Family risk and resiliency factors, substance use, and the drug resistance process in adolescence

Abstract

Recent approaches to drug prevention have emphasized risk and resiliency factors. Two models have been developed to explain these factors, one which posits that separate elements make up each set and the other which posits that a single factor can be either a risk or a resiliency factor depending on, for example, if it is present (resiliency) or absent (risk). This study tested these models and attempted to compare the effects of risk and resiliency across gender and ethnicity. Results support the model in which risk and resiliency are discrete sets of factors and demonstrate that overall resiliency factors play a larger role than risk factors in substance use and drug resistance processes. However, gender proved to be an important moderator of these effects. For adolescent males, resiliency has an indirect effect on overall substance use through age of first use, while risk has a direct effect on overall substance use. For adolescent females, resiliency has a direct effect on overall substance use and risk has an indirect effect through age of first use. This indicates that while early interventions are important for both genders, resiliency factors must be dealt with before initiation of substance use for males. Findings did not differ substantially across ethnicity, although the small African-American sample size may have limited power to detect differences.

Full text

Rapid #: -11726176
CROSS REF ID: 202981
LENDER: CNO :: Main Library
BORROWER: CS1 :: Main Library
TYPE: Article CC:CCL
JOURNAL TITLE: Journal of drug education
USER JOURNAL TITLE: Journal of drug education.
ARTICLE TITLE: FAMILY RISK AND RESILIENCY FACTORS, SUBSTANCE USE, AND THE DRUG RESISTANCE
PROCESS IN ADOLESCENCE.
ARTICLE AUTHOR:
VOLUME: 30
ISSUE: 4
MONTH:
YEAR: 2000
PAGES: 373-398
ISSN: 0047-2379
OCLC #: 1022664
Processed by RapidX: 3/29/2017 12:42:29 PM
This material may be protected by copyright law (Title 17 U.S. Code)

J. DRUG EDUCATION, Vol. 30(4) 373-398, 2000
FAMILY RISK AND RESILIENCY FACTORS,
SUBSTANCE USE, AND THE DRUG RESISTANCE
PROCESS IN ADOLESCENCE*
DREAMA G. MOON
California State University, San Marcos
KRISTINA M. JACKSON
University of Missouri, Columbia
MICHAEL L. HECHT
Pennsylvania State University, University Park
ABSTRACT
Recent approaches to drug prevention have emphasized risk and resiliency
factors. Two models have been developed to explain these factors, one which
posits that separate elements make up each set and the other which posits
that a single factor can be either a risk or a resiliency factor depending on,
for example, if it is present (resiliency) or absent (risk). This study tested
these models and attempted to compare the effects of risk and resiliency across
gender and ethnicity. Results support the model in which risk and resiliency
are discrete sets of factors and demonstrate that overall resiliency factors play
a larger role than risk factors in substance use and drug resistance processes.
However, gender proved to be an important moderator of these effects. For
adolescent males, resiliency has an indirect effect on overall substance use
through age of first use, while risk has a direct effect on overall substance use.
For adolescent females, resiliency has a direct effect on overall substance use
and risk has an indirect effect through age of first use. This indicates that
while early interventions are important for both genders, resiliency factors must
be dealt with before initiation of substance use for males. Findings did not
differ substantially across ethnicity, although the small African-American sample
size may have limited power to detect differences.
*Supported under grant 5 RO1 DA 05629-10A1 from the National Institute of Drug Abuse.
373
Ó2000, Baywood Publishing Co., Inc.

Adolescent substance prevention research has developed two general approaches to
understanding substance use and prevention. The first suggests that certain factors
place a person at risk for substance use and abuse and argues that interventions target
these factors for elimination or reduction by, for example, decreasing family stress
[1]. Others have focused on resiliency or protective factors which are thought to
allow individuals to overcome risk and avoid use and abuse. Interventions evolving
from a resiliency perspective focus on enhancing those factors thought to protect
against or reduce substance use [2]. Recently, researchers have begun to advocate
blended models that attempt to both eliminate and/or reduce risk factors while
simultaneously strengthening or creating factors protective against substance use
and abuse. Research using this blended model has begun to assess the relative
impacts of risk and resiliency on substance use (e.g., [3, 4]). This study is designed to
extend this line of research by examining two different models of risk and resiliency,
comparing the effects of selected risk and resiliency factors on substance use, and
considering the roles of gender and ethnicity.
FAMILY RISK AND RESILIENCY FACTORS AND SUBSTANCE USE
Risk and resiliency factors fall into interrelated domains which include indi-
vidual factors, interpersonal forces (i.e., school, peers, and family), and social and
cultural environment [2, 4]. Of all the contexts in which potential risk and
protective factors occur, the family is thought to be the most powerful as it exerts
the most influence over the child’s emotional, social, psychological, and physical
environment [5]. Family risk factors of chronic family tension and discord [6],
parental drug abuse [5], parental nondirectiveness, permissiveness, and inade-
quate supervision [5], and living in a neighborhood that is perceived as unsafe [7]
have been found to be particularly important.
It has been argued that the single most important family protective factor may
be having a warm, positive relationship with an adult caretaker [7-9]. Rutter found
that having a warm and supportive relationship with one parent can offset the
negative influence of a dysfunctional parent or of living in a family with a great
deal of conflict and tension [7]. Other critical family protective factors are parental
expectations that the child will not drink or use drugs [10, 11], having to assume
family responsibilities or chores [12], parental discipline and monitoring of their
children’s behavior [12], and religiosity [13].
Until recently, resiliency and risk factors have been conceptualized as repre-
senting separate sets of factors [3, 14]. However, this may not be the case.
Newcomb and Felix-Ortiz argue that a single factor may be either a resiliency/
protective factor or a risk factor depending on where along its continuum a person
or family falls [14]. For example, high religiosity may be a resiliency factor while
low religiosity may be a risk factor. The alternative approach is to classify
religiosity as either a risk or a resiliency factor, but not both. Our first goal was to
test these competing models:
374 / MOON, JACKSON AND HECHT

RQ1: Can risk and resiliency best be described as two ends of a
single continuum or as separate sets of factors?
While it is clear that both risk and resiliency affect substance use, research has
not as yet established the relative influence of each set of factors and it is possible
that certain factors may be more powerfully protective or risk-inducing than others
[14]. For example, parental drug abuse is associated with adolescent drug use [5],
while having a warm, supportive relationship with an adult caretaker has been
shown to offset the negative influence of a dysfunctional parent [7]. Perhaps
resiliency factors such as having a positive relationship with an adult caretaker are
more strongly related to adolescent substance use than are risk factors such as
parental drug abuse. Some studies indicate that adolescents may experiment with
certain drugs even when not exposed to risk conditions and clearly, not all
adolescents exposed to high risk situations abuse or even use drugs [15]. The
presence or absence of resiliency or protective factors may account for these
variations. An effective model of substance use may need to consider both sets of
factors or one or the other. This leads to the second research question:
RQ2: Are risk or resiliency factors more important predictors
of substance use?
Moderating Effects of Gender and Ethnicity
Scholarly work on ethnic differences in substance use which indicates that
some ethnic groups are more at risk for use than others [15-19, 21, 22] has
sometimes led to the treatment of ethnicity as a risk factor itself. Likewise,
males tend to exhibit greater risk for substance use than females [18, 20, 23, 24].
Other research has examined risk and resiliency factors across ethnic groups
and gender, suggesting that what constitutes risk and/or resilience for some
groups may differ for others and that ethnicity and gender may moderate the
effects of specific risk and protective factors [3, 4]. Following this second line of
inquiry, we examine ethnicity and gender as moderating influences on selected
family risk and resiliency factors.
Research Hypotheses
In line with previous findings, it is expected that high risk and/or low resiliency
adolescents should report a higher level of involvement with drugs than high
resiliency and/or low risk adolescents. Further, we believe that the effects of risk
and resiliency are moderated by gender and ethnicity. We posited the following
hypotheses:
H1: Risk and resiliency have independent effects on substance
use.
FAMILY RISK AND RESILIENCY FACTORS / 375

H2: Gender and ethnicity moderate the effects of risk and
resiliency on substance use.
DESIGN OF THE STUDY
Participants
A survey was administered to 995 seventh grade students (57% female) in the
Phoenix metropolitan area. Of these 995 students, 55.3 percent were Mexican
American, 16.5 percent were European American, 10.3 percent of mixed racial
heritage, 9.9 percent African American, 3.6 percent Native American, 2.3 percent
Japanese American, and 2 percent Chinese American. For purposes of analysis,
only Mexican Americans, African Americans, and European Americans were
included, reducing the sample size to 609 students. Of these 609 students, almost
three-quarters (n= 442, 73%) reported being offered a drug and over a third
(n= 161, 36%) reported accepting the drug offer.
Procedure
Participants were recruited from local middle schools. The principal of each
school was contacted and asked if the school might participate in the study.
Both individual and systemic incentives were offered to encourage participa-
tion. Thirty-three of forty-five invited schools participated.1Once permission had
been obtained, research teams administered questionnaires to students in their
regular classrooms. Students were informed of the nature of the study (i.e., “We
want to understand how kids are offered and say ‘no’ to drugs. If kids have
difficulty saying ‘no,’ we want to understand why” ) and asked to participate. They
also were told that the survey would ask other questions about a number of areas in
their lives (i.e., family, neighborhood). Lastly, the difference between a “test” and
a “survey” was explained to the students and they were informed that their
anonymity would be protected by not requesting that their names appear on
the form.
Once the students were given verbal instructions, they were given a question-
naire and a computer (scantron) answer sheet. Instructions in using the scantron
sheet were given and sample questions answered. Research team members were
available during the administration time to answer questions and/or provide
clarification. Students were familiar with scantron-type format.
376 / MOON, JACKSON AND HECHT
1The forty-five schools are located in a large metropolitan city in the southwestern United States.
They are the feeder schools for the high school unified school district. All are part of a larger drug
prevention project funded under the same grant. Time constraints associated with this preliminary
study allowed us to include only thirty-three of the schools. They represent all areas of the city.

The questionnaire was pre-tested on twelve seventh-grade students recruited
from a local Boys and Girls Club. Four of the youth were females (2 Anglo
American and 2 African American) and eight were males (2 Mexican American
and 6 African American). As a result of the pilot test, four versions of the
survey were developed. The length of each was reduced from 162 questions to
120, and the format of the questionnaire was changed to include more directional
information to assist respondents in keeping on track in answering the questions.
In addition, wording of questions and responses which were thought to be unclear
were changed and simplified. For instance, the test indicated that adolescents
of this age had difficulty understanding the terms “African American” and “Anglo
American,” hence, the alternate terms of “White” and “Black” were utilized.
The four versions were randomly distributed to the sample by giving every
fourth student a different version. The present article reports the findings from one
of these four versions administered to 25 percent of the total sample (N= 995).
Measures
In this section, we first discuss how the independent measures were opera-
tionalized, then the development of the risk and resiliency scale, and lastly, the
operationalization of the outcome measures.
Gender and Ethnicity
Students indicated their gender and ethnicity on checklists. Choices for eth-
nicity were: Chinese/Chinese American, Japanese/Japanese American, Black/
African American, Mexican American/Hispanic/Chicano/Latino, White/Anglo
American, Native American/American Indian, Mixed Heritage, and Other.
Risk and Resiliency
Risk and resiliency were constructed from six factors identified in previous
research. Risk was assessed by family stressor events, unsafe neighborhood,
and parental substance use, while resiliency was assessed by family relations,
family permissiveness, and religiosity. Each of these measures is described next,
followed by a description of the development of the risk and resiliency scales.
Family Stressors
The amount of family stress experienced by the participants was measured
using three items from Gonzales’ family stressor subscale and two items from
Gonzales’ family conflict subscale [25]. The first three items asked respondents if
a close family member or someone they lived with a) got drunk or high, b) had
serious emotional problems, and/or c) had committed a crime/got in trouble with
the law/was sent to jail. The next two items asked if members of the participants’
families had hit or hurt one another and/or refused to speak with one another. For
FAMILY RISK AND RESILIENCY FACTORS / 377

all five items, participants were asked to indicate how often they had experienced
the situation listed in the questions using a 5-point scale which ranged from never
to almost every day. The responses to the five items were summed and a mean
score computed. A confirmatory factor analysis (CFA) indicated that these five
items fit together well, P2(5, n= 40) = 13.62, p= .02; CFI = .96.
Religiosity
A 4-item scale constructed for this study assessed participants’ involvement
with religion. Respondents were asked to indicate their religion, which was
collapsed into “religious” or “non-religious” (“Do not practice or believe in
any religion” ). Moreover, respondents indicated their own level of activity,
their involvement, and family involvement with their religion on 5-point scales
which ranged from “very active (involved)” to “ not at all active (involved).” A
CFA on the religion scale showed good fit, P2(2, n= 3898) = 67.65, p< .001;
CFI = .99.
Parental Drinking Behavior
Two items taken from Dielman, Butchart, and Shope’s parent’s alcohol use
subscale were used to determine how much participants’ parents drank alcohol
[26]. These were “When your father/mother drinks alcohol, how much does he/she
usually drink?” Responses fell on a 5-point scale ranging from “s/he doesn’t drink
alcohol” to “s/he gets drunk.” These two items were correlated r= .40.
Relationship with Parents
Adolescents’ relationships with their parents were measured using six items
from Field and Yando’s mother/father/friend relationship subscale [27]. Items
were designed to assess how much the respondent perceived their mother/father
accepted them, how often they shared inner feelings or secrets with their
mother/father, and how satisfied they were with the relationship they had with
their mother/father (e.g., “How much does your mother accept you no matter what
you do?” ). A 5-point scale ranged from “not at all” to “very much.” A mean score
was computed from the responses for all six items. A higher mean score reflected
positive parental relations. A CFA showed adequate fit, P2(9, n= 456) = 193.05,
p<.001; CFI = .76.
Parental Permissiveness
Two scales assessed respondents’ perception of parental permissiveness.
Dielman et al.’s four-item permissiveness subscale determines how often the par-
ticipants, a) were allowed to go out when they wanted to, b) got away without
doing work they were told to do, c) were let off easy when they did something
378 / MOON, JACKSON AND HECHT

wrong, and d) were allowed to spend money they had earned on whatever they
wanted [26]. The scale had a 4-point response range (i.e., always to never). Again,
mean scores were computed from the four items. Model fit for the CFA was very
good, P2(2, n= 423) = 6.19, p= .045; CFI = .98.
Two items from Dielman et al’s Parent’s Approval of Kid’s Drinking Behavior
subscale were also used to assess parental permissiveness, “Do your parents allow
you to drink alcohol at parties when they are present?” and “Do your parents allow
you to drink at parties when they are not present?” [26]. The two binary (yes/no)
items were moderately correlated, r= .46.
Neighborhood Safety
Perceived neighborhood safety was measured using three items constructed for
this survey. These items were: “How safe do you feel walking alone in your
neighborhood during daylight?,” “How safe do you feel walking alone in your
neighborhood after dark?,” and “How safe do you feel in your home?” A 5-point
scale assessed the extent to which respondents felt “not at all safe” versus “very
safe.” Correlations between these items ranged between r= .33 and r= .56.
Construction of Risk and Resiliency Scales
In order to explore the first research question, separate scales for risk factors
and resiliency factors were created under the assumptions that risk and resiliency
may not be two ends of the same construct [14]. In order to determine which
items were risk factors and which were protective factors, we began by fol-
lowing a procedure utilized by Newcomb and Felix-Ortiz [14]. Newcomb and
Felix-Ortiz designated the upper 20 percent and lower 20 percent of each vari-
able’s distribution as either risk or protective elements. For example, the upper
20 percent of religiosity was designed as being a protective factor, while the
lower 20 percent was designated as risk. For each variable, two dichotomous
variables were created by assigning a value of 1.0 to the upper (lower) 20 percent,
and a value of zero to the remaining 80 percent. Thus, two dichotomous variables
were created per item, one for risk and one for protection, which were then
correlated with the outcome measure, substance use. A single item was desig-
nated as being a risk factor if the dichotomous risk variable (i.e., lower 20% of
religiosity verus the remaining 80%) correlated more with substance use than with
the dichotomous protective variable (i.e., upper 20% of religiosity versus the
remaining 80%) and vice versa for protective factors.
Using this procedure, twelve dichotomous variables were created, one for
each of the six risk/resiliency factors. These dichotomous variables were then
correlated with five variables tapping substance use in the past month and the
average correlation across the five substance use items was calculated. The
resulting average correlations are presented in Table 1. An examination of Table 1
FAMILY RISK AND RESILIENCY FACTORS / 379

suggests that the correlations between the dichotomous risk variables and
substance use do not differ significantly from those between the dichotomous
protective variables and substance use. Thus, it did not seem appropriate to
designate a factor as a resiliency factor simply because its correlation with
substance use (e.g., parent substance use, r= –.07) was nominally stronger than its
correlation with substance use when constructed as a risk factor (e.g., parent
substance use, r= .06). This method seems appropriate only when a correlation
is significantly larger than its counterpart.
Our next step was to examine whether the correlations with substance use were
significantly different when a variable was constructed as a dichotomous risk
factor rather than when it was constructed as a dichotomous resiliency factor. To
do so, we transformed the absolute value of each correlation to a Fischer z’ score
and then tested differences with a z-test [28]. No risk and resiliency correlations
were significantly different. This suggests that the approach utilized by Newcomb
and Felix-Ortiz was not a fruitful one to employ with these data, as we were not
able to clearly differentiate between risk and resiliency. Perhaps some of the
difficulty lies in the fact that the scale properties of our risk/resiliency items were
different than those used by Newcomb and Felix-Ortiz; their scales were 7-point
Likert scales while ours, in some instances, were composites of dichotomous
items. However, Newcomb and Felix-Ortiz also reported difficulty with items
which could not be clearly assigned as either risk or resiliency factors. We
conclude that each variable may be classified as either a risk or a resiliency factor
(RQ1).
Due the aforementioned problems in following the procedures utilized in
Newcomb and Felix-Ortiz, we turned to the theoretical literature on risk and
resiliency and designated the following as risk factors: family stressor events [6],
unsafe neighborhood [7], and parental substance use [5], while family relations
380 / MOON, JACKSON AND HECHT
Table 1. Average Correlations with Substance Use in the Past Month
for the Eight Dichotomous Risk Variables and the
Eight Dichotomous Protective Variables
Variable r, Risk Variable
r, Resiliency
Variable z-valuea
Family Stressors
Unsafe Neighborhood
Parent Substance Use
Religiosity
Family Relationships
Low Family Permissiveness
.160
–.004
.062
.033
.131
–.099
–.095
.042
–.070
–.028
–.095
.120
.95
–.62
–.11
.11
.56
–.32
aZ-test of the difference between the z’ transformed correlations between substance use
and variables as risk factors and as resiliency factors.

[7, 9], low family permissiveness [12], and high religiosity [13] were determined to
be protective factors. We then tested these factors in a 2-factor CFA. The fit of the
model was fair, P2(8, n= 376) = 36.66, p<.001; CFI = .80. After examining modifi-
cation indices, we concluded that neighborhood should be re-designated as a
protective factor and that religiosity should be re-assigned as a risk factor. With these
changes, the fit of the model improved, P2(8, n= 376) = 20.82, p< .01; CFI = .91.
The final set of risk factors included high family stressor events, low religiosity, and
high parental substance use, while the resiliency factors were high family relations,
low parental permissiveness, and high perceived neighborhood safety.
Outcome Measures
Six outcome variables operationalized substance use: 1) substance use in the
past month, 2) whether or not respondents have ever been offered drugs,
3) lifetime frequency of substance use, 4) age of initiation (first use), 5) whether
respondents accepted the last drug offer, and 6) the context of the last drug offer.
Substance Use in the Past Month
Substance use in the past month was assessed by a 6-item scale modeled after
one used by Flannery, Vazsonyi, and Torquati [29]. Participants were asked to
indicate how frequently in the past month they smoked cigarettes or used any
tobacco products, drank alcohol, used uppers, smoked marijuana, and/or used
other drugs such as hallucinogens, crack, and downers. To do so, participants
marked that they had either (0) not used the drug in the last thirty days, 1) used
the drug one or two days, 2) three to seven days, 3) eight to fourteen days, or
4) fifteen days or more. The scores for these questions were summed and a mean
score was computed.
The remaining outcomes measures were derived from previous research in
the same population, which showed them to be valid measures [30-31].
Ever Offered Drugs
The extent to which an adolescent had ever been offered drugs was assessed
by taking a count of whether the following seven drugs were ever offered to
the adolescent: cigarettes/chewing tobacco, beer or wine, hard liquor, marijuana,
hard drugs, uppers, and inhalants. A participant’s possible score ranged from
0 (no drugs were ever offered) to 7 (all seven drugs has been offered).
Number of Drugs Used in Lifetime
To assess the number of drugs an adolescent had used at least once (“lifetime
use” ), a count was taken of ever-use of seven drugs, including cigarettes/chewing
tobacco, beer or wine, hard liquor, marijuana, hard drugs, uppers, and inhalants.
FAMILY RISK AND RESILIENCY FACTORS / 381

Possible scores ranged from 0 (never used any of the drugs to 7 (used all seven
drugs at least once).
Age of Initiation
The age at which adolescents had first used drugs was assessed by taking the
mean of four items: first smoked a cigarette, first drank alcohol, first smoked
marijuana, and first tried hard drugs (i.e., cocaine, crack, psychedelics, inhalants,
narcotics). Possible responses included (0) eight years old or under, 1) nine to ten
years old, 2) eleven to thirteen years old, 3) fourteen years old or older, and
4) never used or tried once.2A CFA showed strong internal consistency for this
scale, P2(2, n= 4085) = 297.16, p< .001; CFI = .92.
Last Offer
Respondents were asked to think about the last time the received a drug offer.
They indicated whether they accepted or rejected the offer.
Context
The setting of the drug offer was examined with a 6-item response including
at a park, at school, at a party, on the street, at a friend’s home, and at your
home.
RESULTS
There was a great deal of attrition over the course of the survey as a number
of adolescents did not get to the later survey items during the given time period.
Of the 609 European-American, African-American, or Mexican-American
adolescents who began the survey, only 279 completed the last item which was
used to calculate resiliency. The regression analyses below reflects this high
attrition rate. All subsequent analyses utilize listwise deletion.
Attrition Analyses
Individuals who completed the survey were compared to those who did not
complete the survey on the variables of interest to this article. Attritors had higher
risk status than retained respondents, t(301) = 3.21, p= .001, and were less likely to
chose White/Anglo American as their ethnic designation, P2(2, n= 609) = 12.82,
p<.01.
382 / MOON, JACKSON AND HECHT
2Age of initiation assesses when adolescents began regular use or extended experimentation.
Accordingly, “tried once” is categorized with “never used,” as a separate category from the age during
which regular use or extended experimentation began.

Gender and Ethnic Differences in Risk and Resiliency
We first examined whether risk and resiliency level differed by gender and/or
ethnicity using a multivariate analysis of variance. Neither gender, ethnicity, nor
the interaction between the two significantly predicted risk or resiliency status.
Essentially, the risk and the resiliency profiles were similar for males and females,
and for African-American, Anglo-American, and Mexican-American adolescents.
Ever Offered, Lifetime Use, Age of Initiation, and
Substance Use in the Last Month
Next, we tested to see if risk and resiliency predicted whether a drug had ever
been offered, lifetime substance use, age of first substance use, and substance
use in the past month using multiple regressions. In addition, we examined
the role of ethnicity and gender as moderators of this relationship. Hence, we
tested a four-way interaction between risk, resiliency, gender, and ethnicity.
Only significant effects that contained risk and/or resiliency are reported here;
significant findings regarding ethnic and gender similarities and differences are
reported in Moon, Hecht, Jackson, and Spellers [32]. Risk and resiliency were
centered around the mean for ease of interpretation [33] and gender and ethnicity
were dummy coded: one dummy code for gender (male/female) and two dummy
codes for ethnicity (one for African American versus others and one for Mexican
Americans versus others). If a significant result was found for ethnicity, the
dummy code which demonstrated significance is indicated in parentheses.
For ever offered, a significant three-way interaction between resiliency, sex,
and ethnicity (the dummy code for Mexican Americans) was observed, R2= .21;
standardized $= .31; t(273) = 2.00, p< .05, see Figure 1 for predicted scores. Low
resiliency Mexican American females were most likely to have ever been offered
drugs, while low resiliency African American and European American males were
most likely to have been ever offered drugs.
For lifetime use, there were main effects for resiliency and risk, R2= .16;
standardized $= –.27; t(289) = –4.68, p< .001; standardized $= .19; t(289) = 3.31,
p< .01, respectively. Low resiliency and high risk adolescents were more likely to
have used more substances than were high resiliency or low risk adolescents.
Neither gender nor ethnicity moderated these effects.
Third, there were main effects for both risk and resiliency for age of initiation,
R2= .14; standardized $= –.14; t(289) = –2.36, p< .05; $= .28; t(289) = 4.80,
p< .0001, respectively. Specifically, high risk and low resiliency adolescents were
more likely to begin substance use at earlier ages. Gender and ethnicity did not
moderate this effect.
Finally, resiliency significantly predicted substance use in the last month,
R2= .09; standardized $= –.22; t(288) = –3.75, p< .001. Low resiliency
FAMILY RISK AND RESILIENCY FACTORS / 383

adolescents were more likely to have used drugs in the past month than were
high resiliency adolescents.
The Drug Offer Situation
The outcome variables assessing whether the most recent offer was accepted
and the context of the drug offer were dichotomous in nature; hence, logistic
regressions were used to examine the prediction of these variables by risk and
resiliency, as moderated by gender and ethnicity. For purposes of interpretation in
presentation of results in figures, risk and resiliency were dichotomized into
“high” and “low” using a mean split. Only adolescents who had ever been offered
drugs were included in these analyses.
Most Recent Offer
The odds of accepting a drug offer were 2.41 times higher for low resil-
iency adolescents than high resiliency adolescents, standardized $= .29; Wald
P2(1, n= 247) = 11.47, p< .001; OR = 2.41.
384 / MOON, JACKSON AND HECHT
Figure 1. Extent to which adolescent reported ever being
offered drugs by resiliency, ethnicity, and gender.

Context of the Drug Offer
Three two-way interactions were observed in terms of whether the setting of
the drug offer was on the street. There was an interaction between risk and
resiliency, an interaction between risk and gender, and an interaction between
resiliency and gender; standardized $= .48; Wald P2(1, n= 167) = 7.37, p< .01;
OR = 6.58; standardized $= .61; Wald P2(1, n= 167) = 4.72, p< .05; OR = 11.80;
standardized $= .81; Wald P2(1, n= 167) = 6.15, p< .05; OR = 24.50, respectively.
Due to the fact that no low-risk, low-resiliency females reported a drug offer on the
street, the model with the three-way interaction between risk, resiliency, and gender
did not converge. Interpretation of the three two-way interactions is presented in
Figure 2. Males were more likely to report that the drug offer occurred on the street,
especially if they were low resiliency; whereas high resiliency females were most
likely to report a drug offer on the street. Further, for males, being high risk and low
resiliency was most associated with a drug offer on the street but for females, low
risk, high resiliency was most associated with a drug offer on the street.
The odds of a drug offer at school was 6.10 times higher for low resiliency than
high resiliency adolescents, standardized $= .61; Wald P2(1, n= 167) = 9.96,
p< .01; OR = 6.10. Finally, the odds of a home setting was 2.71 times higher for
high resiliency than low resiliency adolescents, standardized $= .34; Wald
P2(1, n= 167) = 4.39, p< .05; OR = 2.71.
Models
We examined some of the above relationships in an organized model frame-
work (see Figure 3). We believed that risk and resiliency were indirectly related to
substance use through age of first use. We were also interested in whether this
relationship differed among ethnic groups or gender. Correlation matrices for
these four variables across ethnic group, gender, and for the overall sample are
presented in Table 2. First, we tested equality of the covariance matrices for this
set of variables across gender and across ethnicity using two Box’s M tests. The
Box’s M test for gender was significantly different than zero, Box’s M = 25.31,
P2approximation (10, n= 295) = 24.94, p< .01, suggesting that the models for
males and females should be examined separately.
The Box’s M for ethnicity was marginally significantly different than zero,
Box’s M = 30.61, P2approximation (20, n= 296) = 29.62, p< .08, suggesting that
the covariance matrices across ethnicity were somewhat different. We chose to
examine the three ethnicities separately in three separate models. We did not test
for differences in the covariance matrices for the interaction between gender and
ethnicity as we would not be able to examine African-American males and females
separately due to the limited sample size for African Americans (n= 35).
Latent variable models for each gender and each ethnicity were run using EQS
[34] (see Tables 3 and 4). The models for females and males are presented in
FAMILY RISK AND RESILIENCY FACTORS / 385

386 / MOON, JACKSON AND HECHT
Figure 2. Percent of adolescents who reported that drug offer
location was on the street by resiliency and risk for males (top panel)
and females (bottom panel).

Figure 4.3All together, risk, resiliency, and age of first use accounted for
65 percent of the variance in substance use for females and 59 percent of the
variance in males.
Examination of the model suggests that for females, age of first use mediates the
relationship between risk and substance use, whereas for males, age of first use
mediates the relationship between resiliency and substance use. For males, the
direct path between resiliency and substance use, which was significant when age
was not included in the model ($= –.18; p< .05), became non-significant when
age was included in the model ($= .03; ns). This meets the criterion set forth by
Baron and Kenny to conclude that mediation exists when a) the predictor causes
the mediator, b) the mediator causes the outcome measure, and c) the direct path
between the predictor and the outcome measure changes from significant to
non-significant when the mediator is included in the model [35]. However, as the
direct path between risk and substance use was not significant when age was not
included in the model ($= .11; ns), we cannot conclude that mediation exists for
females.
FAMILY RISK AND RESILIENCY FACTORS / 387
Figure 3. Empirical model of risk, resiliency, age of initiation,
and substance use.
3Due to the fact that the predicted model is saturated, it is a just-identified model and does not allow
for goodness-of-fit indices. Fit indices were obtained by fixing the path between age of first use and
substance use to the values specified in the just-identified model and thus, freeing up a degree of
freedom which allowed for the goodness-of-fit testing. Since the models were saturated, all models
discussed below had perfect fit (c2(1) = 0.00, p= 1.00; CFI = 1.00).

388 / MOON, JACKSON AND HECHT
Table 2. Correlations Between Risk, Resiliency, Age of First Use,
and Substance Use
Risk Resiliency
Age of
Initiation
Substance
Use
Full Sample (n= 296)
Risk
Resiliency
Age of initiation
Substance Use
Males (n= 138)
Risk
Resiliency
Age of Initiation
Substance Use
Females (n= 157)
Risk
Resiliency
Age of Initiation
Substance Use
African Americans (n= 35)
Risk
Resiliency
Age of Initiation
Substance Use
Mexican Americans (n= 186)
Risk
Resiliency
Age of Initiation
Substance Use
Anglo Americans (n= 75)
Risk
Resiliency
Age of Initiation
Substance Use
—
–.34**
–.23**
.28**
—
–.44**
–.29**
.38**
—
–.25**
–.19**
.21**
—
–.49**
–.29
.23
—
–.35**
–.25**
.31**
—
–.15
–.10
.22
—
.33**
–.35**
—
.37**
–.31**
—
.29**
–.40**
—
.09
–.28
—
.42**
–.40**
—
.25*
–.28*
—
–.77**
—
–.75**
—
–.79**
—
–.73**
—
–.77**
—
–.80**
—
—
—
—
—
—
*p< .05
**p< .01

To test whether age of first use significantly mediated the relationships between
risk and substance use and between resiliency and substance use, the standard error
for the mediated effect was calculated using the multivariate delta method [36-39].
The mediated effect was then tested for significance using a z test. Tables 3 and 4
present the mediated effects for risk and resiliency, respectively, for males and
females as well as for each ethnic group.
Examination of the tables suggests that age of first use is a much stronger
mediator for the path between resiliency and substance use than for the path
FAMILY RISK AND RESILIENCY FACTORS / 389
Table 3. For the Relationship Between Risk and Substance Use,
the Mediated Effect, Percent Mediated, and Significance of
Mediated Effect for Age of First Use as a Mediator
Subgroup
Mediated
Effect
Total
Effect
%
Mediateda
Standard
Error
Z-Value of
Mediated
Effect
Females
Males
Anglo American
Mexican American
.19
.22
.12
.17
.23
.58
.46
.38
81
38
26
44
.117
.123
.218
.100
1.62
1.79†
0.54
1.68†
aPercent mediated is calculated by dividing mediated effect by total effect.
†p< .10
*p< .05
**p< .01
Table 4. For the Relationship Between Resiliency and Substance Use,
the Mediated Effect, Percent Mediated, and Significance of
Mediated Effect for Age of First Use as a Mediator
Subgroup
Mediated
Effect
Total
Effect
%
Mediateda
Standard
Error
Z-Value of
Mediated
Effect
Females
Males
Anglo American
Mexican American
–.37
–.42
–.41
–.55
.71
.48
.55
.68
52
87
75
81
.113
.129
.197
.110
–3.25**
–3.26**
–2.09*
–5.03**
aPercent mediated is calculated by dividing mediated effect by total effect.
†p< .10
*p< .05
**p< .01

390 / MOON, JACKSON AND HECHT
Figure 4. Model of risk, resiliency, age of initiation, and
substance use for males (top panel) and
females (bottom panel).

between risk and substance use. However, it is important to note that the mediated
effect for risk was marginally significant for males, and falls just short of marginal
significance for females (critical value for z at p< .10 is 1.645). Thus, one might
also conclude that the paths for risk are mediated by age of first use.
The models for the two ethnic groups that had sufficient sample size (Mexican
American and Anglo American)4are presented in Figure 5. Risk, resiliency, and
age accounted for 68 percent of the variance in substance use for Anglo Americans
and 61 percent of the variance in substance use for Mexican Americans. For both
Anglo Americans and Mexican Americans, the path between resiliency and
substance use appears to be mediated by age of first use. The direct paths between
resiliency and substance use, which were significant when age was not included
in the model, $= –.25; p< .05; $= –.34; p< .001 respectively, became
non-significant when age is included in the model, $= –.06; ns; $= –.06; ns;
respectively.5Examination of Tables 3 and 4 assists in interpretation. Age of
first use was again a much stronger mediator for the resiliency path than for
the risk path.
DISCUSSION
The results of this study make an important contribution to our understanding
of risk and resiliency in substance use. A number of overall findings emerge and
will be discussed first. After reviewing these overall findings, we will examine
some of the more specific findings and conclude by considering implications
for prevention.
First, it appears from our data that, as opposed to Newcomb and Felix-Ortiz’s
findings, risk and resiliency constitute separate sets of factors [14]. Their model
assigning high scores on a variable to one set and low scores to the other set did not
fit our data. In addition, we found that resiliency, in general, plays a more
important role than risk in substance use. Resiliency factors such as positive
family relations, low parental permissiveness, and perceived neighborhood safety
had significant effects on all six substance use factors (number of drugs used in
lifetime, age of initiation, substance use in the last month, ever offered, acceptance
of most recent offer, and three of the contexts of drug offer). Risk influenced only
three of the factors (age of initiation, number of drugs used in a lifetime, and one of
the contexts of drug offer). In addition, all of the risk effects interacted with
resiliency. Yet, resiliency had independent effects on use in the last month and on
FAMILY RISK AND RESILIENCY FACTORS / 391
4The small sample size for African Americans (n= 35) contributed to uninterpretable results and
this model will not be discussed.
5The weights reported here are standardized betas. The weights reported in the figures are
unstandardized. Also, please note that for the direct paths between resiliency and substance use when
age was not included in the model, $= –.25, p< .05; $= –.34, p< .05.

392 / MOON, JACKSON AND HECHT
Figure 5. Model of Risk, Resiliency, Age of Initiation, and
Substance Use for Anglo Americans (Top Panel) and
Mexican Americans (Bottom Panel).

acceptance of most recent offer and two of the drug offer contexts, and interacted
with gender and ethnicity in its effect on whether drugs had ever been offered.
Second, our findings indicate that we cannot fully understand the effects of
risk and resiliency without considering gender and ethnicity. Gender, ethnicity,
or their combination interacted with risk and/or resiliency to explain whether
drugs had ever been offered and the context of offers. In particular, we found
that the most offers are received by low resiliency Mexican-American females,
African-American males, and European-American males. In addition, we found
that males in general and particularly those with high risk and low resiliency are
more likely to be offered drugs on the streets whereas low-risk, high-resiliency
females receive their offers in this setting. Low-resiliency females are more likely
to receive their offers at school and, in general, low-resiliency adolescents are less
likely to receive their offers at home. Thus, we need to consider both ethnicity
and gender if we are to understand who gets offered drugs and where those offers
are made.
More importantly, when examining our models very different conclusions
emerge, particularly when considering gender. For males, risk has a significant,
direct effect on substance use while the effect of resiliency is mediated by age of
first use. This indicates that unless resiliency-building interventions are introduced
prior to initiation they will not be maximally effective in reducing male substance
use. Conversely, for females the strong and direct effect of resiliency indicates the
efficacy of interventions at any time, although for females, as well, the mediated
effect of age of first use indicates that intervening prior to experimentation will
maximize impact. These findings indicate that risk plays a much larger role for
males than females and that resiliency has a direct effect on substance use for
females only. Thus, a model that ignores gender differences will produce much
less useful findings than one which considers this factor.
There are a number of explanations for these findings. We know that, in general,
earlier experimentation increases the likelihood of continued use and abuse [40].
But why is risk a significant factor (direct and indirect) for males and not
for females? We know that males have more rigid and restrictive gender-role
requirements [41-42]. Once males start to use drugs at an early age, they may have
more difficulty in stopping due to the lack of role flexibility. Thus, once they get
started with use and this becomes part of their self-identity and peer relationships,
they may not have the skills necessary to change this pattern.
Second, we must ask, “Why do females have both a direct and indirect effect for
resiliency while males have only an indirect effect?” Resiliency in this study has
more to do with family relationships. Girls are socialized to value connections
with others and to participate in relationships in ways that boys are not [43-44],
thus their coping skills and support system may be more highly developed. Thus,
even if girls do start using drugs at an early age, they also have the necessary
resiliency skills and supportive relationships needed to stop. In other words, girls
are more likely to have a backup system, whereas with males it is more likely to be
FAMILY RISK AND RESILIENCY FACTORS / 393

a “one-shot deal.” However, males may have other resiliency skills other than
family variables and these need to be examined in future research.
The similarity of the models for Mexican Americans and European Americans
suggests that ethnicity effects may be more salient for African Americans. More
research is needed to clarify these relationships as sample size precluded a
thorough examination of this effect.
Limitations
While family risk and resiliency factors provided a powerful predictor of
adolescent substance use, future research can improve this model by identifying
other resiliency factors, particularly those that may be more salient for males. In
this study, we selected family risk and resiliency factors due to their salience in
the lives of adolescents at this stage of development. One factor, androgyny,
may serve as a potential protective factor for both males and females [12].
Androgynous people have elements of both masculine and feminine gender
orientations including a wider repertoire of available skills which they can draw
upon in a variety of situations such as those in which drugs are offered. Other
risk factors might be included, such as difficult childhood mood temperament
(e.g., negative mood swings, temper tantrums), childhood emotional distress
(e.g., depression), behavior problems (e.g., hyperactivity, poor impulse control),
antisocial behavior, school failure, low achievement, low school commitment,
relationships with drug-using peers, positive attitudes towards drugs, alienation,
while additional resiliency factors might include self-efficacy, realism, problem-
solving skills, personal goals, empathy, humor, and parental supervision [2].
Gender-specific resiliency factors also might be considered, such as positive male
role models, greater structure, and parental supervision for males, and absence of
parental over-supervision, independence, and support from primary caretaker for
females [2].
A few important notes of caution must be applied to these findings. First, we
considered a limited range of risk and resiliency factors. While these factors
accounted for a great deal of variance in substance use, a more complete model
is needed to fully support these conclusions. In addition, these effects may
be specific to the early adolescent years because the family exerts its greatest
influence over a child’s emotional, social, psychological, and physical environ-
ment during formative years including childhood and early adolescence [5]. For
this age group, the lack of protective factors may be qualitatively more important
in determining whether a child will use drugs than other factors. Second, we
empirically derived our risk and resiliency constructs, which raises the needs
for replication of these results using a different sample. Finally, sample size
was limited by the length of the instrument and restrictions imposed by the
participating schools. Unfortunately, this may have attenuated the effects of risk
394 / MOON, JACKSON AND HECHT

by eliminating those with the highest risk who did not finish the instrument and
were more likely to be absent on a single administration.
CONCLUSION
In this study, risk and resiliency constitute separate sets of factors, and overall,
resiliency factors played a more significant role in terms of substance use and
resistance processes. In fact, while risk factors such as family stressor events,
perceived neighborhood safety, and parental alcohol use did not significantly
predict substance use in the past month, whether or not drugs had been offered, and
whether they accepted the most recent offer, we found that the lack of protective
factors, such as satisfying family relationships, low parental permissiveness, and
involvement in religion, to be highly predictive of these variables. Both risk and
resiliency seem to affect the age of initiation and the number of drugs offered over
their lifetime. Adolescents low in resiliency and those high in risk were likely to
have started using drugs at earlier ages. In addition, those low in resiliency also
tended to be offered drugs in more public arenas such as schools or parks whereas
those low in resiliency were least likely to be offered drugs in the relative privacy
of their homes.
It appears that strengthening and increasing protective factors needs to be a
primary focus in prevention and intervention efforts with this age group to reduce
substance use, especially at early ages. The current research suggests that early
intervention is particularly important for males. Intervention efforts may need to
include skills training specific to these very different contexts. It may be that
adolescents with less parental monitoring are more likely to be in public arenas
unsupervised and, therefore, more at risk. These findings continue to support the
efficacy of examining resiliency or protective factors as a vehicle to increase
our understanding of substance use and abuse and in the design of effective
interventions.
REFERENCES
1. J. D. Hawkins, R. F. Catalano, and J. Y. Miller, Risk and Protective Factors for Alcohol
and Other Drug Problems in Adolescence and Early Adulthood: Implications for
Substance Abuse Prevention, Psychological Bulletin, 112:1, pp. 64-105, 1992.
2. E. Norman, Personal Factors Related to Substance Misuse: Risk Abatement and/or
Resiliency Enhancement? in Substance Abuse in Adolescence, T. Gullotta, G. Adams,
and R. Montemayor (eds.), Sage, Thousand Oaks, California, pp. 15-35, 1995.
3. M. Felix-Ortiz and M. D. Newcomb, Risk and Protective Factors for Drug Use
among Latino and White Adolescents, Hispanic Journal of Behavioral Sciences, 14:3,
pp. 291-309, 1992.
4. M. D. Newcomb, Drug Use Etiology Among Ethnic Minority Adolescents: Risk and
Protective Factors, in Drug Abuse Prevention With Multiethnic Youth, G. J. Botvin,
S. Schinke, and M. A. Orlandi (eds.), Sage, Thousand Oaks, California, pp. 105-129,
1995.
FAMILY RISK AND RESILIENCY FACTORS / 395

5. S. Turner, Family Variables Related to Adolescent Substance Misuse: Risk and
Resiliency Factors, in Substance Misuse in Adolescence, T. P. Gullotta, G. R. Adams,
and R. Montemayor (eds.), Sage, Thousand Oaks, California, pp. 36-55, 1995.
6. M. Rutter and H. Giller, Juvenile Delinquency: Trends and Perspectives, Penguin,
New York, 1983.
7. M. Rutter, Protective Factors in Children’s Responses to Stress and Disadvantage,
in Primary Prevention of Psychopathology: Vol. 3, Social Competence in Children,
M. W. Kent and J. Rolf (eds.), University Press of New England, Hanover, New
Hampshire, pp. 49-74, 1979.
8. K. L. Kumpfer, Resiliency and Use Prevention in High Risk Youth, Unpublished
manuscript, 1993. [Available from School of Social Work, University of Utah,
Salt Lake City, UT, 84112.]
9. E. Werner, Vulnerability and Resiliency in Children at Risk for Delinquency: A
Longitudinal Study from Birth to Young Adulthood, in Prevention of Delinquent
Behavior, J. Burchard and S. Burchard (eds.), Sage, Newbury Park, California,
pp. 16-43, 1987.
10. B. Benard, The Case for Peers, Western Center for Drug-Free School and Com-
munities, Portland, Oregon, 1990.
11. J. Brook, A. S. Gordon, M. Whiteman, and P. Cohen, Some Models and Mechanisms
for Explaining the Impact of Maternal and Adolescent Characteristics on Adolescent
Stage of Drug Use, Developmental Psychology, 22, pp. 460-467, 1986.
12. E. Werner and R. S. Smith, Vulnerable But Invincible, McGraw-Hill, New York,
1982.
13. D. L. Thomas and C. Carver, Religion and Adolescent Social Competence, in Devel-
oping Social Competency in Adolescence, T. Gullotta, G. Adams, and R. Montemayor
(eds.), Sage, Newbury Park, California, pp. 195-213, 1990.
14. M. D. Newcomb and M. Felix-Ortiz, Multiple Protective and Risk Factors for Drug
Use and Abuse: Cross-Sectional and Prospective Findings, Journal of Personality and
Social Psychology, 63:2, pp. 280-296, 1992.
15. M. D. Newcomb and P. M. Bentler. Consequences of Adolescent Drug Use: Impact on
the Lives of Young Adults, Sage, Beverly Hills, California, 1988.
16. J. G. Bachman, J. M. Wallace, P. M. O’Malley, L. D. Johnston, C. L. Kurth, and
H. W. Neighbors, Racial/Ethnic Differences in Smoking, Drinking, and Illicit Drug
Use among American High School Seniors, 1976-89, American Journal of Public
Health, 81, pp. 372-377, 1991.
17. R. Caetano and J. Schafer, DSM-IV Alcohol Dependence and Drug Abuse/
Dependence in a Treatment Sample of Whites, Blacks, and Mexican Americans, Drug
and Alcohol Dependence, 43, pp. 93-101, 1996.
18. L. D. Johnston, P. M. O’Malley, and J. G. Bachman, National Survey Results on Drug
Use From the Monitoring The Future Study, 1975-1995. Volume I: Secondary School
Students, (NIH Pub. No. 97-4139), National Institute on Drug Abuse, Rockville,
Maryland, 1996.
19. K. L. Kumpfer and C. W. Turner, The Social Etiology Model of Adolescents: Impli-
cations for Prevention, International Journal of the Addictions, 25:4a, pp. 435-452,
1991.
20. M. D. Newcomb and P. M. Bentler, Substance Use and Ethnicity: Differential Impact
of Peer and Adult Models, Journal of Psychology, 120, pp. 83-95, 1986.
396 / MOON, JACKSON AND HECHT

21. M. A. Pentz, E. A. Trebow, and W. B. Hansen, Effects of Program Implementation on
Adolescent Drug Use Behavior: The Midwestern Project, Evaluation Review, 14,
pp. 264-289, 1990.
22. G. J. Warheit, W. A. Vega, E. L. Khoury, A. A. Gil, and P. H. Elfenbein, A Com-
parative Analysis of Cigarette, Alcohol, and Illicit Drug Use among an Ethnically
Diverse Sample of Hispanic, African American, and Non-Hispanic White Adolescents,
Journal of Drug Issues, 26, pp. 901-922, 1996.
23. J. R. Hughes, L. E. Day, R. J. Marcantonio, and E. Torpy, Gender Differences in
Alcohol and Other Drug Use Among Young Adults, Substance Use and Misuse, 32,
pp. 317-342, 1997.
24. D. B. Kandel, J. G. Johnson, H. R. Bird, G. Canino, S. H. Goodman, B. B. Lahey,
D. A. Regier, and M. Schwab-Stone, Psychiatric Disorders Associated with Substance
Use among Children and Adolescents: Findings From the Methods for the Epidemi-
ology of Child and Adolescent Mental Disorders (MECA) Study, Journal of Abnormal
Child Psychology, 25, pp. 121-132, 1997.
25. N. A. Gonzales, M. L. Gunnoe, K. M. Jackson, and R. Samaniego, Validation
of a Multicultural Events Scale for Urban Adolescents: Preliminary Strategies for
Enhancing Cross-Ethnic and Cross-Language Equivalence, Unpublished manuscript,
1995.
26. T. Dielman, A. Butchart, and J. Shope, Structural Equation Model Test of Patterns in
Family Interaction, Peer Alcohol Use, and Intrapersonal Predictors of Adolescent
Alcohol Use and Misuse, Journal of Drug Education, 23:3, pp. 273-316, 1993.
27. T. Field and R. Yando, The Adolescent’s Self-Perception Scales, Unpublished
manuscript, 1991.
28. J. Cohen and P. Cohen, Applied Multiple Regression/Correlation Analyses for the
Behavioral Sciences (2nd Edition), Lawrence Erlbaum, Hillsdale, NJ, 1983.
29. D. J. Flannery, A. T. Vazsonyi, and J. Torquati, Ethnic and Gender Differences in Risk
for Early Adolescent Substance Use, Journal of Youth and Adolescence, 23, p. 195,
1994.
30. J. K. Alberts, M. L. Hecht, M. Miller-Rassulo, and R. L. Krizek, The Communicative
Process of Drug Resistance among High School Students, Adolescence, 27,
pp. 203-226, 1992.
31. R. L. Krizek, M. L. Hecht, and M. Miller, Language as an Indicator of Risk in the
Prevention of Drug Use, Journal of Applied Communication Research, 2, pp. 245-262,
1993.
32. D. G. Moon, M. L. Hecht, K. M. Jackson, and R. Spellers, Ethnic and Gender
Differences and Similarities in Adolescent Drug Use and Refusals of Drug Offers,
Substance Use and Misuse, 34, pp. 1059-1083, 1999.
33. L. S. Aiken and S. G. West, Multiple Regression: Testing and Interpreting Inter-
actions, Sage, Newbury Park, California, 1991.
34. P. M. Bentler, Theory and Implementation of EQS: A Structural Relations Program,
BMDP Statistical Software, Los Angeles, 1989.
35. R. M. Baron and D. Kenny, The Moderator-Mediator Distinction in Social Psycho-
logical Research: Conceptual, Strategic, and Statistical Considerations, Journal of
Personality and Social Psychology, 51, pp. 1173-1182, 1986.
36. D. P. MacKinnon, Analysis of Mediating Variables in Prevention and Intervention
Research, in Scientific Methods for Prevention Intervention Research, A. Cazeres and
FAMILY RISK AND RESILIENCY FACTORS / 397

L. A. Beatty (eds.), NIDA Research Monograph 139, DHHS Pub. No. 94-3631,
Superintendent of Documents, U.S. Government Printing Office, Washington, D.C.,
pp. 127-153, 1994.
37. D. P. MacKinnon and J. H. Dwyer, Estimating Mediated Effects in Prevention Studies,
Evaluation Review, 17, pp. 144-158, 1993.
38. M. E. Sobel, Asymptotic Confidence Intervals for Indirect Effects in Structural
Equations Models, in Sociological Methodology, S. Leinhardt (ed.), American
Sociological Association, Washington, D.C., pp. 290-293, 1986.
39. M. E. Sobel, Some New Results on Indirect Effects and Their Standard Errors in
Covariance Structure Models, in Sociological Methodology, N. Tuma (ed.), American
Sociological Association, Washington, D.C., pp. 159-186, 1986.
40. L. Robins and T. Przybeck, Age of Onset of Drug Use as a Factor in Drug and Other
Disorders, in Etiology of Drug Abuse: Implications for Prevention, C. Jones and
R. Batjes (eds.), National Institute on Drug Abuse, Rockville, Maryland, 1985.
41. J. R. Bardewell, S. W. Cochran, and S. Walker, Relationship of Parental Education,
Race, and Gender to Sex-Role Stereotyping in 5-Year Old Kindergartners. Sex Roles,
15, pp. 275-281, 1986.
42. S. Feinman, A Status Theory of the Evaluation of Sex-Role and Age-Role Behavior,
Sex Roles, 10, pp. 445-456, 1984.
43. E. J. Aries and F. L. Johnson, Close Friendship in Adulthood: Conversational Content
Between Same-Sex Friends, Sex Roles, 9, pp. 1183-1196, 1983.
44. P. H. Wright and M. B. Scanlon, Gender Role Orientations and Friendship: Some
Attenuation but Gender Differences Still Abound, Sex Roles, 24, pp. 551-566, 1991.
Direct reprint requests to:
Dreama Moon, Ph.D.
California State University, San Marcos
Communication Department
San Marcos, CA 92096-0001
398 / MOON, JACKSON AND HECHT