, 1986, p. 137-8).
Marcos et al. (1986) constructed a new sociological model, when they blended
social control theory and differential association theory in attempt to remedy
Hirschis theoretical shortcomings. Marcos et al. (1986) included differential
association theory because of the strong findings regarding religion and peer
influence reflected in their research and for the benefit of social control theory.
At this point, I explore the origins and premises for differential association
theory, then conclude the chapter by defining the differences and similarities of both
Differential Association Theory
Differential association theory is a branch of social learning theory as developed by
Edwin H. Sutherland and Donald R. Cressys (1978) work in
Differential association theory is a branch of social learning theory as
developed by Edwin H. Sutherland and Donald R. Cresseys (1978) work in

which they argue that behavior patterns are primarily derived from ones close
personal acquaintances. In the words of the authors, differential association
is a learning theory, which focuses on the processes by which individuals
come to commit criminal acts (Sutherland and Cressey 1978).
In contrast to the social control theory belief that delinquency comes from
a breakdown in community allegiance at large, Sutherland and Cressey assert
the following. That a person becomes delinquent because of an excess of
definitions favorable to violations of law over definitions unfavorable to
violations of law and through interactions with intimate personal contact
(Sutherland and Cressey, 1978).
Moreover, Sutherland and Cressey point to the ratio of pro-deviance to anti-
deviance definitions (what is acceptable or not acceptable behavior determined by
their influence groups or individuals), and not just to the level of pro-deviance
definitions and the proximate cause of deviance (Sutherland and Cressey, 1978).
Sutherlands path, modeled by Marcos et al. (1986), is outlined in Marcos et al.
(1986)s article.
Let me explain how differential Association /bonding theory (DA) agrees with
some aspects of social control, while disagreeing in other as to how delinquency

forms. Both theories agree that bonding occurs through relations with community,
education, parents and peers. Differential association disagrees with social control
regarding peers, parents and religion. Where social control argues that parents .are the
primary influence for deterring delinquency, DA argues that peers, hot parents are the
major influence for preventing or encouraging delinquency. Regarding religion,
social control theory never addresses the influence of religion, whereas DA believes
that religion plays a salient role in deterring youth from delinquent behavior. For
Differential association, a central theme is
That deviant or criminal behavior (such as teenage drug use) is
primarily influenced by the association one has with definitions or
behavior patters that either promote (reinforce) or proscribe
(punish) such behavior. An important corollary is that those
definitions or behavior patterns are primarily derived from ones
close personal acquaintances.
(Sutherland and Cressey, 1978)
Logically, youth who are involved with religion (at least in the past) would
tend to pick peers who were more likely to be anti-drug that youth that were not
involved with religion.
In sum, the logic of this perspective is that those who use drugs will likely
have drug-using friends and those who do not use drugs will have friends that do not.
According to the research, youth pick up their friends definitions (of right or wrong

regarding using or not using drugs, stealing, vandalism etc.) or values conducive to
drug use (e.g. my friends or peers or parents do drugs so it must be ok) (Matsueda
1982; Jacquith 1981). And Jeanne Jenkins found that, consistent with prior research,
the strongest correlate of gateway an hard drug use across all grade levels was
affiliation with drug-using friends (Jenkins, 1996).
Since social control and differential association are similar yet different
enough that they complement each other in the blend, Marcos et al. (1986) blended
the two theories in their research project. In the words of Marcos et al. (1986) we
Attempt to add to our understanding of the processes by which
adolescents become involved in drugs by delineating a model of
adolescent drug use biased on social control and differential
association theories, and then test that model using four different
categories of drugs alcohol, tobacco, marijuana, and
amphetamines and depressants (Marcos, 1986, p. 136).
In summary, Marcos et al. (1986) believed that social control theory as
designed by Travis Hirschi omitted necessary variables (religion generally) or
misinterpreted salient influence of variables (peer influence specifically) to explain
why youth would begin to use drugs. To correct for Hirschis omissions and
limitations, Marcos et al. (1986) combined social control theory with differential
association theory (religion and peer influence derived higher salience in differential

association theory) to construct a more complete model that would explain why youth
would become involved with drug use.
Empirical Review
Marcos et al. (1986) reports that they used the Monitoring the Future (MTF)
dataset 1984 Questionnaire Responses from the Nations High School Seniors.
Jerald G. Bachman, Lloyd d. Johnston, and Patrick M. OMalley, from the University
of Michigan Institute of Behavioral Science, collected and maintained the MTF
database. Marcos et al. (1986) reports that the data was collected in the
southwestern United States using a self-report questionnaire to students in five high
The authors report that there were 2703 students reporting with 53% female
and 47% male respondents. These students were a random sample of required
classes from within each grade (ninth, tenth, eleventh & twelfth) representing 28%,
27%, 24%, and 21% respectively (later the sample is identified as high school

Regarding race, there were 82% reporting as white, 13% Hispanic, and 5%
from other minority groups. Regarding rural and urban population, the authors report
that 84% were form a metropolitan area and 14% represented a rural population.
To test for reliability, they conducted a test-retest collected from 149 juniors
and senior high school students in four schools not included in the sample. Ninety-
four percent of the responses at Time 2 were identical to those at Time 1, thus
showing reporting consistency. Also, the authors report the rate of nonresponse to
drug use questions were less than one-half of one percent and were similar to non-
response rates on other types of questions indicating self-reporting drug use as
The following are operational constructs that Marcos et al. (1986) used to
define the independent and dependent variables for their research. The intentions for
these definitions are to match as much as possible the operational definitions that
Hirschi constructed. However, recall from the discussion on page 6 of this paper, that
Marcos et al. (1986) had to infer the constructs for some of the variables because
Hirschis social control theory lacks both conceptual clarity and specification of
causal ordering... and that the social control influences of parents, as one construct
(Marcos, 1986, p. 137-8).

Marcos et al. (1986) measured drug use by self-reports on the use of four
categories of drugs alcohol, cigarettes, marijuana, and amphetamines and
depressants. Students were asked how often they had ever used each drug and how
often they used that particular drug during the past month. Marcos et al. (1986) only
reported lifetime use findings for each drug. The response categories were scaled
ranging from 0 for never to 4 for ten or more times.
Parental attachment combines the sum of four items tapping affection and
bonding between parent and child. The students were asked whether their family had:
1) any things in which they participated together that were fun, 2) whether they could
talk to their mother about problems, 3) whether they could talk to their father about
problems, and 4) whether in their free time their parents usually knew where they
Educational attachment combines the sum of five elements: feelings toward
school; the importance of good grades; time spent on schoolwork; grades received
and educational expectations.

Religious attachment combines the sum of two elements. Students were asked
how often they attended church and how important religion was in their lives.
Conventional values combines the sum of four elements: the importance of
following rules and obeying the law; the acceptability of stealing under specified
circumstances; and attitudes toward sneaking into a movie or ballgame without
paying. It is important to note that these are generalized conventional values items,
not drug-specific attitudes or definitions.
Systematic Review
In contrast to Hirschis theoretical constructs of Attachment, Belief,
Commitment, and Involvement, Marcos et al. (1986) construct path analysis based
upon their construct definitions previous discussed. They begin with parental
attachment as the primary, exogenous variable, meaning that in the model, there are
no other variables influencing this variable. Educational Attachment, Religious
Attachment, Conventional Values and Drug-Using Friends exist as intervening
Marcos et al. (1986) report conducting a path analysis using a trimmed
regression model. According to SPSS users guide 11.0, the trimmed arithmetic mean
is calculated when the largest 10% and the smallest 10% of the cases have been

eliminated. Eliminating extreme cases from the computation of the mean results is a
better estimate of central tendency, especially when the data are non-normal. (SPSS
Users Guide, 11.0.1).
Marcos et al. (1986) conduct path models for each of the five different
dependent variables (alcohol, tobacco, marijuana, amphetamines and barbiturates)
using the primary exogenous variables, and the intervening variables. The findings
from the Marcos et al. (1986) study are discussed based on the effects of the path
analysis and measured against each of the dependent variables.
Regarding alcohol, drug-using friends had a very strong direct effect (.53),
while religious attachment has a small but significant negative, direct effect (-.15).
They report that the model explains 34% of the variation in lifetime alcohol use.
They also report that when drug-using friends are taken out of the model, the other
four independent variables only explain 16% of the variability.
Regarding cigarettes, drug-using friends showed a strong direct effect (.46),
while educational attachment showed a small, negative, direct effect of (-.13). The
explained variation is 27 percent for all five independent variables and the explained
variation drops to 13% when drug-using friends are taken out of the model.

Regarding marijuana, drug-using friends showed a strong direct effect (.53)
while conventional values (-.11) and educational attachment (-13) have relatively
small direct, negative effects. This model explains 42% of the explained variance,
and drops to 21% when drug-using friends are removed from the model.
In sum, the study found that the predominate influence on adolescent drug
use is having friends who use drugs (Marcos et al. 1986). Parental influence rarely
shows any effect on drug use directly while religious attachment has the most distinct
pattern of effects. Additionally, religious attachment even manages to have a direct
influence on alcohol use that is independent of peer group processes (Marcos, 1986).
The first weakness that I find is regarding generalizability. The authors report
that the data was collected from five high schools in the southwestern region.
Without a nationwide sample that is randomly selected, making a nationwide
inference would be inappropriate. According to Ron Larson (Penn State University at
Erie) and Betsy Farber (Bucks County Community College) the most common abuse
(or misuse) of statistics is using a sample that does not represent the entire population
of the study (Farber and Larson, 2003).

At no time in the study do the authors address the target population for the
findings of this study. Thus, the only population that we can generalize to would be
high school students in the southwestern region.
The second weakness deals with the sample representation. Marcos et al.
(1986) referenced the Questionnaire Responses from the Nations High School
Seniors, but report that they had a sample of 9th, 10th, 11th and 12th graders. There
seems to be a discrepancy about the sample and the dataset that they report using.
The third weakness is found with the control variables that they never
included in the model. It is common practice in sociology to include race and sex as
independent variables in a model and the authors only report the statistics of the
sample regarding race and sex, but never include race or sex in the path analysis
model. It is the goal of my research to remedy this omission by including race and
sex as key variables to see how the regression model is affected globally.
The fourth weakness I find in the Marcos study regard region. Discussed
later, studies suggest that there are significant findings regarding how region plays a
salient role in youth drug choices. Much of the current research does not include
region in the analysis. One of the reasons stems from the lack of substantive region
data being collected.

Just as Joy Dryfoos reports in her book, Adolescents at Risk: Prevalence and
Prevention, that gender has just recently been regarded as a salient variable, Region is
only now being realized for its importance (Dryfoos, 1996). I predict in the future
that databases will include exhaustive raw data collection regarding region for
researches to explore.

Gender, the Chronology
There is a plethora of research regarding gender, and for many years the
research has showed the saliency for adding gender into a research model. As far
back as 1934, researchers argued that boys are more prone to higher delinquency rates
as a result of less supervision than girls (Glueck and Glueck, 1934). And in 1977
Anthony Harris showed that gender is one of the strongest correlates of crime and
delinquency (Harris, 1977).
Research in 1985 showed the same finding that gender is one of the strongest
correlates of crime and delinquency (Hagan, Gillis, and Simpson 1985). In 1985
research also found that because girls are supervised more closely they have stronger
emotional bonds to families, and thus are less free to break the law (Hagan et al.
According to research done by the National Institute of Drug Abuse (NIDA,
1988), while earlier studies of substance abuse pointed toward greater usage among
young males, more recent studies show a decrease in sex differences (Gersick et al.,
1985). Moreover, the same study found female high school seniors are more likely
than males to smoke and use some illicit drugs, such as amphetamines, and they use
alcohol and marijuana at almost the same rates as male seniors (Gersick et al., 1985).

According to Joy Dryfoos in her book, Adolescents at Risk: Prevalence and
Prevention, there is evidence that males are much more involved in heavy drinking
and drunk driving than females (Dryfoos, 1996).
With all of this research regarding gender prior to the Marcos et als (1986)
research project, it is unclear why they would omit gender from their model. So it is
incumbent upon this project to include gender into the model.
The National Household Survey on Drug Abuse (2003) reports that although
males were more likely than females to have used marijuana during the past year
among adults ages 18 or older, there were no gender differences in past year
marijuana use among youths (
I will be using Karen Heimers research as the model for including gender in
my analysis. Heimer was chosen for two reasons. First, her research is timely and
current and second, her research uses the blended models of differential association
and social control, which resemble the Marcos et al. (1986) research model.
Karen Heimer conducted research in 1996 using the National Youth Survey
dataset from 1985. Her research showed that U.S. boys and girls ages 11 to 17 from
1976 differ in the process by which social controls are transformed into self-control in
delinquent situations among girls as compared with boys (Heimer 1996). She also

found that there are gender differences in the role taking process leading to
delinquency suggesting an important difference in the role-taking process by which
group social controls are transformed into self control in delinquency situations
among girls as compared with boys (Heimer, 1996).
In other words, there are clear differences in how boys and girls are influenced
and that these influences lead both girls and boys toward delinquency but that the
paths they take appear to by clearly different.
Although the goal of this research is not to confirm Heimers findings, her
research shows that sex must be included in any social research as a control factor
because girls and boys are consistently found to be different in their paths toward
delinquent behavior.
Moreover, Joanne Belknap from the University of Colorado, Boulder argues
in her book, Invisible Woman: Gender Crime, and Justice,
Whereas social class, access to opportunities to learn
crime, and area of residence in a city have been used to
explain boys likelihood of turning to crime, the causes
of girls criminality have rarely been examined until
recent years. Additionally, criminological theory
historically tended to view women as driven to crime
because of biological influences, whereas men were
viewed as turning to crime because of economic or
sociological forces. ...Thus girls delinquency was

seen as neither interesting nor important until the past
couple of decades.
Lastly, more recent research by Kristan G. Erickson, Robert Crosnoe, and
Sanford Dombusch in an article entitled, A Social Process Model of Adolescent
Deviance, found that gender subgroups indicate that the deviance of males was more
strongly affected by the actions of friends than was deviance of females (2000).
Erickson et al. were researching the social control/differential association
blended theory to determine how susceptibility to negative peers could be mitigated
through strong social bonds. They looked at youth from high schools in California
and Wisconsin from 1987 through 1990. They measured social bonds (parental
attachment, parental supervision, teacher attachment, educational attachment, &
community involvement), peer deviance, susceptibility, and adolescent deviance.
Conducting a structural equation model, the study found that greater social
bonds was significantly associated with reduced susceptibility (p. 410). They also
report that strong parental attachment and a commitment to education decreased the
likelihood of students to have close friends who frequently engaged in delinquent acts
(Erickson et al. 2000)
The work of Heimer, Belknap, Erickson and other feminist/criminology
theorists are bringing sex and gender to the forefront of theoretical modeling.

Because of their work, clearly it is incumbent upon researchers to include sex/gender
as a control variable in any delinquency analysis. This is a benefit of my study
According to the National Household Survey on Drug Abuse (NHSDA)
report, July 19 2002, percentages of youth aged 12 to 17 reporting past year
marijuana use by the following regions. West -14.8%, Northeast -14.4%, Midwest -
13.4% and South -12.1%. The NHSDA is an annual survey sponsored by the
Substance Abuse and Mental Health Services Administration (SAMHSA). The 2000
data are based on information obtained from nearly 72,000 person aged 12 or older,
including more than 25,000 youths aged 12 to 17. The survey collects data by
administering questionnaires to a representative sample of the population through
face-to-face interview at their place of residence (
One of the influences for including region in my study derived from a year
that I spent working with the Center for the Study and Prevention of Violence
(CSPV) in Boulder, CO. under the direction of Dr. Delbert Elliot. Delbert Elliot is
the primary researcher for both the National Youth Survey, and the Denver Youth
Survey which are respected and renowned, perennial databases that track youth
behaviors and attitudes at the national level and for city of Denver respectively.

While working for CSPV, I traveled to 12 states and 750 schools monitoring a
grant study on behalf of the Department of Justice Office of Juvenile Justice and
Delinquency Prevention. It became abundantly clear that in each region, youth had
different drugs available to them. Administrators and teachers would informally
report the drugs their students had access to and were using.
Based on a summary from a 1986 Monitoring the Future National Survey of
High School Seniors collected and reported by L. Johnston, P. OMalley, and J.
Bachman, Region showed a difference with the South using less drugs. This team
published an article entitled, National Tends in Drug-use and Related Factors among
American High School Students and Young Adults, 1975 1986.
The Johnston et al. (1985) report; suggests that substance abuse, problems,
defined by heavy use, are experienced more by males, whites, and youth without
college plans. Continuing, these problems are more common on the East and West
Coasts for illicit drug-use and less common in the South for all substance abuse
(Johnston et al., 1985).
Deduction would lead us to believe that illicit drug-use may be less common
in the South because of religious influence as well as region. Based upon this
information (or lack of information) regarding region differences for drug-use, region

will be included to determine if there are significant drug-use differences among
Because so little research has been conducted regarding region, I explore the
variable, Region, using a crosstab to determine how the data counts appeared when
compared to the dependent variable, marijuana use. The results of the crosstab can be
found in appendix A, Figure 2a. The results suggest that most of the respondents are
reporting that they are not engaging in marijuana use (51.8%) and that the South
(when not using marijuana) reports less marijuana than any other region. However,
when we examine the responses for yes to drug use, the south seems to report the
highest number of responses South, 1-5 times (5.2%), 6-9 times (2.0%), 10-19 times
However, when looking at the most frequent use of marijuana (20 40+ times)
the results suggest that the NE and the S are approximately equal in response. One of
the most significant suggestions is that youth are either not engaging in drug use
much or at all (66.8% of the Total), or that they are consistently using marijuana
(22.3% of the Total). Based on this information I decided to conduct a more
exploratory analysis called Analysis of Variance. So the claim by Johnston et al. may
lack full merit.

I also conducted a One-way Analysis of Variance (ANOVA) test to determine
if Region showed significance when tested against the dependent variable, lifetime
marijuana use (discussed below). An ANOVA is a hypothesis-testing technique that
is used to compare means from three or more populations (Larson, Faber, 2003). The
null hypothesis that I tested is as follows there would be no significant difference
among regions when tested against the dependent variable, lifetime marijuana use.
The results of the test can be found in appendix A, Figures 2b, 2c, 2d.
Because the ANOVA showed a significant main effect for Region (F = 9.30, p
< .05) I conducted a pairwise multiple comparison post-hoc test. According to the
SPSS Guide 11.0
Once you have determined that differences exist among the means,
post hoc range tests and pairwise multiple comparisons can
determine which means differ. Range tests identify homogeneous
subsets of means that are not different from each other. Pairwise
multiple comparisons test the difference between each pair of
means, and yield a matrix where asterisks indicate significantly
different group means at an alpha level of 0.05.
In the post hoc test for multiple comparisons, for NE there was a significant
difference from the South. For North Central and West, there was a significant
difference. For the West there was a significant difference from the North Central
and South (the mean differences tested were significant at the .05 level).

Based on these preliminary findings, Region appears to be a viable candidate
for the social control model. Once the multiple regression is complete, if region
shows a significant effect, we will be able to determine the strength and direction
from the correlation coefficient (R). We will also be able to determine how much
marijuana use is explained by region through the coefficient of determination (R2).
Joy Dryfoos reports the race issue is complicated by the under-
representation of black inner-city youth in the two national surveys (Dryfoos, 1996).
Additionally, Dryfoos reports that according to the book, Young Black Males and
Substance Abuse, by A. Brunswick, aggregate black rates are weighted by rural
southern blacks, who are much less likely than other black youth to experiment with
drugs (Dryfoos, 1996). Inherently, this will cause spurious results to be reported and
lead to conclusions that are not necessarily reliable.
Additionally, the National Institute on Drug Abuse, 2002 reported that it is
white females and males that have the highest current prevalence rates for all
substances except cocaine: and that Black and Hispanic females consistently show the
lowest substance abuse rates.

Clearly race has contentious debate surrounding it, thus it makes it important
to include race in the model as a control factor. This was a major shortcoming for the
Marcos et al. (1986) study, but subsequent social control research has incLudedrace in
the model (Johnston et al, 1985, Erickson, 2000). Likewise, my model will include
race as a control factor to expand the Marcos et al. (1968) study.

A Multiple Regression
Model of Social Control/Bonding
and Youth Drug-Use
This project investigates Marcos et al. (1986) Social control/bonding theory,
by constructing a multiple regression model to analyze two substantive issues: (1) the
models ability to explain delinquency when sex, region and race are added to the
model; and (2) the models ability to explain the influence of region on delinquency.
The Monitoring the Future: A Continuing Study of American Youth (12th-
Grade Survey) data began being collected in 1984. It was funded by grants from the
National Institutes on Drug Abuse (NIDA, 1988), a part of the National Institutes of
Health (NIH). The University of Michigan Survey Research Center carried out the
survey under the supervision of Jerald G. Bachman, Lloyd D, Johnston, and Patrick
M. OMalley (
The data being use for my research also comes from the MTF collected in
2000. Following Marcos et al (1986), the respondents are 12th grade students who

initially participate in school by completing a self-administered, machine-readable
questionnaire in the normal classrooms, administered by University personnel
following standardized procedures detailed in a project instruction manual
The questions in the MTF are designed around self-report which brings
problems regarding exaggeration or minimization regarding drug involvement
(McCord, 1990). One option is to review official arrest records and adjudication
records, but there are other issues related to arrest records because the processing of
youth offenders is often tainted by racial and class bias (Farrington and Tarling, 1985;
McCord, 1990; West, 1982). Because of these disadvantages, the preferred source of
information for deviance research and the most commonly employed is self-report
(Hindelang et al., 1981)
Student respondents are randomly selected from both public and private high
schools nationwide. The students are randomly assigned to complete one of six
questionnaires, each with a different subset of topical question but all containing a set
of core questions on demographics and drug-use. There are about 1400 variables
across the questionnaires. Drugs covered by this survey include among other drugs:
tobacco, alcohol, marijuana, amphetamines (stimulates), barbiturates (tranquilizers).

Other items include attitudes toward religion, parental influences, and educational
aspirations (
The follow-up questionnaires are mailed to respondents with a return, self-
addressed, stamped envelope and a small monetary gift of ten dollars from the
University of Michigan as a token of appreciation (
Based on the variables described above, and after extensive searching; the
model will be constructed from both the core data and form 1 (out of 6 forms)
because these data are particularly well suited to test social control/bonding theory.
There were approximately 420 public and private high schools and middle
schools selected to provide an accurate representative cross-section of students
throughout the coterminous United States at each grade level
(monitoringthefuture. org).
The sample was constructed from the core data and form 1 (discussed above)
which consisted of 2242 12th grade students. According to Larson and Farber, 2003,
determining minimum sample for population proportions can be described with the
f v

Where p-hat and q-hat are the population proportion and its complement
respectively. Because I dont have a preliminary estimate for p, I am required to use
the most conservative estimate for p and q, .5 and .5 respectively meaning that I must
pay the penalty by using a largest sample possible (Larson Farber, p.297).
Zcritical is a standard score determined by alpha .05 two-tailed (Zcrit = 1.96). E is
the maximum error of estimate allowed for accuracy. I chose to be accurate within -
3.5% (.035). The minimum sample required to be accurate within 3.5% of the
population is 784 respondents. With a sample size of 1015,1 am well within the
minimum sample size to estimate the population proportion.
Missing data was removed from the SPSS statistics package (pre-composite)
by recording all the variables (dependent and independent) into same variable and
recoding the response -9 (identified as missing by dataset curators) into system
missing variable.
Next, scales were created to match the Marcos et al. (1986) model to test
social control theory (discussed below) and when complete; there were 1015
Regarding race, there were 1293 (57.5%) whites and 321 (14.3%) blacks.
Regarding gender, girls represented 999 (44.6%), and boys represented 968 (43.2%)

of the sample. Regarding region, in the Northeast (NE) there were 517 (23.1%), in
the North-Central (NC) there were565 (25.2%), in the South (S) 786 (35.1%), and in
the West there were 374 (16.7%) (sample descriptives are available in appendix A)
I constructed the variables based upon Marcos et al. (1986)s constructs of
Parental Attachment, Religious Attachment, Educational Attachment, Drug-using
friends, and Conventional Values.
To maintain the studies integrity, not only am I using the MTF database, as
did Marcos et al. (1986). but I will also be constructing this model using confirmatory
factor analysis with a varimax rotation, and forming composite independent variables,
all modeling Marcos et al. (1986).
Independent Variable
Factor analysis was used to confirm my construct for the independent
variables. The varimax rotated factor loadings revealed that the items measuring the
constructs into five separate components. Varimax rotation is an orthogonal rotation
method that minimizes the number of variables that have high loadings of each factor.
It simplifies the interpretation of the factors

These components became the variables for the multiple regression analysis
(see appendix C).
All measures of the independent variables were originally constructed in two
ways. First, the raw items in each construct were summed to form a composite score.
Second, using factor analysis, each item was standardized and weighted by its rotated
factor score, and then all items measuring a given construct were summed to form the
scale (Marcos et al. (1986). For the original variables, coding, and additive scores,
(see appendix B).
According to the SPSS 11.0 Users Guide,
Factor analysis attempts to identify underlying variables, or factors,
that explain the pattern of correlation within a set of observed
variables. Factor analysis is often used in data reduction to identify a
small number of factors that explain most of the variance observed in a
much larger number of manifest variables.
Once the four components were extracted, I examined the underlying
variables to determine how the components fit the Marcos et al. (1986) model.
Explanations for each component follow (labels are in parenthesis).
Component 1 resembled peer attachment (drugpeer), component 2 resembled
religious attachment (religatt), component 3 resembled parental attachment (pamtatt),

component 4 resembled conventional attachment (convatt). Component 5 resembled
educational attachment (eductatt).
For component 4, it is important to be married and have a family (impmarfa)
loaded lower than (.6) in the rotation. However, for theoretical considerations, I
chose to include the variable in the analysis. The obtained scores is not that
dissimilar (.528).
The independent variable drugpeer consisted of three variables with a
reliability score (alpha = .80). The independent variable religatt consisted of two
variables with a reliability score (alpha = .76). The independent variable parental
attachment consisted of two variables with a reliability score (alpha = .64). The
independent variable conventional attachment consisted of three variables with a
reliability score (alpha = .67). The independent variable educational attachment
consisted of two variables with a reliability score (alpha = .73)
There are also three independent variables used as control variables included
in the model and they are sex, race and region (see model discussion on page 38 for
more information).
In sum, there are eight variables used to construct the model:
1. Drug using peers
2. Parental attachment
3. Educational attachment

4. Conventional attachment
5. Religious attachment
6. Race
7. Region
8. Sex
Dependent Variable
For the purpose of this study, the dependent variables, lifetime marijuana use
(lftmmj) was chosen to test the model. Lifetime marijuana use was measured on a
response scale of 1-6 (1 = never used marijuana, 7 = 40+ times used). I recoded the
dependent variable into 5 response categories called mjlftm. I combine response
categories 2 & 3 = 2 and 6 & 7 = 5.
Each of the other responses remained the same
1=0 times;
2 = 1-5 times;
3-6-9 times;
4 = 10 19 times;
5 = 20 40+ times.
By collapsing the data there are fewer response categories and so the number
of responses increases, thus giving us greater variation within the matrix.

To recap the variables for the model there are eight independent variables,
five of which are composite independent variables. The composite variables are:
drug-using peers (drugpeer), religious attachment (religatt), parental attachment
(pamtatt), educational attachment (eductatt) and conventional attachment (convatt)
For the dependent variable, the recoded lifetime marijuana variable will be
The next step is to define and discuss the hypothesis tests used to test the
social control theory, including the new variables.

Using the five theoretical variables and the three key variables (sex, race, and
region), I will be conducting multiple regression tests to explore the effects of region
on drug use. Region has been turned into a dummy variable for all four regions in the
model (Northeast, North Central, West and South). Three hypotheses will be tested
using regression analyses to test the drug-use variable, lifetime marijuana use
(lftmmj). The hypothesis discussions follow.
. Hypothesis 1
Regarding region, I claim that there will be a significant difference among
youth drug use when considering the dependent variable lifetime marijuana use.
Based on the NIDA findings, that the South reports less marijuana use, we should
find that marijuana use should be less in the South than in the Northeast, North
Central and West.

Hypothesis 2
Regarding race, I claim that there will be a significant difference among youth
drug use when considering the dependent variable marijuana use. We should find
that whites should report more use when considering the drug marijuana. This
follows the findings of the NIDA findings in 2002 that whites were more likely to use
drugs than Blacks and Hispanics.
Hypothesis 3
Regarding sex, I claim that there will be no significant difference among
youth when considering the dependent variable marijuana use. According to the
research discussed above (Heimer, Belknap, and Erickson), girls use drugs differently
than boys. However, the study by the NIDA (1988) suggests that contrary to previous
research, there are no differences between boys and girls in drug use. So, I claim that
there will be no significant difference between boys and girls

Using the eight independent variables, I am constructing a model that will be
called, A Multiple Regression Model of Social Control/Bonding and Youth Drug Use,
that will be used to test the hypotheses just discussed.
Multiple regression is a better prediction model for a dependent variable
(response) variable by using more than one independent (explanatory) variable
(Larson & Farber, 2003). The method of multiple regression that I ran was called
stepwise multiple regression. A stepwise method examines the variables in the
block at each step for entry or removal. This is a forward stepwise procedure (SPSS
Users Guide, 2001). A forward procedure submits each variable one step at a time to
determine how well the variable correlates. Each variable is submitted at a .05 level
of significance and removes at a .10 level of significance. The level of significance is
more conservative (on type I error) to enter the model, and more liberal (on type I
error) to be removed from the model.
The regression model appears where y = the dependent variable, B = the y
intercept, m; = the slope for each independent variable, x; = the independent variable,
and e = the unexplained error. The model follows:

y B + (miXi(race) ITI2X2 (sex) IH3X3 (region) III4X4 (pamtatt) HI5X5 (eductatt) 1 H16X6 (peerdrug)
+ m7x7 (convatt)+ mgXg (reiigatt)) + £ (unexplained error). (A correlation matrix can be seen in
Appendix A, Table 3a).
The overall correlation model shows a moderate, positive relationship among
the variables, (R = .530) and an overall weak explanation for marijuana use with a
correlation of determination (R2 = .281) (Appendix A, Table 3b)
For the model, the ANOVA reports an overall significant correlation (F = 49.158, p
< .05).
The correlation matrix shows the following independent variables have a
significant correlation with marijuana use. Drug-using peers show the strongest,
positive relationship (.454); conventional attachment has a weak, negative
relationship (-.118); a weak, negative relationship with educational attachment; a
weak, negative relationship with religious attachment (-.233); a weak, negative
relationship with sex (-.099); and a weak, negative relationship with parental
attachment (-.042). A discussion of the regression model follows.
This model
y ~ B + (miXi(race) m2X2 (sex) ^03^3 (region) ITI4X4 (pamtatt) ITI5X5 (eductatt) "t" mgXg (peerdrug)
+ m7X7 (convatt) m8X8 (reiigatt)) £ (unexplained error).

will be used in a multiple regression model to determine how much explained
variance in drug use occurs when tested with the social control/bonding theory.
Additionally, this model will be used to conduct an analysis that focuses on
dependent variable, mjlftm.
Although this model will be testing the entire social control/bonding theory,
region is the critical variable because little research exists in social control that
addresses the affect of region.
Theoretically, youth who score high on the bonding variables (other than
drug-using peers) should have a lower occurrence of drug use in both analyses.
Additionally, drug choices should be different based on the region when considering
the drugs tobacco, amphetamines and barbiturates.
According to social control/bonding theory, drug-using peers consistently
show a significant effect on youth using drugs. We should find the same outcome in
this model. Meaning that if youth report having more drug-using friends, they should
also respond with higher marijuana use.
We should also find that religious attachment should influence youth behavior
by reducing drug use. That is to say, the more they report that religion is important,
the less they should be using drugs.

High parental attachment should also show less drug use among youth; we
should also find that high conventional attachment should show less drug use among
I conducted a multiple regression testing all eight of the independent variables
against the dependent variable, lifetime marijuana use. The variable region was
recoded with the four dummy variables representing the NC, NE, W and S. A
discussion for how the dummy variables were created follows.
To test the region hypotheses, I created a dummy variable for region so that
we could isolate different parts of the country. There are four dummy variables to
represent Northeast (regdumne =1), North Central (regdumnc = 2), West (regdumw
= 3 South, and (regdums = 4) (variable labels/values are in parentheses). Once the
dummy variables were created, they were included in the regression model and
measured against both of the dependent variable, lifetime marijuana use.
Table 4.2 reports the regression coefficients obtained during the multiple
regression (R). The following discussion explains the results of the test. The beta
coefficients determine the strength and direction of the variable. That is to say that a
high score reports a stronger relationship for the model. Likewise, a negative sign on

the beta coefficient determines that the relationship is a negative relationship,
meaning that as one variable score (independent variable) increases, the other
(dependent variable) score decreases.
Table 4.2 also represents the model tested against the dependent variable
drugtab. The constant is the dependent variable, lifetime marijuana use. The
Standard error is
a measure of how much the value of a test statistic varies from
sample to sample. It is the standard deviation of the sampling
distribution for a statistic. For example, the standard error of the
mean is the standard deviation of the sample means.
(SPSS Users Guide, 2001)
The Beta is
a standardized regression coefficient, when all variables are
expressed in standardized (z-score) form. Transforming the
independent variables to standardized form makes the coefficients
more comparable since they are all in the same units of measure
(SPSS Users Guide, 2001).
Larson & Farber (2003) explain t as an obtained standard score that can
be used to test whether the correlation between two variables is significant. The
test statistic is r and standardized test statistic is
t = R\standard error (where R = correlation coefficient for the multiple

The correlation coefficient is a measure of the strength and the direction
of a linear relationship between two variables (Larson & Farber, 2003). The
symbol R represents the sample correlation coefficient. The formula for R is
R = n the sum of x*y (sum of x)*(sum of y)\square root of n*the sum of x
squared (sum of x) squared the square root of n* the sum of y squared (the
sum of y) squared (Larson & Farber, 2003).
In Table 4.2, the bold significant scores represent those variables that were
found to be significant when tested against the dependent variable. The significance
is the probability of a Type I error. Type I error occurs if the mill hypothesis is
rejected when it is actually true (Larson & Farber, 2003).
The null hypothesis (Ho), is a statistical hypothesis that obtains a statement of
equality, such as less than or equal to, equal to, or greater than or equal to (Larson &
Farber, 2003).

Hypothesis 1 stated that region would show a significant difference regarding
the South, meaning that the South would use less marijuana.
These results show that the dummy variables for North Central and
Northeast regions are not significant in their correlation. The dummy variables West
and South were both removed from the analysis. According to the warning that SPSS
generated during the analysis, West was removed from the analysis because it is
either a constant or because of missing correlations. When I recoded region into the
dummy variables, West was the last of the dummies created; therefore it is the
constant dummy variable meaning that recoding the raw data was unnecessary.
I attempted to correct this problem by recoding the raw data for the West from
the constant to a 1 (which was successful), but the same warning generated.
Regarding the South region, it was excluded for collinearity considerations
(tolerance = .000) in both Analyses. After a review of the raw data, all of the
response categories were coded with 1, with none of the dummy variables having
matches. I am uncertain about any reasons why this collinearity exclusion existed.
West is the constant not South. None of the cells in the response cells matches the

other region cells. It may be that there was some error made during the data
collection that is beyond my understanding.
I will now continue with the discussion about the findings explaining the
relationship with the literature and the hypothesis tests.
Hypothesis 2 stated that there would be a significant difference among races
when considering the dependent variable, lifetime marijuana use. I claimed that
whites (labeled 0) would report more marijuana use than their Black (labeled 1)
counterparts. For Hypothesis 2, the analysis suggests that race does show a
significant difference between Blacks and whites regarding lifetime marijuana use
The theory suggested that race would show a significant difference among the
races. The statistics obtained (t = 2.377, p < .018) in the analysis suggests that race
does have a weak positive, significant effect (Beta = .069) for why youth become
involved with marijuana use.
This research suggests that Blacks are more likely to engage in marijuana use
than their white counterparts. This finding is contrary to my claim, because the
research suggested that whites are more likely to use marijuana. My claim followed
the research. In this study, Blacks are more likely than whites to report using drugs.

One caution should be mentioned regarding race. The researchers from MTF
collected data as a dichotomous variable, white and Black. By not including other
races in the data, full interpretation is limited to the existing variables, white and
Hypothesis 3 states that there would be no significant difference among the
sexes. According to the literature reviews, the findings were mixed. The NIDA
(1988) suggested that there are no significant differences between the sexes, but this
analysis suggests that there is a significant difference (t = -2.036, p < .05). The
statistics suggest that there is a weak, negative significant effect (Beta = -.063)
meaning that boys (labeled 1) are more likely to engage in marijuana use than girls
(labeled 2).
Theoretical Variables
For the five social control, theoretical variables, only drug-using peers,
conventional attachment and religious attachment showed significant effect on
marijuana use. It is interesting that parental attachment and educational attachment
do not influence youth marijuana use.

The data suggest that drug-using peers appear to influence youth marijuana
use (t = 16.998, p< .05). The model suggests that there is a moderate, positive
significant effect. This means that as youth report more drug-using peers, the more
likely they are to engage in marijuana use.
Regarding conventional attachment and religious attachment, the data suggest
that there is a significant effect (t = -1.987, p < .05) (t = -5.231, p < .05) respectively.
For both variables the data suggest that there is a weak, negative effect (Beta = -.057)
(Beta = -.153) respectively. Both of these finding suggest that as youth report
stronger relationships with conventional attachments (leadership, marriage and
family, and having a purpose in life) there is less marijuana use reported by these
youth. Likewise, as religious attachment increases (attendance and importance) there
is less marijuana use reported by these youth.
Even though the first analysis showed no significant effect for region, I also
conducted a multiple regression analysis with region as a 4-category dummy variable.
Because the data in this model are the same, the previous analysis explanation
remains the same. In this analysis I will focus exclusively on the region variables:
regdumnc, regdumnc, regdumw, and regdums. The results of the correlation
coefficients are contained in Table 4.2

This study explored and attempted to expand the social control/bonding theory
of Marcos et al (1986). Marcos and his team conducted an analysis on five
theoretical variables derived and blended from Hirschis social control theory and
Sutherland and Cresseys differential association theory. Social control/bonding
theory argues that the higher a youth reports attachment to parents, religion,
education, and convention, the lower the drug use by the same youth.
Likewise, for youth that report having drug-using friends, the more likely the
these youth report engaging in drug use themselves.
Marcos et al. conducted their study by measuring youth drug use in five
separate analyses of alcohol, tobacco, marijuana, amphetamines and barbiturates.
They found that a higher level of the theoretical variables showed lower reports of
drug use by youth. Drug-using peers were found to increase the reports of drug use
by responding youth, corroborated by this research.
There were several additions/changes made to my model in contrast to the
Marcos et al. model. The changes were intended to expand their research.

First, I included three additional variables (race, sex, and region) that I believe
should have been included in the Marcos et al. model. By including these variables, I
expected a clearer explanation for why youth engage in drug use.
Second, in contrast to the Marcos et al. model, I used one dependent variable:
lifetime marijuana use.
Drug-Using Peers
The analysis for marijuana use supported social control/bonding theory.
However there were a few significant findings. First, regarding marijuana, drug-
using peers reported a significant influence on youth drug use, corroborating previous
social control research including Hirschi and Marco et al. (1986)
The analysis suggests that drug-using peers do negatively influence their peers
when considering marijuana. Softer drug use may be influenced by peers, more
because these drugs are considered the gateway drugs, which may suggest that
peers would hold a more salient role because they introduce their friends to the drugs.
Marcos et al. (1986) found that drug-using friends showed a strong direct
effect (.53); likewise this study made found a similar finding (.46).

Educational Attachment
In the analyses, educational attachment did not maintain a salient role. For
youth who have aspirations for future education, the data suggests that youth
marijuana use is not significantly influenced. Marcos et al. (1986) found a weak
negative effect (.13), so the no significant finding is not out of realm for what we
might expect for seniors and marijuana smoking.
This no significant finding regarding marijuana use might stem from youths
perceptions of college students using these drugs or a drug use may offer a feeling of
adult behavior and may influence their choice. This question of why seniors engage
in more drugs would be a good topic for future research.
Parental Attachment
Analysis of parental influence showed no significant results. For marijuana,
parents influence tends not to reduce reports of marijuana use. One reason may be
that the respondents are senior high school students. Therefore, because of their age,
they may feel older or may classify themselves as adults so that parental influence
may not be as strong for these older students compared to their younger peers.

The literature suggests that parental influence is equivocal as stated by
Aseltine (1995). His research found a weak relationship between parental influence
and delinquency, so the no significant findings of this research suggest that the
findings from Hirschi and Marcos et al. (1986) may not be accurate. Further research
with a multiple item, composite variable would be necessary before any conclusive
decisions can be made.
Additionally, future research with separate analyses including additional age
groups (freshmen, sophomores, and juniors) would offer a clearer understanding
regarding parental influence on youth.
Conventional Attachments
The analyses suggest that conventional attachment does have a significant
influence on youth marijuana use. These findings imply that as youth have higher
beliefs about marriage and family, leadership and purpose in life, the more likely they
are to reduce/refrain from marijuana use. These findings support the research that
both Hirschi and Marcos conducted.
Conventional attachment can mean so many things, from desires, goals,
ambitions, etc,, about the community. Community can be large or small, from the
family and neighborhood, to the national conventions of paying taxes, and national

organizations. To say what is conventional becomes a subjective matter. To fully
explore what the convention means can be exhaustive ad nauseum.
In sum, what may be conventional years ago may not be today. And
convention today may not hold in the future. Because of this phenomenon, social
control theory will play a part in social theory for years to come.
Religious Attachment
The analyses suggest that religious attachment has a significant effect in
reducing marijuana use. These findings support the findings of Marcos et al. (1986)
when they argue that religious attachment should have been included in Hirschis
construct of social control. These findings also corroborate Marcos et al.s findings
regarding religious attachment decreasing reports of marijuana use.
To make any conclusive remarks regarding religious attachment will be
contentious. In the Welsh study (1983), religion showed a significant effect where
secular controls are absent or weak. This leads one to ask what the secular controls
were for each of these students. It also leads one to ask whether the findings of the
Welsh study were correct. It is difficult to say if secularism has increased or
decreased since 1983 as compared with religious influence. The arguments become
contentious, so to draw conclusive decisions is not advised.

Again, there is support for Marcos et als assertions that religious controls
should have been included in Hirschis constructs. Last, religious attachment has
consistently included two variables (importance, attendance). But there are many
other questions that could be asked: for instance, which religion a person follows or
whether they believe in a certain sect within each religion. Additionally, questions
regarding ethics and morals could be assessed to determine how a person would
behave in certain situations.
So, to say that religious attachment makes a significant difference is tenuous
because there are so many more variables that should be included.
Marcos et al. (1986) never included race in their research. This research
suggests that there is a significant difference between Black and White students,
where Blacks report using marijuana more than whites. Although the findings are
significant, the Beta (-.069) shows a weak relationship. There are two factors to
consider with these results. First, because the findings are weak, it would be
imprudent to make any ultimate conclusions regarding the effect of race. Second,
because race was dichotomous, external application to other races is not possible and

thus a very limiting effect for this study. Future research would benefit from multiple
categories for race.
The analysis suggests a significant effect regarding sex, where boys are more
likely than girls regarding reporting marijuana use.
Sex implies boys and girls, but research in other theoretical disciplines also
includes gender as a variable. Sex is the difference that we are bom with (ascribed),
gender is what we learn through the socialization process (achieved). The importance
of this distinction cannot be understated. Sex as a variable serves the purpose of a
control variable, and although this study reports differences between boys and girls, if
gender were to be measured and tested, gender may show different findings.
For future studies in social control, it would be important to include not only
sex, but also gender in the exploration of drug use and youth to achieve a clearer
understanding of youth drug use.

Three dummy variables for region were created called Northeast, North
Central and South. In the analysis, South was removed from the analysis due to
collinearity considerations.
The analysis showed mixed results for region. When considering the
crosstabulation and ANOVA, there is a significant difference among the regions, with
the South reporting less marijuana use by youth. However, when considering the
multiple regression, there was no significant effect. There may be some problem with
the raw data and that is why the South was removed for collinearity issues. When
considering the removal of West from the analysis, it was removed from the analysis
because it was the constant. I attempted to make a change regarding the data by
recoding region West into a new code, but to no avail.
The results from this study give some corroborating evidence for the
research, but to make any definitive statements would be disadvantageous.
One limitation was the number of classifications for region. The number was
too low to accurately assess how different (if any) each region is from another. As
time passes, databases will begin to include many more classifications for region so
that differences may be explored.

I continue to believe that region remains a salient variable for investigation.
Of course it should be said that region in and of itself is construed as a
causal/determinate variable. Rather region is a corollary for outside influences. Some
of the influences that may lead to drug availability to youth may be access to shipping
ports, military posts etc. Likewise, region in future studies can be broken into two
more distinct areas, urban and rural. I believe that the rural/urban factor would lead
use into production or availability of certain drugs.
Overall the model showed a low level of explanation for why youth become
involved with drug use (R = .28), meaning that 28% of the variation in marijuana use
can be explained by the eight variable model. However, the findings regarding region
suggest that further research is necessary to understand the influence of region on
drug use.
The intention for this study was to repair the omissions of the Marcos et al.
(1986) study by including race, sex, and region in the analysis.
Explanations regarding race are contrary to my hypothesis and the literature
review. Also, because of the original data configuration, race was collected as a
dichotomous variable; therefore, explanations are limited. The data collection was

limited in its classifications for race and due to this limitation, any subsequent
discussion became limited regarding race. However, the inclusion of this variable as
a control variable suggests that there are differences regarding race. This
corroborates the discussion by NIDA, Dryfoos, Brunswick and Heimer regarding
race: that there is a significant difference among the races. Future research regarding
drug use will require that race be classified with more variation.
Research by Dryfoos, Heimer, and Erickson et al. showed the importance of
sex in model construction because there are clear differences between boys and girls.
According to the research, the differences lie in the paths to drug use, not necessarily
in the drugs they use. Nevertheless, sex was included in the analysis and showed no
significant difference with a marijuana drug use.
Region was the critical variable for this study. The results showed mixed
results regarding region. Regarding marijuana use, the South was significantly
different from the North Central region in the ANOVA and crosstabulation. Again,
the limited classifications make deeper interpretations impossible. There were only
four levels of region classification, and South was removed due to collinearity issues.
However, preliminarily the significant finding with the ANOVA and
crosstabulation allows us the first look into region as a salient variable for social

control theory and may partially explain why youth become involved with drugs.
This is important when we consider drug intervention. In the United States, we have
a system of universal application for drug intervention with youth. We have come to
know it as D.A.R.E. (Drug Abuse Resistance Education), which in its own right has
had several shortcomings which are beyond the scope of this paper.
Anyway, by attempting universal application for youth drug intervention, we-
miss the importance of regional (community, city, county, and even state) differences,
which a significant finding in region may suggest. Further research with a database
including more than three or four classifications for region would allow us to explore
the region affect more thoroughly. If future region studies suggest that there is a
significant difference, then a new approach to youth drug interdiction would be
By having at least one region variable showing significant results, we have a
window of opportunity to explore the importance of region as an influence for why
youth become involved with drug use. Last, future research could also explore
whether region plays a salient role with a composite dependent variable (combining
multiple drug-use categories into a scale). It may be that composite measures of drug
use and region show even more significant findings.

One thing is certain, the findings regarding sex and race when exploring youth
drug use while using demonstrates that explanations regarding youth use of drugs
may not be as simple as first imagined.
With these findings, social control theory could set itself apart from other
social theories if more creative research constructs were developed and tested.

Table A-l
Descriptive Statistics
N Mean Std. Deviation
Statistic Statistic Std. Error Statistic
WILL DO 2YR CLG 1785 2.07 .03 1.138
WILL DO 4YR CLG 1849 3.31 .02 .995
WILL DO GRD/PRF 1804 2.60 .02 1.025
ALL FRD DRINK ALCOHOL 1982 3.61 .03 1.147
ALL FRD SMOKE CIG 1997 2.73 .02 1.016
ALL FRD SMOKE MARIJUANA 1997 2.60 .02 1.087
FATHER IN THE HOUSE 1988 .73 .01 .443
MOTHER IN THE HOUSE 1988 .90 .01 .303
IMPORT TO BE A LEADER IN THE 2208 2.34 .02 .978
MARRIAGE AND FAMILY IS 2208 3.63 .02 .761
IMPORTANT TO HAVE A PURPOSE 2211 3.41 .02 .850
LIFETIME MARIJUANA USE 2131 2.87 .05 2.359
MOTHERS EDUCATION LEVEL 1911 4.01 .03 1.284
FATHERS EDUCATION LEVEL 1818 3.99 .03 1.355
RACE 1614 .20 .01 .399
HOW OFTEN DO YOU ATTEND 1566 5.5945 .0480 1.89863
SEX Valid N (listwise) 1967 1016 1.51 .01 .500
The total, 13286, are obtained when using all 6 forms and the core data. The valid
cases, 714, were obtained when using form 1, the core data, & after removing the
missing data.

Table A-2
LFTMMJ Marijuana use in your life REGION-4 Categories
LFTMMJ 1 no times Count NE:(1) 248 NC:(2) 276 S'.(3) 420 W:(4) 159 1103
Marij use Expected Count 254.1 277.4 383.5 187.9 1103
Region % within 50.5% 51.5% 56.7% 43.8% 51.8%
Total % 11.6% 13.0% 19.7% 7.5% 51.8%
2 1-5 times Count 64 93 111 51 319
Expected Count 73.5 80.2 110.9 54.3 319
Region % within 13.0% 17.4% 15.0% 14.0% 15.0%
Total % 3.0% 4.4% 5.2% 2.4% 15.0%
3 6-9 times Count 14 28 42 23 107
Expected Count 24.7 26.9 37.2 18.2 107
Region % within 2.9% 5.2% 5.7% 6.3% 5.0%
Total %of .7% 1.3% 2.0% 1.1% 5.0%
4 10-19 Count 33 26 39 29 127
times Expected Count 29.3 31.9 44.2 21.6 127
Region % within 6.7% 4.9% 5.3% 8.0% 6.0%
Total % 1.5% 1.2% 1.8% 1.4% 6.0%
5 20-40 Count 132 113 129 101 475
times Expected Count ' 109.4 119.5 165.2 80.9 475
Region % within 26.9% 21.1% 17.4% 27.8% 22.3%
Total % 6.2% 5.3% 6.1% 4.7% 22.3%
TOTAL COUNT 491 536 741 363 2131
Total Region % within 100% 100% 100% 100% 100%
% of Total 23% 25.2% 34.8% 17% 100%

Table A-3
ANOVA Analysis of Variance
Between-Subjects Factors_____
Value Label N
2 NC:(2) 536
3 S:(3) 741
4 W:(4) 363
1 = North East
2 = North Central
3 = South
4 = West
Table A-4
Tests of Between-Subjects Effects
Source Type III Sum of Squares Df Mean Square F Sig.
Corrected Model 74.796 3 24.932 9.305 .000
Intercept 11196.959 1 11196.959 4178.748 .000
REGION 74.796 3 24.932 9.305 .000
Error 5699.298 2127 2.680
Total 17249.000 2131
Corrected Total 5774.094 2130
Dependent Variable: LFTMMJ Marijuana use in your life
R Squared = .013 (Adjusted R Squared = .012)

Table A-5
. Coefficients: Number of Respondents, n = 1016
Standardiz ed Coefficients t Sig.
Model Std. Error Beta
1 (Constant) .549 2.548 .011
Convatt .037 -.057 -2.014 .044
Drugpeer .023 .473 17.345 .000
Educatt .037 -.012 -.422 .673
Parntatt .026 -.015 -.535 .592
RACE .165 .061 2.124 .034
RegdumNC .157 -.043 -1.391 .165
RegdumNE .138 -.033 -1.094 .274
Religatt .033 -.152 -5.222 .000
SEX .120 -.070 -2.557 .011
a Dependent Variable: MARJ/LIFETIM

Table A-6
Multiple Comparisons
Dependent Variable: LFTMMJ Marijuana use in your life
Tukey HSD________________________________________
Mean Difference 0-J) Std. Error Sig. 95% Confidence Interval
(1) 001 (J) 001 Lower Upper
:SCHL :SCHL Bound Bound
1 NE:(1) 2 NC:(2) .20 .102 .215 -.07 .46
3 S:(3) ,35 .095 .002 .10 .59
4 W:(4) -.16 .113 .517 -.45 .14
2 NC:(2) 1 NE:(1) -.20 .102 .215 -.46 .07
3 S:(3) .15 .093 .373 -.09 .39
4 W:(4) -.35 .111 .008 -.64 -.07
3 S:(3) 1 NE:(1) -.35 .095 .002 -.59 -.10
2 NC:(2) -.15 .093 .373 -.39 .09
4 W:(4) -.50 .105 .000 -.77 -.23
4 W:(4) 1 NE:(1) .16 .113 .517 -.14 .45
2 NC:(2) .35 .111 .008 .07 .64
3 S:(3) .50 .105 .000 .23 .77
Based on observed means.
* The mean difference is significant at the .05 level.

Table A-7
Correlation Matrix
MJLFTM Convatt Drugpeer Educate Race Region Religatt SEX Parntatt
Pearson MJLFTM Correlation DeptVar 1.000
Convatt -.118 (.000) 1.000
Drugpeer .494 -.052 1.000
(.000) (.048)
Educate -.077 .180 -.066 1.000
(.007) (.000) (.018)
Race -.031 .162 -.123 .132 1.000
(.160) (.000) (.000) (.000)
Region -.030 .123 -.074 .067 .219 1.000
(.168) (.000) (.009) (.016) (.000)
Religatt -.233 .251 -.166 .178 .197 .247 1.000
(.000) (.000) (.000) (.000) (.000) (.000)
SEX -.099 .157 -.031 .069 .070 .069 .095 1.000
(.001) (.000) (.165) (.014) (.013) (.014) (.001)
Parntatt -.042 .021 -.014 .211 -.175 -.050 .064 -.091 1.000
(.091) (.248) (.330) (.000) (.000) (.055) (.020) (.002)
a. Predictors: (Constant), PARNTATT Parental Attachment, DRUGPEER
Drug Using Peers, CONVATT Conventional Attachment, REGION -4 CAT,
SEX, EDUCATT Educational Attachment, RACE, RELIGATT Religious
Bold Variable = Dependent Variable: MJLFTM Lifetime Marijuana Use

Table A-8
Model Summary: Number of Cases n = 1015__________________________
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .530 .281 .275 1.860
Table A-9
Model Sum of Squares df Mean Square F Sig.
1 Regression 1360.490 8 170.061 49.158 .000
Residual 3483.727 1007 3.460
Total 4844.217 1015
a. Predictors: (Constant), PARNTATT Parental Attachment, DRUGPEER
Drug Using Peers, CONVATT Conventional Attachment, REGION -4 CAT,
SEX, EDUCATT Educational Attachment, RACE, RELIGATT Religious
Bold Variable = Dependent Variable: MJLFTM Lifetime Marijuana Use

Independent Variables
Religious Attachment (2-8 composite range)
1. How often do you go to church? A. Never (1); B. Rarely (2); C. About once or
twice a month (3); D. Once a week or more (4).
2. How important is religion to you? A. Not important (1); B. Little important (2); C.
Pretty important (3); D. Very important (4).
Educational Attachment (2- 8 composite range)
3. Will you attend a two-year college? A. Definitely not (1); B. Probably wont (2);
C. Probably will (3); D. Definitely will (4).
4. Will you attend a four-year college? A. Definitely not (1); B. Probably wont (2);
C. Probably will (3); D. Definitely will (4).
Parental Attachment (2 -14 composite range)
5. What is your fathers education level? A. Grade school (1); B. some high school
(2); C. High school grad (3); D. Some college (4); E. College grad (5); F.
Graduate school (6).
6. What is your mothers education level? A. Grade school (1); B. some high school
(2); C. High school grad (3); D. Some college (4); E. College grad (5); F.
Graduate school (6).
Drug-using Friends (3 -15 composite range)
7. Do all your friends drink alcohol? A. None (1); B. A few (2); C. Some (3);
D. Most (4); E. All (5).

8. Do all your friends smoke cigarettes? A. None (1); B. A few (2); C. Some (3);
D. Most (4); E. All (5).
9. Do all your friends smoke marijuana? A. None (1); B. A few (2); C. Some (3);
D. Most (4); E. All (5).
Conventional Attachment (3-16 composite range)
10. How important is it to be married and have a family? A. Not important (1); B.
Somewhat important (2); C. Quite important (3); D. Extremely important (4).
11. How important is it to be a leader in your community? A. Not important (1); B.
Somewhat important (2); C. Quite important (3); D. Extremely important (4).
12. How important is it to find a purpose in your life? A. Not important (1); B.
Somewhat important (2); C. Quite important (3); D. Extremely important (4).
Dependent Variables
Lifetime Marijuana (recoded)
How many times have you smoked marijuana in your life? A. Zero occasions (1);
B. One to two times (2); C. Three to five times (3); D. Six to nine times (4); E.
Ten to nineteen times (5); F. Twenty to thirty-nine times (6); G. Forty + times (7).
How many times have you smoked marijuana in your life? A. Zero occasions (1);
B. One to five times (2); C. Six to nine times (3); D. Ten to nineteen times (4); E.
Twenty to Forty + times (5).

Table C-l
Rotated Component Matrix
1 2 3 4
CLLG2YR .848
CLLG4YR .813
Extraction Method: Principal Component Analysis. Rotation Method: Varimax
with Kaiser Normalization,
a. Rotation converged in 5 iterations.

Aseltine, Robert H., Jr. (1995). A Reconsideration of Parental and Peer Influences on
Adolescent Deviance. Journal of Health and Social Behavior, 36 (2), 103-121.
Bachman, Jerald G., OMalley, Patrick M., and Johnston, Lloyd D. (1991). Change
and Consistency in the Correlates of Drug-use Among High School Seniors 1975-
1986 (Monitoring the Future Occasional Paper 21). Ann Arbor: Institute for
Social Research.
Bailey, Susan L. and Hubbard, Robert L. (1990). Developmental Variation in the
Context of Marijuana Initiation Among Adolescents. Journal of Health and
Social Behavior, 35, 58-70.
Brook, Judith S., Gordon, Ann S., Brook, Adam, and Brook, David W. (1989). The
Consequences of Marijuana Use on Interpersonal and Interpersonal Functioning
in Black and White Adolescents. Genetic, Social, and General Psychology
Monographs, 115, 349-69.
Brunswick, A. (1988), Young Black Males and Substance Abuse. In J. Giws( Ed.),
Young, Black, and Male in America: An Endangered Species, (pp. 167-87). Dover,
MA.: Auburn.
Chilcoat, Howard D., and Breslau, Naomi. (1996). Alcohol Disorders in Young
Adulthood: Effects of Transitions into Adult Roles. Journal of Health and Social
Behavior, 4, 339-349.
Conger, R. D. (1976). Social Control and Social Learning Models of Delinquent
Behavior: A Synthesis. Criminology, 14, 17-40.
Crutchfield, Robert D. and Pitchford, Susan R. (1997). Work and Crime: The Effects
of Labor Stratification. Social Forces, 76,93-118.

Donovan, John E., and Jessor, Richard (1985). Structure of Problem Behavior in
Adolescence and Young Adulthood. Journal of Consulting and Clinical
Psychology, 53, 890-904.
Dryfoos, Joy G. (1990). Adolescents at Risk: Prevalence and Prevention. New York:
Oxford University Press.
Erickson, Kristan G., Crosnoe, Robert, and Dombusch, Sanford G. (2000). A Social
Process Model of Adolescent Deviance: Combining Social Control and
Differential Association Perspectives. Journal 'of Youth and Adolescence, 29,
Farrington, D., and Tarling, R. (1985). Prediction in Criminology. Albany, NY:
SUNY Press,
Glueck, Sheldon, and Glueck, Eleanor T. (1934). Five Hundred Delinquent Women.
New York: Knopf.
Grumberg, M. E., and Baum, A. (1984). Social Psychological Approaches to
Cigarette Smoking Behavior. Journal of Applied Social Psychology, 14,181-83.
Hagan John, Simpson John, and Gillis A.R.. (1987). Class in the Household: A
Power-Control Theory of Common Delinquent Behavior. American Journal of
Sociology, 92, 788-816.
Harris, Anthony R. (1977). Sex and Theories of Deviance: Toward a Functional
Theory of Deviant Type-Scripts. American Sociological Review, 42, 3-16.
Heimer, Karen. (1996). Testing a Theory of Differential Social Control. Social
Psychology Quarterly, 59, 39-61.
Hindelang, Michael. (1971). Age, Sex, and the Versatility of Delinquent
Involvements. Social Problems, 18, 522-535.
Hindelang, M., Hirschi, T., and Weis, J. (1981). Measuring Delinquency. London:

Hirschi, Travis. (1969). Causes of Delinquency. Berkeley: University of California
Humble, C. G., Samet, J.M., Pathak, D.R., and Skipper, BJ. (1985). Cigarette
Smoking and Lung Cancer in Hispanic Whites and Other Whites in New
Mexico American Journal ofPublic Health, 75, 145-48.
Jacuith, S. M. (1981). Adolescent Marijuana and Alcohol Use: An Empirical Test of
Differential Association Theory. Criminology>, 19, 271-80.
Jenkins, Jeanne E. (1996). The Influence of Peer Affiliation and Student Activities
on Adolescent Drug Involvement. Adolescence, 31, 297-307.
Johnston, L., OMalley, P., and Bachman, J. (1987). National Tends in Drug-use and
Related Factors among American High School Students and Young Adults, 1975-
1986. (DHHS Pub. No. ADM 87-1535). Washington, D. C.: U.S. Government
Printing Office.
Kandel, Denise B., Treiman, Donald, Faust, Richard, and Single, Eric. (1976).
Adolescent Involvement in Legal and Illegal Drug-use: A Multiple Classification
Analysis. Social Forces, 55, 438-458.
Kempf-Leonard, Katherine. (1992). Race, Gender, and the Prehearing Detention of
Juveniles. Thousand Oaks, CA: Sage Publications.
Krohn, Marvin D,. Massey James L., and Zielinski, Mary. (1988). Role
Overlap, Network Multiplexity, and Adolescent Deviant Behavior.
Social Psychology Quarterly, 51, 346-56.
Larson, Ron, and Farber, Betsy. (2003). Elementary Statistics: Picturing the World
(2nd ed.). Upper Saddle River, NJ: Prentice Hall.
Lichtenstein, E. (1982). The Smoking Problem: A Behavioral Perspective.
Journal of Consulting and Clinical Psychology, 50, 804-19.

Macdonald, Donald I. (1984). Drugs, Drinking and Adolescents. Journal of American
Medicine (JAMA), 138 (2), 117-125
Marcos, Anastasios C., Bahr, Stephen J., and Johnson, Richard E. (1985).
Test of a Bonding/Association Theory of Adolescent Drug-use. Social
Forces, 65,135-161.
Matsueda, Ross L. (1982). Testing Control Theory and Differential Association: A
Causal Modeling Approach. American Sociological Review, 47,489-504.
Matsuada, Ross L. and Heimer, Karen. (1987). Race, Family Structure, and
Delinquency: A Test of Differential Association and Social Control Theories.
American Sociological Review, 52, 826-840.
McCord, J. (1990). Problem Behaviors. In Feldman, S., and Elliott, G. (Eds.), At the
Threshold. Cambridge, MA: Harvard University Press.
McNeal, Ralph. B., Jr. (1999). Parental Involvement as Social Capital: Differential
Effectiveness on Science Achievement, Truancy, and Dropping Out. Social
Forces, 78, 117-144.
Mensch, Barbara and Kandel, Denise B. (1988). Do Job Conditions Influence the
Use of Drugs? Journal of Health and Social Behavior, 29,164-184.
Mensch, Barbara and Kandel, Denise B. (1988). Dropping Out of High School and
Drug Involvement. Sociology of Education, 61, 95-113.
Messerschmidt, James W. (1979). School Stratification and Delinquent Behavior.
Stockholm: Gotab.
Messerschmidt, James W. (1993). Masculinities and Crime. Blue Ridge Summit,
PA: Rowman and Littlefield Publishers, Inc.
Miller, Peter V. (1997). Review: Is Up Right? The National Household
Survey On Drug Abuse (in The Polls). Public Opinion Quarterly, 61,

Morgan, F. J., and Sonquist, J. (1969). Multiple Classification Analysis.
Ann Arbor: University of Michigan, Institute for Social Research.
Polk, K. (1971). A Reassessment of Middle-class Delinquency. Youth and Society, 2,
Schulenberg, John, Bachman,Gerald G., OMalley,Patrick M., and Johnston, Lloyd
D. (1994). High School Educational Success and Subsequent Substance Use: A
Panel Analysis Following Adolescents in to Young Adulthood. Journal of Health
and Social Behavior, 35,45-62.
SPSS for Windows, Standard Version. (15 November, 2001). Chicago, IL: SPSS Inc.
West, D.J. (1982). Delinquency: Its Roots, Careers, and Prospects. Cambridge, MA:
Wiatrowski, Michael D., Griswold, David B., and Roberts, Mark K. (1980). Social
Control Theory and Delinquency. American Sociological Review, 461, 525-541.