AFRICAN AMERICAN ADOLESCENTS’ PARENTAL, PEER, AND PARTNER
RELATIONSHIPS AND SEXUAL RISK
By
Mercedes M. Morales-Alemán

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Psychology
2011

ABSTRACT
AFRICAN AMERICAN ADOLESCENTS’ PARENTAL, PEER, AND PARTNER
RELATIONSHIPS AND SEXUAL RISK
By
Mercedes M. Morales-Alemán
African American youth are disproportionately affected by STDs, including HIV.
Empirical evidence suggests that distinct aspects of parent, peer, and sexual partner relationships
are important influences on adolescent sexual risk taking. Research has consistently shown that
adolescents who report involved parenting (including a high quality relationship and supervision)
are significantly less likely to engage in sexual risk taking (DiClemente et al., 2001; Henrich,
Brookmeyer, Shrier, & Shahar, 2006; Huebner & Howell, 2003; Li, Stanton, & Feigelman, 2000;
Rodgers, 1999; Stanton et al., 2002; Sturdivant, 2007), while adolescents with risky friends are
considerably more likely to be risky themselves (Bearman & Bruckner, 1999; Lyons, 2009;
Miller, Forehand, & Kotchick, 2000; Rai et al., 2003; Romer, Black, Ricardo, Feigelman, & et
al., 1994). Additionally, some research has found that the level of risk of a person’s sexual
partner(s) is more predictive of negative sexual health outcomes than his or her own previous
behavior (Staras, Cook, & Clark, 2009). However, research has not yet examined whether
involved parenting can protect youth who have sexually risky peers and/or sexually risky
partners from sexual risk taking (i.e., lack of condom use). This study sought to determine
whether involved parenting was a significant buffer against high risk sex for adolescents whose
peers were sexually risky and whether involved parenting was a significant buffer against high
risk sex for adolescents whose partners were sexually risky. It also sought to determine whether
gender moderated the relationship between (1) peer sexual risk and sexual risk taking and (2)
partner sexual risk and risk taking.

This study used data drawn from the Bayview Network Study (BNS), a longitudinal study
conducted in San Francisco, CA from 2000-2002. Recruitment for the BNS was a combination
of population-based random sampling and snowball sampling of friends and sexual partners.
Youth were considered eligible if they identified as African American, were between the ages of
14-19, and lived in the Bayview area. Data from 2 waves (12 months apart from one another)
were analyzed. Structural equation modeling was used to examine a conceptual model of parent,
peer, and partner influences on sexual risk taking and STD diagnosis. Data from 199 index
participants and their social friends were analyzed.
Results showed that more than half of the youth reported inconsistent use of condoms
within the last 3 months at both waves. Additionally, 18% of the sample either reported a past
STD or had a positive result on the STD test that they took as part of the study. Structural
equation modeling analyses demonstrated that involved parenting did not have the predicted
effect on sexual risk taking in this sample of youth. However, (1) increased peer risk approached
significance in predicting decreased condom use behaviors and (2) riskier partners were a
significant predictor of sexual risk taking. Moderated analyses of the effects of involved
parenting on the relationship between peer risk and condom use showed that involved parenting
did not buffer the impact of sexually risky peers on decreased condom use. Similarly, moderated
analyses of involved parenting on the relationship between sexually risky partners and condom
use showed that involved parenting was not protective against sexual risk taking. Finally, gender
was not a significant moderator of the effects of any of the aforementioned relationships. Study
results suggest that although involved parenting was not an effective protective factor for high
risk sex in this population, peers and partners may be important points of intervention in
preventing sexual risk behaviors in African American adolescents.

Copyright by
MERCEDES M. MORALES-ALEMÁN
2011

ACKNOWLEDGEMENTS
This work would not have been possible without the support of my mentors, colleagues,
family and friends. First and foremost, I would like to thank to my dissertation committee
members Drs. Robin L. Miller, Jonathan Ellen, Deborah Bybee, Deborah Johnson and Jennifer
Watling Neal. I am very grateful to each of you for your guidance. Dr. Robin L. Miller, thank
you for your mentorship and instruction. Reading your work as an undergraduate student at the
University of Puerto Rico inspired me to become a community psychologist. It has truly been an
honor to work with you. Dr. Ellen, thank you for granting me the opportunity to work with the
Bayview Network Study data. I have grown a great deal as a researcher and feel privileged to
have had this chance. Dr. Bybee, thank you for your kindness and instruction with the
methodology of this study. It was wonderful to learn from you. Dr. Johnson, thank you for
offering your expertise and guidance. I truly appreciate your feedback and time. And Dr. Neal,
thank you for your attention to detail and encouragement. It was a pleasure to work with you.
I would also like to thank my Eco colleagues, Dr. Cidhinnia Torres-Campos and Dr.
Sinead Younge, for their mentorship. I am fortunate to have ―learned the ropes‖ from such
incredible women. Drs. Sheila F. LaHousse, Marisa L. Beeble, Lauren F. Lichty and (soon to be
Dr.) Jason Forney thank you for the countless hours of feedback, advice, and intellectual
community. I would also like to thank Dr. Steven Pierce for his patience, guidance and
encouragement in my moments of statistics-related anxiety.
I would also like to acknowledge the American Psychological Association’s Minority
Fellowship Program, the Michigan State University Psychology Department and the Graduate
School. I could not have completed this work without their generous support.

v

I also must express my gratitude to the women who helped me begin this journey. Dr.
Milagritos González, thank you for introducing me to community-based research. Dr. Deborah
Salem, thank you for providing the tools and encouragement necessary to complete the first
steps.
Finally, I would like to thank my family. Thank you to my amazing parents, Dr. José R.
Morales and Mrs. Edlyn M. Rodríguez, for never doubting that I could accomplish this. On the
worst days, you always made me feel cherished and capable. Thank you for teaching me to
always do my best, help others, and pursue what I love. I am eternally grateful to my husband,
Carlos E. Alemán. I cannot begin to express what your support has meant to me. I am blessed to
have such an amazing teammate. We did it!

vi

TABLE OF CONTENTS
LIST OF TABLES ......................................................................................................................... ix
LIST OF FIGURES ........................................................................................................................ x
INTRODUCTION .......................................................................................................................... 1
Understanding Sexual Risk in Adolescent Populations .............................................................. 1
Review of the Literature ............................................................................................................. 4
Normative Adolescent Sexual Involvement ........................................................................... 5
Ecodevelopmental Theory ...................................................................................................... 6
Multilevel Approach ........................................................................................................... 7
Developmental Perspective ................................................................................................. 8
Understanding Interactional Processes ............................................................................... 9
Ecodevelopmental Theory: Previous Research ............................................................ 10
Ecodevelopmental Perspective on Familial, Peer and Partner Microsystems ...................... 12
Parents and Families ......................................................................................................... 12
Peers .................................................................................................................................. 15
Sexual Partners.................................................................................................................. 16
Adolescent Parent Relationships, Peer Relationships and Sexual Risk: An Interactions
Approach ........................................................................................................................... 17
Do family and peer systems interact to influence boys and girls similarly? ................ 20
Adolescents’ Parent Relationships, Sexual Partners, and Sexual Risk: Considerations on
interactions and gender ..................................................................................................... 23
Studying Social Influences on Sexual Risk Taking: Bridging the Gap between Theory and
Research Methodology ............................................................................................................. 25
Hypotheses and Research Questions ........................................................................................ 26
Research Questions ............................................................................................................... 26
Exploratory Questions ...................................................................................................... 27
Hypotheses ........................................................................................................................ 27
METHOD ..................................................................................................................................... 30
Research Design........................................................................................................................ 30
Participants ................................................................................................................................ 30
Procedure .................................................................................................................................. 31
Recruitment ........................................................................................................................... 31
Consent and Confidentiality ................................................................................................. 32
Survey Administration .......................................................................................................... 33
Measures ................................................................................................................................... 34
Demographics ....................................................................................................................... 34
Involved Parenting ................................................................................................................ 34
Parental Monitoring .......................................................................................................... 35
Parental Support ................................................................................................................ 35
Sexual Risk: Condom Use ................................................................................................... 36
Sexually Risky Peers............................................................................................................. 37
Sexually Risky Partners ........................................................................................................ 38
vii

Perception of STD and HIV Risk ......................................................................................... 40
STD Status ............................................................................................................................ 40
Control Variables and Preliminary Analyses ........................................................................ 41
Missing Data, Normality, and Outliers ..................................................................................... 45
Missing Data and Outliers .................................................................................................... 45
Data Analytic Strategy .............................................................................................................. 46
Model Fitting Process ........................................................................................................... 47
RESULTS ..................................................................................................................................... 50
Confirmatory Factor Analysis: Involved Parenting as a Latent Construct .............................. 51
Path Analysis Model Results by Research Question ................................................................ 57
RQ #1: Is involved parenting a significant buffer against high risk sex for adolescents
whose peers are sexually risky? ............................................................................................ 58
RQ #2: Does gender moderate the relationship between peer sexual risk and sexual risk
taking? ................................................................................................................................... 63
RQ #3: Does higher risk taking predict higher incidence of STD diagnosis? ..................... 68
Exploratory Questions .............................................................................................................. 70
EQ #1: Is involved parenting a significant buffer against high risk sex for adolescents
whose partners are typically sexually risky? ........................................................................ 70
EQ #2: Does gender moderate the relationship between risky sexual partners and sexual
risk taking? ............................................................................................................................ 76
EQ #3: Does perception of risk moderate the relationship between risky sexual partners and
sexual risk taking?................................................................................................................. 81
DISCUSSION ............................................................................................................................... 83
Theoretical Implications ........................................................................................................... 93
Implications............................................................................................................................... 96
Limitations .............................................................................................................................. 100
Future Research ...................................................................................................................... 100
Conclusion .............................................................................................................................. 102
APPENDIX A ............................................................................................................................. 104
Items in Scales ........................................................................................................................ 104
APPENDIX B ............................................................................................................................. 109
IRB Documentation ................................................................................................................ 109
APPENDIX C ............................................................................................................................. 111
Correlations Tables ................................................................................................................. 111

viii

LIST OF TABLES

Table 1. Participant Demographic Characteristics: Participants with 1 or 2 friends who were
interviewed versus participants with no friends who were interviewed …..…….……….......44-45
Table 2. Confirmatory Factor Analysis: Model Fit Statistics ….………..…….……..........…….53
Table 3. Peer Sexual Risk Model: Unstandardized Parameter Estimates and Significance
Values......……………..………………………………………………...………………….……61
Table 4. Chi-Square Test for Involved Parenting and Peer Risk Model: Testing significance of
interaction between involved parenting and sexually risky peers .…………………………..….63
Table 5. Chi-Square Test for Involved Parenting and Peer Risk Model: Testing significance of
interaction between sexually risky peers and gender …………………………………...……….66
Table 6. Chi-Square Test for Involved Parenting and Peer Risk Model: Testing significance of
interaction between involved parenting and gender ……...…………………………………..…68
Table 7. Sexually Risky Partners Model: Unstandardized Parameter Estimates and Significance
Values …………..…………………………………………………….………………...……71-72
Table 8. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing
significance of interaction between involved parenting and risky sexual partner….….………...76
Table 9. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing the
significance of the interaction between risky sexual partners and gender ………………………79
Table 10. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing
the significance of the interaction between involved parenting and gender ………………….…81
Table 11. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing
significance of interaction between risky sexual partners and perception of
Risk…………………………………………………..………………………………………..…82
Table 12. Bivariate Correlations for all Study Variables………………………...………..113-118

ix

LIST OF FIGURES

Figure 1. Conceptual Model of Parent, Peer, and Partner Influences on Sexual Risk
Taking ……………………………...………………………….…………...………...………….29
Figure 2. Confirmatory Factor Analysis of Involved Parenting as a Latent Construct …………52
Figure 3. Confirmatory Factor Analysis of Involved Parenting: Final Model (Standardized
Parameter Estimates)…………………………………………………………………………….55
Figure 4. Structural Equation Model of Involved Parenting and Peer Sexual Risk: Interaction
Term Excluded (Standardized Parameter Estimates) ………………………...…………………59
Figure 5. Structural Equation Analysis of Involved Parenting and Peer Sexual Risk: Gender
Moderation Analyses (Sexually Risky Peers X Gender) ……………….………….….………...65
Figure 6. Structural Equation Analysis of Involved Parenting and Peer Sexual Risk: Gender
Moderation Analyses…………………..………………..……………………………………….67
Figure 7. Structural Equation Model of Involved Parenting and Sexually Risky Peers:
Mediational Analysis ……..………….………………………………………………………….69
Figure 8. Structural Equation Model of Involved Parenting and Risky Sexual Partners:
Interaction Terms Excluded (Standardized Parameter Estimates)………………...……………..73
Figure 9. Structural Equation Model of Involved Parenting and Risky Sexual Partners: Gender
Moderation Analyses (Risky Sexual Partners X Gender) …..……………………….…………..78
Figure 10. Structural Equation Model of Involved Parenting and Risky Sexual Partners: Gender
Moderation Analyses …..………………………………………………………………………..80

x

INTRODUCTION
African American adolescents are disproportionately affected by STDs and HIV in the
United States (Centers for Disease Control and Prevention, 2008a). According to data from the
National Youth Risk Behavior Survey (YRBS) 67.6% of African American students report ever
having had sexual intercourse and about 31.1% report that they did not use a condom the last
time they had sex (Centers for Disease Control and Prevention, 2006). In relation to HIV/AIDS,
the disparities are striking. In 2006, 70% of HIV/AIDS diagnoses in 13 to 19 year olds occurred
in African Americans, although this group accounts for only 17% of adolescents 13 to 19 years
of age (Centers for Disease Control and Prevention, 2008b). African American adolescents are
also disproportionately affected by other STDs (Centers for Disease Control and Prevention,
2008, March). In 2006, the rates of Chlamydia infection and Gonorrhea were much higher for
African American adolescents than for White adolescents (7.5 times greater for Chlamydia and
4.4 times greater for Gonorrhea) (Centers for Disease Control and Prevention, 2006).
Understanding Sexual Risk in Adolescent Populations
Given the serious health consequences associated with STDs and HIV (including
infertility, chronic pain, cancer, and death) and the high costs of treatment (Cates, Herndon,
Schulz, & Darroch, 2004) prevention has long been considered very important in reducing the
public health impact of these epidemics. In order to develop effective prevention efforts, it is
crucial to conduct research that will uncover the antecedents and mechanisms associated with
sexual risk behaviors in adolescent populations (DiClemente, Salazar, & Crosby, 2007).
Understanding these epidemics is the first step to effective prevention.

1

Until recently, most of the research on sexual risk in adolescent and other populations
explored individual level factors related to sexual risk taking and therefore focused on the
psychological or cognitive determinants of risk behaviors (DiClemente, et al., 2007). This is
largely due to the fact that, historically, most sexual health research has been guided by
individual-level theories of behavior (e. g., Bandura’s self-efficacy theory, Prochaska’s
Transtheoretical Model of Behavior, and Azjen & Fishbein’s Theory of Reasoned Action) which
focus on the importance of cognitive determinants of risk behaviors such as knowledge, selfefficacy, intentions and attitudes. Studying sexual risk from a solely individual perspective is
problematic, however, because it ignores the larger social context in which adolescents are
embedded (Hobfoll, 1998). Youth live and develop within complex social networks and their
relationships play a crucial role in their sexual behavior (Perrino, Gonzalez-Soldevilla, Pantin, &
Szapocznik, 2000; Szapocznik & Coatsworth, 1999).
In fact, certain types of parent-child relationships have been linked with lower levels of
sexual risk in adolescents (DiClemente, et al., 2007). Specifically, adolescents who report that
their parents monitor them (i.e., know where they are at night, have a curfew etc.) and support
them (i.e., are close to them and care about them etc.) are considerably less likely to report
sexual risk than their counterparts who do not report high parental monitoring and support
(DiClemente, et al., 2001; Henrich, et al., 2006; Huebner & Howell, 2003; Li, Stanton, et al.,
2000; Miller, et al., 2000; Rodgers, 1999; Stanton, et al., 2002; Sturdivant, 2007; Voisin,
DiClemente, Salazar, Crosby, & Yarber, 2006). In addition to parents, adolescents’ peers have a
strong influence on sexual behavior (Bearman & Bruckner, 1999; East, Khoo, & Reyes, 2006;
Scaramella, Conger, Simons, & Whitbeck, 1998). Teenagers who perceive their peers to be
sexually active and sexually risky (i.e., engage in unprotected sex) are more likely to engage in

2

these activities themselves (Lyons, 2009; Miller, et al., 2000; Rai, et al., 2003; Romer, et al.,
1994). Teenagers whose peer groups are sexually risky (i.e., low condom use, multiple sex
partners etc.) are also more likely to be risky themselves (Bearman & Bruckner, 1999; Henry,
Schoeny, Deptula, & Slavick, 2007). Finally, characteristics of adolescents’ sexual partners have
been found to be closely related to sexual risk taking in youth (Auerswald, Muth, Brown, Padian,
& Ellen, 2006; Staras, et al., 2009). Overall, these relationships have been well established by
research.
However, youths’ social environments are fluid and interact with one another
(Szapocznik & Coatsworth, 1999). As researchers, we don’t yet understand how parent, peer,
and partner influences come together to impact sexual risk. It is important to explore how
different aspects of adolescents’ social systems interact with one another to affect sexual risk
taking in youth.
Additionally, we do not yet understand whether these processes play out in similar or
dissimilar ways in boys versus girls. Gender norms and sexual socialization work differently
depending on a child’s gender. Therefore, there is reason to believe that the ways in which
parent, peer, and partner characteristics would impact sexual risk taking might be different for
boys than they are for girls. Finally, as will be discussed in the review of the literature, a great
majority of the research focusing on social relationships’ impact on sexual health falls short of
using methods that reflect a shift toward a contextualized perspective (Luke, 2005). The tools
and techniques that psychologists have historically used in the collection and analysis of data are
not necessarily the best suited for understanding social influences on behaviors (Luke, 2005),
including health-related behaviors. This dissertation utilized multi-informant data (i.e., data

3

provided by participants and their peers), allowing for a more contextualized understanding of
social influences on sexual risk taking.
This research sought to achieve three main goals: (1) understand if and how parent, peer
and partner influences interact with one another to impact sexual risk taking (i.e., condom use)
and sexual health outcomes (i.e., STD) in African American youth (2) explore what effect, if
any, a child’s gender has on these relationships and (3) attempt to bridge the gap between theory
and methodology by utilizing methods that reflect a contextual perspective.
Review of the Literature
The following sections include the foundational literature on which this study is based.
Firstly, normative adolescent sexual involvement is described with the purpose of providing
some background about what it means to be a teenager in the United States in relation to sex
initiation and condom use. Secondly, the Ecodevelopmental Theory is described. The
Ecodevelopmental Theory is the theoretical basis for this study (Szapocznik & Coatsworth,
1999). This theory is well suited for this research given that it was developed with the purpose
of understanding how extra-individual influences can impact development of risk trajectories (or
protective trajectories) over time. This work utilized it as an organizing framework with which
to understand parent, peer, and partner influences in relation to sexual risk taking in African
American youth. Thirdly, the literature on parental, peer and partner influences and how they
interact to impact sexual risk taking in adolescents will be discussed, including gendermoderated effects of these interactions. Major points of consensus will be explored and issues
that warrant further research will be addressed in detail.

4

Normative Adolescent Sexual Involvement
In order to understand sexual risk behaviors in adolescents, it is important to first get a
sense of what normative sexual behavior consists of in American youth. Overall, sexual
initiation during adolescence is very common. The average age of sexual initiation in the United
States is 17.3 for young women and 17.0 for young men (New Strategist Publications, 2007).
However, when examined more closely, it becomes clear that many youth initiate sexual
intercourse at much earlier ages. According to data from the Youth Risk Behavior Survey
(2007), 38.1% of males and 27.4% of females report that they initiated sexual intercourse by 9th
th

grade (about 14-15 years old). By 12 grade, 52.8% of males and 66.2% of females report that
they have had sexual intercourse (about 17-18 years old) (Centers for Disease Control and
Prevention, 2009). In relation to condom use, boys and younger adolescents are more likely to
report condom use at last intercourse than girls and older adolescents (Centers for Disease
th

Control and Prevention, 2009). Although 75.8% of 9 grade boys report condom use at last
th

sexual intercourse, only 59.6% of 12 grade boys report condom use at last intercourse (Centers
th

for Disease Control and Prevention, 2009). Similarly, 61% of 9 grade girls report condom use
th

at last sex whereas only 49.9% of 12 grade girls report condom use at last sex (Centers for
Disease Control and Prevention, 2009).
On average, African American youth report earlier sexual initiation than other ethnic
groups. However, African American youth are also more likely to report condom use (Centers
for Disease Control and Prevention, 2009). Average age of first intercourse for African
American youth is about 16.3 years for females and 15.5 for males (New Strategist Publications,
2007). Approximately 67% of African American high school students (compared to 43.7% of
5

White students and 52% of Hispanic students) report having had intercourse (Centers for Disease
Control and Prevention, 2009). However, out of these sexually active youth, 67.3% report using
condoms at last intercourse (compared to 61.4% of White youth and 61.4% of Hispanic youth)
(Centers for Disease Control and Prevention, 2009).
Generally speaking, sexual initiation is normative within the context of adolescence.
Unfortunately, low levels of condom use are very common as well. This dissertation focused
specifically on sexual risk taking (i.e., condom use) instead of sexual initiation. A great deal of
the research in the field of adolescent sexual health has focused on younger adolescents and how
different factors relate to early initiation. Although sexual initiation can be problematic for
younger teens, especially given its link to later instances of sexual risk taking (O'Donnell,
O'Donnell, & Stueve, 2001), this may not be the case for older teens. Sexual initiation is, in fact,
a normative part of most adolescents’ sexual development trajectories. The more relevant
question, in relation to HIV and STD risk is what factors are related to sexual risk taking (i.e.,
lack of condom use) and protective behaviors in teens. A focus on condom use is warranted
given that (1) as outlined above, many teens do not use them consistently and (2) among
contraceptive devices, only male and female condoms are able to effectively prevent HIV and
(most) STD infections.
Developing a deeper understanding of which factors relate to sexual risk taking in
adolescents is crucial in order to create appropriate interventions, implement effective policies,
and inform practitioners in the field (DiClemente, et al., 2007).
Ecodevelopmental Theory
Theoretical frameworks are important tools in achieving a more complete understanding
of sexual risk taking in adolescents. One theory which has focused on organizing the knowledge
6

of how social influences impact sexual risk is the Ecodevelopmental Theory (Szapocznik &
Coatsworth, 1999).
The Ecodevelopmental Theory was developed from several decades of empirical research
on family therapy and family-based interventions conducted at the University of Miami, Center
for Family Studies (Szapocznik & Coatsworth, 1999). This theory integrates concepts from: (1)
Bronfenbrenner’s (1979) social ecology model of human development, (2) a lifespan
developmental approach (Baltes, Reese, & Nesselroade, 1977), (3) structural family theory
(Minuchin, 1974), and (4) multisystemic interventions (Henggeler & Bourdin, 1990).
Ecodevelopmental Theory attempts to build upon these models to create a framework that
represents a ―next step‖ in the study of adolescent risk and protective behaviors (Szapocznik &
Coatsworth, 1999).
Three guiding perspectives inform how this theory represents risk and protective factors: (1)
a multilevel approach, (2) a developmental perspective, and (3) an understanding of interactional
processes.
Multilevel Approach
Firstly, Ecodevelopmental Theory draws upon the central concept of Social Ecology Theory,
that the individual is nested within a set of structures that are, in turn nested inside other
structures like Russian dolls (Bronfenbrenner, 1979). Within Social Ecology Theory, four
primary systems are identified: microsystems (settings in which the adolescent participates
directly, like the family), mesosystems (interactions amongst microsystems such as between
family members and peers), exosystems (systems independent of the adolescent that may
influence her or him indirectly such as a parent’s job), and macrosystems (cultural patterns and
values, such as cultural norms that confer more power to men than women within the context of
7

sexual relationships) (Bronfenbrenner, 1979). Ecodevelopmental Theory organizes these levels
in the same manner as Social Ecology Theory with a small reformulation to the mesosystemic
level. While typical interpretations of Social Ecology Theory have been more interested in how
functioning in one domain affects functioning in another domain, Ecodevelopmental Theory
proposes that direct interactions between the people involved in the distinct microsystems must
be examined more closely (Szapocznik & Coatsworth, 1999). For instance, parents may directly
influence their children’s peers through interacting with them and providing some level of
parenting to those friends. This represents a direct connection between two microsystems where
members of two social spheres, the family system and the peer system, are interacting directly
with one another. Essentially, Ecodevelopmental Theory proposes that social scientists must
look at protective and risk factors imbedded in different systems in order to accurately
understand adolescent risk (Szapocznik & Coatsworth, 1999).
Developmental Perspective
Secondly, Ecodevelopmental Theory draws on a developmental approach. It posits that risk
and protective factors are often interrelated along a developmental trajectory (Szapocznik &
Coatsworth, 1999). However, it is easy to misinterpret these factors as independent if they are
not studied with a developmental lens. Organizing risk and protective factors along a
developmental path helps to more accurately reflect how problem behaviors develop
(Szapocznik & Coatsworth, 1999). For instance, a child who does not feel supported by his or
her family members might turn to his or her peers for closeness and support. If those peers are
sexually risky, she or he may feel pressured to engage in sexual risk behaviors in order to
maintain those close relationships with her peers.

8

This theory also goes past the traditional conceptualizations of ―development‖ as just
involving intrapersonal processes (Szapocznik & Coatsworth, 1999). Ecodevelopmental Theory
utilizes the term ―ecodevelopment‖. Ecodevelopment is defined as ―the complex set of features
that emerge over time within the child and in the child’s social ecosystems and the nature of the
interactions within and among these systems as they change and influence each other over time‖
(Szapocznik & Coatsworth, 1999) (pg. 342). In other words, ecodevelopment has three distinct
components: (1) development of the adolescent’s intrapersonal features (such as physical and
cognitive attributes), (2) development within the systems that surround the adolescent (such as
changes that occur at the family level), and (3) the interactions within these systems and between
these systems. Ecodevelopmental Theory proposes that all of these aspects of ecodevelopment
must be taken into account in understanding and organizing risk and protective behaviors in
relation to adolescent risk (Szapocznik & Coatsworth, 1999).
Finally, this theory also postulates that the family is, in most cases, the most proximal and
important system in relation to adolescent development (Szapocznik & Coatsworth, 1999). It
forms the foundation from which subsequent child development occurs. When children are
infants, this is the main system at work in their lives, and the first one to increase in complexity
(Szapocznik & Coatsworth, 1999). As they get older, different aspects of their microsystems
begin to play progressively larger parts in risk and protective processes (Szapocznik &
Coatsworth, 1999).
Understanding Interactional Processes
Thirdly, Ecodevelopmental Theory emphasizes the interactions among different systems
(Szapocznik & Coatsworth, 1999). Within this framework the term ―interaction‖ is defined as
―the patterns of relationships and direct transactions among individuals within and across the
9

different domains and levels of the social ecology‖ (Szapocznik & Coatsworth, 1999) (pg. 345).
Therefore, there is a focus on processes that occur between the persons imbedded in different
social systems and how these result in protective or risk trajectories. Ecodevelopmental Theory
posits that the quality of these inter-system interactions and connections determine the degree to
which risky and protective behaviors occur and are sustained (Perrino, et al., 2000).
In summary, Ecodevelopmental Theory suggests that in order to understand any type of
adolescent risk behavior it is necessary to explore: (1) the social systems within which the risk
and protection behaviors occur (Szapocznik & Coatsworth, 1999), (2) the developmental
processes within the adolescent and the systems of interest that surround her or him and (3) the
interrelationships that occur between the risk and protective factors imbedded in the different
social spheres of the adolescent’s social ecology (Szapocznik & Coatsworth, 1999).
Ecodevelopmental Theory: Previous Research. Ecodevelopmental Theory was originally
devised to study ecodevelopmental trajectories that lead to adolescent drug use (Szapocznik &
Coatsworth, 1999). However, its concepts have been applied to other adolescent risk and
protective behaviors (Donenberg, Wilson, Emerson, & Bryant, 2002; Perrino, et al., 2000).
Some researchers have used this framework to understand sexual risk taking in adolescents
(Donenberg, et al., 2002; Perrino, et al., 2000). For instance, Perrino et al. (2000) used
Ecodevelopmental Theory to organize their findings in their review on the role of families in
adolescent HIV prevention. Overall, the review found that parents can be a protective influence
on their children’s problem behaviors (such as drug use) and sexual risk behaviors. Perrino et al.
(2000) also organized risk and protective factors for risk behaviors along the Ecodevelopmental
levels of influence. Perrino et al. (2000) discussed how these may play out developmentally.
For instance, they suggested that some youth enter risky developmental trajectories where an

10

unsupportive family environment leads to involvement with deviant peers (such as peers who use
drugs or engage in criminal activity), and, subsequently, to risk behaviors (Perrino, et al., 2000).
Donenberg et al. (2002), on the other hand, used the theory to guide their research on
family factors related to sexual risk taking among adolescents in psychiatric care (Donenberg, et
al., 2002). They examined how parental monitoring and permissiveness were related to sexual
risk taking in adolescent boys and girls. Donenberg and colleagues (2002) theorized that family
relationships and patterns that were problematic or dysfunctional could become models of
behavior that were then carried out with peers and partners; becoming risk behaviors.
Donenberg et al. (2002) tested whether those individuals who had low parental monitoring and
high permissiveness would be more likely to be involved in sexual risk taking and found that this
was indeed the case. Both of these examples illustrate the advantage of utilizing a framework
such as the Ecodevelopmental Theory. In both pieces, the authors were able to place risk and
protective factors along a developmental trajectory in their discussion of sexual risk.
Additionally, they were able to draw on the fact that there are often multiple social influences to
sexual risk taking and these do not tend to work independently.
As mentioned earlier, this study utilized this theory as a guiding framework with which to
understand and organize social influences on adolescent sexual risk behaviors.
This dissertation focused specifically on micro and mesosystemic forces on adolescent
sexual risk. The purpose was to understand how the familial, peer and partner microsystems can
impact sexual risk taking at the mesosystemic level (which examines interactions between the
systems).

11

Ecodevelopmental Perspective on Familial, Peer and Partner Microsystems
Parents and Families
Families are a crucial microsystem for adolescent sexual development. They play an
important role in keeping adolescents safe and parents are a major influence on their children’s
behavior throughout their development (DiClemente, et al., 2007; Voisin, et al., 2006).

As

suggested by Ecodevelopmental Theory, most adolescents spend a great deal of time in their
family unit which makes it an important point of influence (Donenberg, et al., 2002).
As mentioned above, two aspects of familial microsystems have been found to be very
important for adolescent sexual health: parental or familial support and parental monitoring
(DiClemente, et al., 2007; Kirby & Lepore, 2007). A wide body of literature has linked these
―parental involvement‖ constructs to sexual health outcomes in youth. The following paragraphs
provide a brief summary of this work.
Parental support is usually defined as the degree to which adolescents believe that their
parents care about them and are close to them. Other related constructs that appear in the
adolescent sexual health literature include parental closeness, connectedness and quality of
relationship (Regnerus & Luchies, 2006; Rink, Tricker, & Harvey, 2007). In the adolescent
sexual health literature, support, closeness, connectedness and quality of relationship have
consistently been linked with protective sexual health outcomes in general adolescent
populations including fewer opportunities for sex, later sexual debut or onset, less STD risk, and
lower scores on general measures of sexual risk (Crosby, DiClemente, Wingood, Cobb, et al.,
2001; Henrich, et al., 2006; Resnick et al., 1997; Roche, Ahmed, & Blum, 2008; Rose et al.,
2005; Voisin, et al., 2006). Similarly, work which has explored the relationship between support
and sexual risk taking in African American adolescents has found that increased support is
12

associated to decreased incidence of sexual initiation, anticipated level of sexual activity, and
unprotected sex (Browning, Leventhal, & Brooks-Gunn, 2005; Crosby, DiClemente, Wingood,
Cobb, et al., 2001; Rose, et al., 2005).
Perhaps adolescents who have close relationships with their parents (who are protective
themselves) are more likely to adopt positive parental attitudes, modeling, or protective efforts in
relation to sex (Crosby, DiClemente, et al., 2001b). That is, those teenagers who feel close to
and supported by their parents may internalize protective attitudes and behaviors in relation to
sexual risk taking to a greater degree than those who do not have these close relationships.
Given that the first microsystem that youth are deeply involved in and influenced by is their
familial unit (Szapocznik & Coatsworth, 1999), a deep connection with their families early on
could be very protective for adolescents as their peer microsystems begin to expand.
In addition to support, parental monitoring, or the degree to which parents know and regulate
where their children go and what they do, has also been found to be protective of sexual risk in
adolescent populations (Donenberg, et al., 2002; Huebner & Howell, 2003; Longmore, Manning,
& Giordano, 2001; Miller, et al., 2000; Rodgers, 1999). Overall, a great deal of empirical
evidence suggests that higher levels of parental monitoring result in better sexual health
outcomes for adolescents from delayed onset of sexual activity to decreased chance of pregnancy
(Donenberg, et al., 2002; Huebner & Howell, 2003; Longmore, et al., 2001; Miller, et al., 2000;
Rodgers, 1999). Similarly, the literature on African American adolescents and parental
monitoring indicates that monitoring is inversely related to increased opportunities to engage in
risky sex, decreased instances of sexual initiation, less sexual involvement, fewer instances of
unprotected sex, and reduced chances of STD infection and pregnancy in this population

13

(Baptiste, Tolou-Shams, Miller, McBride, & Paikoff, 2007; DiClemente, et al., 2001; Paikoff,
1996; Rose, et al., 2005; Sturdivant, 2007).
From a developmental perspective, it seems that the mechanisms through which parental
monitoring exerts its protective effect may actually differ for younger teens who are not sexually
involved and older teens who have initiated sexual activity. That is, some research has pointed
to the fact that when teens are younger parental monitoring is protective in that it prevents them
from spending unsupervised time with potential sexual partners, therefore delaying initiation of
sexual activity (DiLorio, Dudley, Soet, & McCarty, 2004; Paikoff, 1995). However, as youth get
older and have more freedom to spend time with their opposite sex peers, parental monitoring
may actually be protective in that it signals to the teens that they are ―cared for’ (Longmore, et
al., 2001). This may, in turn, make them less likely to engage in sexual risk taking (such as noncondom use) to avoid negative consequences (including STDs, HIV and pregnancy) and
maintain good family relations.
Generally speaking, both parental monitoring and support can be understood as indicators of
involved parenting. Empirical research clearly indicates that this parental involvement can be
protective from sexual risk taking; although the underlying mechanisms of this relationship have
not yet been established in the literature. It is important to point out, however, that some research
on parent-child relationships in African American families does provide some insight about some
important underlying aspects of protective parenting. For instance, one study, conducted by
McBride Murry et al. (2007) found that involved parenting (including monitoring and racial
socialization) resulted in increased youth pride (defined in terms of strong racial identity, self
esteem and body image) and decreased peer orientation. Structural equation modeling
demonstrated that these two processes mediated the relationship between involved parenting and

14

sexual initiation (sexual risk was not assessed given the young age of the participants) in 5th
grade youth (McBride Murry, Berkel, Brody, Gibbons, & Gibbons, 2007). The authors explain
that African American parents, unlike ethnic majority parents, have an added responsibility of
educating their children on how to cope with situations in which their racial group may be
devalued. Therefore, it is possible that involved parenting may be related to some important
aspects of self worth that, in turn, protect youth from peer risk and sexual risk taking.
Peers
In addition to these aspects of familial involvement, empirical evidence suggests that peers
are very important influences on adolescent sexual risk (Bearman & Bruckner, 1999; East, et al.,
2006; Scaramella, et al., 1998). Peer sex behaviors (perceived and actual) have consistently been
linked to negative individual sexual health outcomes in adolescents. Three main findings emerge
from this body of literature: (1) adolescents who perceive that their friends are having sex are
more likely to have sex themselves (O'Donnell, Myint, O'Donnell, & Stueve, 2003; Rai, et al.,
2003; Stanton, et al., 2002), (2) sexually active adolescents who perceive that most of their
sexually active friends use condoms, are more likely to use condoms themselves (Miller, et al.,
2000; Rai, et al., 2003; Romer, et al., 1994), and finally (3) two studies have linked actual peer
sexual behaviors (as opposed to perceived peer behavior) to individual sexual risk (Bearman &
Bruckner, 1999; Henry, et al., 2007).
Adolescents are in a developmental stage when they may be distancing themselves from their
parents as they seek more autonomy and independence (Roche, et al., 2008). They may be
looking to their peers to establish their own sense of identity (Sturdivant, 2007). Therefore, this
is a period when friends can play increasingly important roles in influencing their behaviors,
including sex behaviors (Barnes, Reifman, Farrell, & Dintcheff, 2000; Li, Feigelman, & Stanton,
15

2000; Sturdivant, 2007). Social Learning Theory (1977) suggests that peer influence can take
place via modeling and positive reinforcement (Bandura, 1977; Lyons, 2009). That is, attitudes
and behaviors that validate the collective norms around sexuality will be reinforced and
rewarded by the adolescent’s peers whereas others may be ignored and punished (Lyons, 2009).
Therefore, adolescents imbedded in peer groups where risky sexual behaviors are normative
may, over time, adopt these behaviors themselves.
Sexual Partners
Finally, some empirical research indicates that sexual partner risk is more predictive of
negative sexual health outcomes than a person’s own previous behavior (Staras, 2009). For
instance, Staras et al. (2009) found that certain partner characteristics (participant-partner age
discordance of 5 years or more and having an STD in the past year) predicted STD diagnosis.
Interestingly, an indexed measure of sexual risk behavior at the participant level (including age
of first intercourse, number of sexual partners, casual sex in the past 12 months, sex under the
influence of alcohol, and condom use) was associated with STD diagnosis when considering
demographics but not when also considering partner characteristics. Essentially, partner
characteristics were more predictive of STD diagnosis than individual sexual risk behaviors
(Staras, et al., 2009). Similarly, Auerswald et al. (2006) found that higher rates of sexually
transmitted infections (STIs) in African American adolescent girls were determined more by the
level of risk of their sexual partners than their individual level of risk. This study investigated
whether partner incarceration (this partner risk characteristic was chosen because it was the only
partner characteristic associated with both gender and STI diagnosis) mediated the relationship
between gender and rate of STIs. Controlled analyses demonstrated that the gender difference in
STI rates (girls had higher diagnosis than boys in this sample) was eliminated when controlling
16

for partner incarceration, suggesting that this construct indeed mediated that relationship. In
other words, when the effects of partner incarceration were accounted for in analyses, gender
differences on STI diagnosis disappeared (Auerswald, et al., 2006). Girls were simply more
likely than boys to have partners who had been incarcerated. These analyses suggest that this is
the reason why girls had higher STI diagnosis (despite the fact that boys actually had higher
levels of individual level sexual risk).
It is important to note that having risky partners is often conceived as the outcome of
parent and peer influences. However, partners are a vital part of youths’ microsystems and
therefore, may also be conceptualized of as an important proximal influence on adolescents’
sexual behaviors.
The following sections will discuss the ways in which protective parenting (including
parents monitoring and support) may protect youth from sexually risky peers and partners.
Firstly, the relationship between parenting and peer risk is discussed.
Adolescent Parent Relationships, Peer Relationships and Sexual Risk: An Interactions Approach
Overall, a good deal of research has established that both parent and peer microsystems are
important influences on adolescents’ sexual development trajectories (risky or protective).
However, parental and peer social systems do not work independently. As suggested by the
Ecodevelopmental Theory, different aspects of parent and peer systems may work together to
influence sexual risk taking from childhood through young adulthood.
Given this, two important questions emerge: (1) What is the nature of this interaction? and
(2) What exactly does it mean for these systems to interact with one another in relation to sexual
risk taking? The Ecodevelopmental Theory would suggest that the nature of the interaction is
multi-systemic. According to this theory, parental and peer microsystemic forces cascade
17

downward and are manifested at the individual level (Szapocznik & Coatsworth, 1999). For
instance, a child who has internalized protective sexual attitudes as a result of an involved
familial relationship may be less likely to internalize or act upon the risky sexual norms of his or
her peer group. As mentioned earlier, the primary microsystem for youth is their family. If
strong, protective relationships are formed early on within this social sphere, that child may be
less invested in following risky peer norms. The adolescent’s peers are not the only source of
support; the adolescent also has a strong familial relationship which he or she may want to
preserve. Therefore, it seems plausible that, given the proper developmental circumstances,
involved parenting could buffer the negative effects of sexually risky peers.
Some empirical evidence provides support for this hypothesis. For instance, Miller et al.
(2000) examined whether different systems act in an additive manner. They studied multiple
variables from different levels of influence (self system, family system and extra-familial
system) and found a linear relationship between the number of factors from different systems
(e.g., familial, peer etc.) identified as ―at-risk‖ and indicators of adolescent sexual risk behavior.
Multiple regression analyses revealed that different variables at multiple levels had a significant
relationship to sexual health outcomes (including intercourse frequency, number of sex partners,
age of first intercourse and condom use). For example, Miller and colleagues (2000) were able
to determine that number of sex partners was significantly predicted by level of parental
monitoring and peer group risk. In other words, the more ―at risk‖ indicators a participant had in
each of the levels of ecology (self, familial and extrafamilial) the more likely they were to have
engaged in sexual risk taking (Miller, et al., 2000). This study suggests that distinct aspects of
parent and peer microsystems can act in concert to affect sexual risk taking. Therefore, it seems
plausible that poor (or lack of) parental involvement may, in fact, facilitate the influence of high

18

risk peers on sexual behavior. However, this study simply looked at multiple social influences in
an additive manner. It did not explore how distinct social influences might interact with one
another. Additionally, like much of the work on this subject, the data were egocentric. In other
words, only participant data were analyzed.
In addition to Miller’s (2000) work, only one study, conducted by East, Khoo and Reyes
(2006), has tested how protective parenting (including monitoring) and peer sexual risk may
interact to influence sexual risk outcomes in adolescents. Specifically, East et al. (2006)
examined whether protective parenting (conceptualized as strictness, monitoring, educational
expectations, and disapproval of teen sex and parenting) buffered against peer sexual risk
(operationalized as the proportion of friends who had had intercourse and the level of peer
pressure to have sex). They tested the interaction between protective parenting and peer sexual
risk in a sample of Latina and African American girls from a high risk environment (N=128) and
found the interaction to be significant (East, et al., 2006). Involved parenting appeared to protect
against pregnancy even in the presence of peer risk (East, et al., 2006). Adolescents who had
high risk peers were less likely to get pregnant if they also had high levels of protective
parenting. Furthermore, protective parenting helped maintain low pregnancy rates, even when
adolescents had high risk peers. In fact, within the group who received protective parenting 32%
of girls who had high peer risks and 32% of girls who had low peer risks became pregnant. In
contrast, 77% of the girls who received low protective parenting and were exposed to peer risks
became pregnant. (Twenty-three percent of girls who had both low protective parenting and low
peer risks had a pregnancy (East, et al., 2006).) This study suggests that protective and involved
parenting might act as a buffer against pregnancy. This is a heartening finding that should be
corroborated in other high risk populations, given that it could inform effective prevention and

19

intervention efforts. However, East et al.’s (2006) (like Miller’s study and a great majority of the
work on parental and peer influences to sexual risk taking) examined how aspects of parent and
peer systems interact from an egocentric perspective which did not take ―true‖ peer risk into
account. This dissertation sought to build upon this work by utilizing multi-informant data and
investigating other sexual risk behaviors and outcomes including condom use and STD
diagnosis.
Do family and peer systems interact to influence boys and girls similarly? In addition to
considering mesosystemic level interactions of parent and peer microsystems it is also important
to examine how gender may be involved in these associations. It seems likely that gender could
be an important moderator of these relationships given that the main outcome of interest in this
study is condom use; a behavior over which boys have more power (Wingood & DiClemente,
2000). Additionally, boys and girls are socialized differently, expectations (from parents and
peers) regarding sexual behaviors differ between them. Therefore, it seems likely that these
relationships would not be the same for boys as they are for girls.
There is some empirical evidence which seems to provide support this argument. Firstly, we
know that girls are monitored more heavily than boys (Borawski, Ievers-Landis, Lovegreen, &
Trapl, 2003; Donenberg, et al., 2002; Li, Feigelman, et al., 2000; Li, Stanton, et al., 2000;
Longmore, et al., 2001). It is possible that this differential treatment may be the result of
predominant gender norms which are generally more accepting of risky male behavior than risky
female behavior (Browning, et al., 2005). This disparity could have important implications for
parent and peer system level interactions. It seems plausible that, given that girls receive more
monitoring, they also engage in less risk even in the presence of high risk peers. Research has
not yet examined whether this is the case. However, some evidence has pointed to the fact that

20

parental monitoring may play a more important role in protecting girls from sexual risk than
boys (Donenberg, et al., 2002). For instance, Donenberg et al. (2001) found that parental
monitoring predicted fewer sex partners and higher levels of sex with condoms in girls but not
boys (Donenberg et al., 2002). (The girls in this study were more highly monitored than the
boys.) The authors suggest that ―girls may be more influenced by parents than are boys because
they place a high value on maintaining important relationships, and girls may be more responsive
to parental monitoring or permissiveness in order to protect the parent-adolescent relationship‖
(pg. 141).
On the other hand, some research suggests that parental monitoring may actually have a
stronger association with boys’ protective sexual behaviors than girls’. Sturdivant (2007) found
that males who had higher levels of parental monitoring also had fewer sex partners (Sturdivant,
2007) in a sample of 156 African-American preadolescents. This was not true for females in the
study. The young women in this study also reported higher levels of parental monitoring than
the young men. The author suggests that females may be monitored at such high levels that
increases in monitoring do not have an impact on number of sexual partners (Sturdivant, 2007).
Finally, Borawski et al. (2003) found that the girls in their study were monitored more
heavily than the boys. However, this did not seem to have a significant impact on sexual risk
activities (operationalized from 5 yes-or-no questions about number of sexual partners, being
diagnosed with an STD, carrying a condom, refusing unprotected sex and consistent condom
use). Highly monitored males, on the other hand, were more likely to use condoms than their
counterparts.
Overall, it appears that higher levels of parental monitoring are protective for both boys
and girls. However, it is possible that the threshold for impact on girls’ behaviors, in the

21

presence of high levels of sexual risk within her peer group, may be lower than it is for boys.
Perhaps, as suggested by Donenberg (2001) girls are more interested in preserving familial
relationships than boys. Regardless, it is clear that further work is needed to elucidate if the
relationship between parental monitoring and sexual health outcomes is different for boys than it
is for girls.
In addition to this work on parental monitoring, one study investigated how support is
related to sexual risk in boys versus girls. Roche, Ahmed and Blum (2008) explored how family
socialization (including familial closeness) in adolescence was associated with sexual onset and
number of partners in early adulthood in male and female participants. Structural equation
modeling analyses demonstrated that girls with more family closeness during early adolescence
were less likely to report onset of sexual intercourse by middle adolescence and had fewer sex
partners in early adulthood (Roche, et al., 2008). For males, they found that family closeness
was unrelated to sexual initiation by middle adolescence but was directly associated with having
fewer sex partners during early adulthood (Roche, et al., 2008). Authors suggest that given
societal norms which condone sexual behavior in boys, it is not surprising that stronger parental
relationships do not necessarily result in later onset of sexual intercourse. However, this does
not explain why closeness would be related to fewer sexual partners in early adulthood. It might
be the case that these boys with strong familial relationships internalized attitudes regarding
―conventional‖ and protective norms from their families in regards to sexual risk taking (Voisin,
et al., 2006).
Interestingly, this study points to a developmental trajectory of protection in girls. That
is, at middle adolescence closeness appears to protect them from early intercourse, whereas in
early adulthood, it appears to protect them from having many sexual partners. This is compelling

22

evidence that perhaps the effects of these protective parental relationships can take place along a
developmental path. Understanding if this would still be the case, even in the presence of a
sexually risky peer group, would add another important component to our current knowledge of
adolescent sexual development trajectories.
Finally, some research suggests that peer influences could differ by gender as well. For
instance, Henry et al. (2007) studied a multi-ethnic sample from the Add Health database and
found that girls whose friends perceived fewer costs to sex were more likely to engage in
intercourse without a condom 1 year later. This was not observed among boys. The authors
suggest that girls may experience greater levels of interdependence in their relationships, making
them more susceptible to their friends’ influence in relation to sex without condoms. On the
other hand, Miller (2000) found that gender did not moderate the relationship between peer
group sexual intercourse and frequency of sexual intercourse in a African American and
Hispanic, geographically diverse sample. It seems possible that ―frequency of sexual
intercourse‖ as a measure of sexual risk may not be as susceptible to peer influence.
Overall, it is important to note that evidence regarding the link between peer risk and
individual risk taking is contradictory and largely reflective of the specific risk outcomes
measured in these different studies.
Adolescents’ Parent Relationships, Sexual Partners, and Sexual Risk: Considerations on
interactions and gender.
As mentioned above, adolescent sexual risk literature points to the importance of considering
partner characteristics in understanding sexual risk taking (Staras, et al., 2009). However studies
have not yet examined whether protective parenting may be able to buffer some of those effects.
Given the fact that this area of research remains underexplored, this dissertation study sought to

23

conduct preliminary, exploratory analyses about the intersection between aspects of parentadolescent and partner-adolescent relationships. Specifically, this study tried to understand
whether engaged parents (who support and supervise their children) can protect a child by
bolstering the odds that they will use condoms with their sexual partners who are typically
unsafe (i.e., have multiple sexual partners or have been diagnosed with an STD). It seems
possible that, in this situation, the same mechanisms via which involved parenting can protect
youth from a sexually risky peer group might protect them from a sexually risky partner. That is,
it is plausible that a teen who feels that her parent cares about her might be more likely to use
condoms even if she has a typically risky partner. However, it is important to point out that
within partner relationships there are other factors at play when it comes to sexual risk. Firstly,
as mentioned earlier, condom usage usually falls to the male partner, often resulting in a power
imbalance in the context of male-female sexual relationships (Wingood & DiClemente, 2000).
Therefore, it seems plausible that boys may be more likely to engage in safer sex, even if their
partner is typically unsafe, in the context of a protective familial environment. They might
simply be more empowered to enact this behavior.
Secondly, perception of sexual risk, with any particular sexual partner, will play an important
role on whether or not condom use will take place with that partner. Thus, a strong familial
relationship may not necessarily result in protective behavior with a sexual partner in the absence
of perception of risk. Male and female power differentials and perception of risk should be
considered in attempting to understand how parental relationships might affect sexual health
outcomes in the context of sexually risky partners.

24

Studying Social Influences on Sexual Risk Taking: Bridging the Gap between Theory and
Research Methodology
As mentioned above, another important issue to consider in conducting research on social
influences on sexual risk taking in adolescents is the current gap between the theory and the
methodology. There is wide agreement in the literature on adolescent sexual risk of the
importance of studying the social forces that can influence risk-taking in youth (DiClemente, et
al., 2007). In addition, many theories have been devised to try to understand contextual
influences on risk behaviors. However, very few studies take on the task of using methods that
take those contexts into account (Luke, 2005). Specifically, assessing social influences
independently of the participants’ perceptions is still very uncommon in the literature on
adolescent health. An important step in learning how social contexts can influence individual
behavior is to capture those social contextual influences to allow for a more nuanced and
complete understanding of the social phenomena of interest and how it can impact a specific
sexual risk outcome such as non-condom use. If, as social scientists, we intend to understand
how a particular social force, such as peers, can influence sexual risk taking, it is important to
measure the level of sexual risk in a participants’ peers. This study attempted to bridge the gap
between theory and methodology by utilizing multi-informant data provided by index
participants as well as those participants’ peers.
Additionally, Ecodevelopmental Theory proposes that risk and protection often occur
along developmental trajectories. In relation to sexual risk taking and STD diagnosis, this
suggests that protective and risk factors must precede sexual taking and STD diagnosis. This
study attempted to address this aspect of the Ecodevelopmental Theory by analyzing data across
two waves (conducted 12 months apart from one another) and trying to determine whether parent

25

and peer influences at an earlier time had an effect on sexual risk taking and STD diagnosis at a
later time. (Given the fact that even with consistent condom use, STD diagnosis might occur if
one’s partner has an STD, it was proposed that the relationship between risky sexual partners and
STD status would be partially mediated through condom use behaviors.) Overall, this study
examined whether protective parenting could result in a protective sexual behavior trajectory in
relation sexual risk taking.
Additionally, given the age group of interest (14 to 19 years old) the outcome of interest
in this study was sexual risk taking (i.e., condom use) as opposed to sexual initiation (which is
the outcome variable of interest of a great majority of the literature in this field). As mentioned
earlier, this outcome seems more developmentally appropriate for teens in this age group given
that sexual initiation is a normative part of adolescent sexual development.
Ecodevelopmental Theory also suggests that distinct social systems interact with one
another in impacting adolescent risk. Moderation analyses were conducted to determine if this is
the case within the study sample. Specifically, the interactions between involved parenting and
sexually risky peers, and involved parenting and sexually risky partners were examined to verify
if involved parenting can buffer the effects of risky peers and partners.
Hypotheses and Research Questions
In summary, empirical evidence clearly suggests that parent, peer, and partner influences
impact sexual risk taking. However, it seems possible that these interactions might be different
for boys than they are for girls.
Research Questions
This study attempted to answer the following questions:

26

1. Is involved parenting a significant buffer against high risk sex for adolescents whose
peers are sexually risky?
2. Does gender moderate the relationship between peer sexual risk and sexual risk taking?
3. Does higher risk taking predict higher incidence of STD diagnosis?
Exploratory Questions
1.

Is involved parenting a significant buffer against high risk sex for adolescents whose
partners are typically sexually risky?

2. Does gender moderate the relationship between risky sexual partners and sexual risk
taking?
3. Does perception of risk moderate the relationship between risky sexual partners and
sexual risk taking?
Hypotheses
Main Effects
1. (Path a) It was expected that individuals with high levels of involved parenting (Time 1)
will have lower levels of sexual risk-taking (Time 2).
2. (Path b) It was hypothesized that adolescents whose peers engage in high levels of sexual
risk (Time 1) will have higher levels of sexual risk-taking (Time 2).
3. (Path c) It was expected that individuals with high risk sex partners (Time 2) will engage
in higher levels of sexual risk taking (Time 2).
4. (Path d) It was hypothesized that participants reporting higher levels of perception of risk
(Time 2) will have lower levels of sexual risk taking (Time 2).

27

5. (Path e) It was hypothesized that individuals with higher levels of sexual risk taking
(Time 2) will be more likely to have a positive STD diagnosis (Time 2).
Moderating Effects
1. (Path not modeled) It was expected that involved parenting will act as a buffer in the
relationship between sexually risky peers (Time 1) and risky sexual behavior (Time 2).
2. (Path not modeled) It was expected that respondent’s gender will moderate the impact of
involved parenting (Time 1) on sexual risk (Time 2).
Mediated Effects
1. (Paths a and e) It was hypothesized that the path from involved parenting to STD
status will be fully mediated through sexual risk taking.
2.

(Paths b and e) It was expected that the path from sexually risky peers to STD status
will be fully mediated through sexual risk taking.

3. (Paths c and e) It was expected that the path from risky sexual partners to STD status
will be partially mediated through sexual risk taking.
4. (Paths d and e) It was hypothesized that the path from perception of risk to STD
status will be fully mediated through sexual risk taking.

28

Figure 1. Conceptual Model of Parent, Peer, and Partner Influences on Sexual Risk Taking
Time 1
Involved
Parenting
(Monitoring
and Support)

Time 2

a
e

Sexual
Risk
Taking

b
Sexually
Risky Peers

c

Risky Sexual
Partners

29

d

Perception of
Risk

STD
Status

METHOD
Research Design
The data for this dissertation were drawn from the Bayview Network Study (BNS). The
BNS was a multi-wave prospective study of African American youth and members of their social
and sexual networks. The study was conducted in San Francisco, California from 2000-2002 and
it was designed with the purpose of studying risk factors for STIs in African American
adolescents. Teens were eligible to participate if they were African American, between 14 and
19 years of age, and living in Bayview-Hunter’s Point (a low income area in San Francisco).
Index participants, their friends and their sexual partners were recruited to participate throughout
3 waves of data collection. Friends and partners were considered eligible to participate if they
were at least 14 years old and living in San Francisco.
Participants
This dissertation study utilized the data collected from index participants and their social
friends at waves 1 and 3. Specifically, given the aims of this study, the sample consisted of index
participants who reported having had sexual intercourse with an opposite sex partner in the past
6 months at wave 3 (n=199).
Overall, 58% (n=116) of index participants were female and 42% (n=83) were male. The
average age of the index participants was 17.2 (SD=1.4). Most youth reported that they lived
with one parent or guardian 65.8% (n=131); 28.6% reported that they lived with 2 parents or
guardians (n=57) and 5.5% reported that they lived with someone who was not their guardian or
by themselves (n=11).

30

Procedure
Recruitment
The recruitment effort for the BNS was a combination of population-based random
sampling and snowball sampling of friends and sexual partners. Index participants were located
and recruited through random digit dialing of phone numbers (corresponding to the 94124 ZIP
code).
Recruitment took place in several steps. At baseline, random digit dialing was used to
recruit index participants into the study. In order to locate and recruit index participants, over
40,000 numbers were dialed (Ellen et al., 2005). Out of these, 22,000 were identified as
households. Ninety-seven percent were considered ineligible because they were out of the target
geographical area or no African American adolescents, between the ages of 14-19, lived there.
After this process, a total of 673 teenagers living in 470 eligible households were identified.
Ninety-three of the potential participants were excluded because they were actually duplicates.
Potential participants who could not be reached by phone were referred to the University of
California, San Francisco (UCSF) staff. Staff attempted to locate participants with home visits,
word-of-mouth, queries, public records, and assistance of key informants and community leaders
(Ellen, et al., 2005). Ultimately a total of 580 household youth were offered the opportunity to
complete the interview (Auerswald, et al., 2006). Out of these 580, a total of 350 were
interviewed for a response rate of 60.3%.
Reasons for not participating in the study broke down in the following way: 20%
guardian refusal, 13% respondent refusal, 4% informant refusal, 3% respondent not available
(Auerswald, et al., 2006). Overall, males and younger adolescents were least likely to accept
participation.
31

Those 184 index participants who reported that they had initiated sexual activity were
asked to name two social friends. They also gave brief descriptions of those friends.
Finally, at wave 3, conducted 12 months after baseline, index participants and friends
were interviewed again. Those sexually active participants (within the past 6 months) were
asked to name sex partners in the past 6 months, since the last interview. After that, these sex
partners were recruited into the study and asked to name up to 6 partners in the past 6 months.
The partners of the partners were then pursued for recruitment.
Data from 199 participants who had been sexually active in the last 6 months at Wave 3
(12 months after Wave 1) were utilized in the analyses of this study. Out of those 199
participants, 102 had at least 1 social friend who was also interviewed in the study. Data from
those 102 social friends was utilized to create a ―peer sexual risk‖ scale that is described below.
Consent and Confidentiality
After youth’s eligibility was assessed, informed consent was obtained. Parental consent
was obtained for teens younger than 18 years old. Parental consent was obtained by phone and a
copy of the consent form was mailed and returned with the parent’s signature (Auerswald, et al.,
2006). Consent was obtained directly from participants 18 years of age and older prior to the
start of the interview. Participants who were younger than 18 years old gave assent. Consent or
assent was obtained from youth by telephone or in person. The study was approved by the
1

Institutional Review Board (IRB) at the UCSF .

1

Michigan State University’s IRB was consulted in regards to this research project. They informed the
investigators that IRB approval from the MSU IRB would not be necessary given that the data utilized for
analyses is completely de-identified. The corresponding IRB documentation is included in Appendix B.

32

Survey Administration
A majority of the interviews were conducted over the phone although some were
conducted in person (Ellen, et al., 2005). When possible, named friends and sex partners were
recruited and interviewed by phone as well. Participant recruitment and interviews were
conducted by University of California at Berkeley’s Survey Research Center. UCSF staff (in
addition to locating hard-to-reach participants) collected urine and vaginal swabs to determine
STD status.
During the interview process, interviewers entered the responses directly into a computer
utilizing the CASES software program (Computer-assisted Survey Methods, USC Berkeley,
California). The interview lasted an average of 45 minutes and participants received $25 for
their participation in the interview and $10 for sexually transmitted disease testing (Ellen, et al.,
2005).
All wave 1 and 3 participants were tested for STIs and all female participants were tested
for pregnancy (Ellen, et al., 2005). (Girls who reported that they were virgins were tested for
pregnancy as well, to prevent exposing adolescents’ sexual initiation status to their peers.)
Gonorrhea (Neisserie gonorrhoeae) and Chlamydia (Chlamydia trachomatis) tests were
conducted on men’s urine samples and women’s vaginal Q-tip swabs. Young women were also
asked to provide a urine sample to test for pregnancy. Information was also collected regarding
STI symptoms. Those participants who tested positive were notified and offered free treatment.
Youth undergoing STI testing were met in the field by study staff. The staff members provided
written materials to the adolescents on how to properly take the tests. Convenient locations were
used for testing including home, school or workplace (Auerswald, et al., 2006).

33

Measures
Demographics
Demographic characteristic data were gathered in relation to age, race, gender and with
whom the adolescents lived. Age was measured with the question ―How old are you?‖ Race
was measured with the question ―What is your race or ethnicity?‖ (As mentioned above, all
index participants were African American given the eligibility criteria.) Gender was assessed
with the question ―Are you male or female?‖ Finally, who the adolescent lived with was
measured with the following questions: (1) Who are you currently living with, both your parents,
only one parent, someone else, or are you currently living by yourself? (2) Are you living with
someone who is your female guardian like a grandmother, aunt or foster mother, or someone else
who is your female guardian? (3) Are you living with someone who is your male guardian like a
grandfather, uncle, or foster father, or someone else who is your male guardian. Given that some
literature suggests that two parents or guardians can be more protective (presumably because
there are more opportunities for monitoring and support) (Bearman & Bruckner, 1999), the items
above were utilized to create a new variable with the following categories: (1) adolescent lived
with one parent or guardian, (2) adolescent lived with two parents or guardians, (3) adolescent
lived by themselves or with someone else who was not their parent or guardian.
Involved Parenting
A confirmatory factor analysis was utilized to test involved parenting as a latent
construct. Two scales or sub-constructs were entered into the CFA for this variable. These were
parental monitoring and parental support.

34

Parental Monitoring
Parental monitoring was measured on a 6-item, 3 point scale. Items in the scale asked
how much parents tried to know about their teen’s whereabouts at night, how much they tried to
know about what they did with their free time, and how much they tried to know about their
whereabouts during the day. It also asked how much their parents actually knew about their
whereabouts at night, how much they actually knew about what they did with their free time and
how much they actually knew about their whereabouts during the day. The response set for all
of the questions consisted of the following options: ―not at all‖, ―a little‖ and ―a lot‖. When the
items were averaged, they produced a mean of 2.44 (SD = .45, Range= 1-3) and a satisfactory
Cronbach’s alpha of .79. Higher scores indicated higher levels of parental monitoring.
These items were centered (by subtracting 2 from each item) before importing the dataset
into MPlus. Centering was conducted to account for collinearity amongst variables, given that
the items were used to create cross-product interaction terms in subsequent analyses. This
centering method (subtracting 2 from each item) was chosen given that parental monitoring was
scored on a 3 point scale, ranging from 0 - 3. Therefore, centering at 2 resulted in whole
numbers, allowing for easier interpretation of item values.
Parental Support
Parental support was measured with four items. These items asked about closeness to
their female and male guardians and/or their biological mother and father. Questions about
biological parents were only asked if the adolescent reported that they did not live with their
biological mother or father. The items were worded in the following way: ―How close do you
feel to your female guardian? By close I mean, feeling like you can count on her to talk with you
about your personal problems and to provide you with emotional support.‖ The same question
35

was asked in regards to their male guardian, biological mother (only asked if they did not live
with her) and biological father (only asked if they did not live with him). The response set
included the options: ―Not close at all‖, ―Not too close‖, ―Somewhat close‖, and ―Very close‖.
Some participants could have more than one parental figure of each gender with whom they had
routine contact (e.g., stepfather and biological father or a mother and stepmother), and therefore
two parental support scores for male parental figures and/or female parental figures. In these
cases, those two scores were averaged together to create an overall ―male/female parental figure
support‖ score. The reason why these support items were handled this way, instead of simply
using the guardian scores, is that almost a third of the sample reported that they did not live with
their biological father but did have a relationship with him (n=59, 30%). Specifically, out of
those 59 individuals, 79% (n=46) said that they saw their biological father at least once a week,
8.4% (n=5) said that they saw him less than once a week, and 5.1% (n=3) said they saw him less
than once a month. Therefore, this made it possible to retain information about those
relationships within the score. In the end, each participant had two support scores: one for their
male parental figure(s) and one for their female parental figure(s). Given that this was a 4-point
scale, with no true 0 point, each item was mean centered before importing the dataset into
MPlus. The mean parental support score from female parents or guardians was 3.4 (SD = 0.84,
responses ranged 1-3). The mean parental support score from male parents or guardians was
2.47 (SD = 1.16, responses ranged 1-3)
Sexual Risk: Condom Use
Condom use was measured with the question: ―How often have you and [partner 1] used
condoms in the last 6 months – every time, most of the time, a few times or never?‖ (This
question was asked in relation to 1- 6 sexual partners.)
36

An index of sexual risk/non-condom use was calculated for each index participant by
summing their scores, across all partners, to the question(s) above. Therefore, a participant who
had 2 partners and used condoms most of the time (score of 1) with partner 1 and a few times
(score of 2) with partner 2, received a score of 3. Summing was chosen over averaging in
creating this index, because conceptually it captures the notion of cumulative risk. In other
words, a person who has more partners with whom he or she is ―sexually risky‖ will have a
higher score than a participant who is ―sexually risky‖ with just one partner.
Overall, participants could receive a score ranging from 0-18, with higher scores indicating
higher levels sexual risk taking or non condom use. The mean score across partners was 1.3
(SD=1.62) and the median was 1.0. Scores ranged from 0-9.
This scale was found to be positively skewed and kurtotic. As suggested by Tabachnick and
Fidell (2006), a log transformation was conducted in order to correct for this (Tabachnick &
Fidell, 2007). Prior to transformation the skewness was 1.40, after transformation it was .42;
prior to transformation the kurtosis was 2.40 and after transformation the kurtosis was -1.27).
Use of the ML estimation in structural equation modeling, assumes a normal distribution,
therefore, this transformation was conducted in order to better met that assumption.
Sexually Risky Peers
The measure of sexual risk taking of peers was constructed in the same way as the measure
of sexual risk taking in index participants (data from 102 of the index participants’ social friends
was used). ―Sexually risky peers‖ was measured with the question: ―How often have you and
[partner 1] used condoms in the last 6 months – every time, most of the time, a few times or
never?‖ (This question was asked in relation to 1-6 sexual partners.) Level of sexual
risk/condom use was calculated for each social friend by summing their scores, across partners,
37

to the question(s) above. Overall, participants could receive a score ranging from 0-18, with
higher scores indicating higher levels sexual risk taking or non condom use. The mean score
across partners was .94 (SD=1.44) and the median was 0. Scores ranged from 0-8. (Friends who
had not initiated sexual intercourse or had not had sexual intercourse in the last 3 months,
received a score of 0, indicating a non-sexually risky friend.)
Some index participants had data for one close friend (n= 65), whereas others had data for
two close friends (n=37). All participants said that they had at least one close friend, therefore,
for those participants who had data for two close friends (n=37), one of the two was chosen at
random. The sexual risk score for the randomly chosen friend became the measure of ―peer
sexual risk‖ for the corresponding index participant.
Finally, normality analyses showed that this index was negatively skewed (-2.04) and highly
kurtotic (5.02). However, several transformations were attempted, none of which improved the
distributional properties of the variable. (A log transformation, worsened the skewness and a
square root transformation worsened the kurtosis.) Therefore, the decision was made to use the
original variable; capped at a score of 4. This was a meaningful score because any given
participant could use condoms every time (score of 0), most of time (score of 1), a few times
(score of 2), or never (score of 3). Therefore, anyone who had a 4 or higher, had inconsistent
condom use with at least 2 partners. Additionally, only 6 individuals had scores higher than 4
(with the highest being a score of 8).
Sexually Risky Partners
―Sexually risky partners‖ was measured utilizing two items and one indexed measure
(age difference between participant and partner). The two items (asked of 1-6 sexual partners)

38

were: (1) Did [he/she] ever have a sexually transmitted disease? and (2) Did [he/she] have any
other sex partners in the last 6 months?
Finally, age difference between respondents and their partners was utilized as an
indicator of ―risky sexual partners‖. Previous empirical work has suggested that an age
difference of over 5 years between sexual partners can result in increased sexual risk taking due
to power differentials in negotiating safer sex (Wingood & DiClemente, 2000).
A total of 32 participants (16%) reported that they had one partner who was 5 years older
than they were and 4 participants (2%) reported that they had two partners who were 5 years
older than they were (at wave 3). Additionally, 23 participants (11.6%) reported that they had at
least one partner who had been infected with an STD and 57 participants (28.6%) reported that at
least one of their partners had had other partners within the last 6 months. (Approximately 20.1%
of participants said one of their partners had other partners, 7.0% said that two of their partners
had other partners, 1.5% said that three or more of their partners had other partners)
The answers to these three items were dichotomized so that each positive response
received a 1 and each negative response received a 0. Participants could receive a score from 018, with higher scores indicating higher levels of perceived sexual risk in their sexual partners.
Scores ranged from 0 to 5 (Mean = 0.73, SD = 0.98).
Normality analyses showed that this index was positively skewed. A square root
transformation was conducted to help correct for this (prior to transformation the skewness was
1.40, after transformation the skewness was .51). This variable was also mean centered before
the dataset was imported into MPlus.

39

Perception of STD and HIV Risk
Perception of (partner-specific) STD and HIV risk was measured with one item each: (1)
―How likely do you think it is that [he/she] (could/did) give you a sexually transmitted disease?
Would you say very likely (score of 4), somewhat likely (score of 3), somewhat unlikely (score
of 2), or very unlikely (score of 1)?‖ and (2) ―How likely do you think it is that [he/she]
(could/did) give you HIV? Would you say very likely, somewhat likely, somewhat unlikely, or
very unlikely?‖ Each question was asked about 1-6 sexual partners. When averaged across each
participant’s partners, the mean level of perception of STD risk was 1.53 (SD=.81) and the mean
level of perception of HIV risk was 1.41 (SD=.71). The scores to these two items were averaged
together to create an index of HIV/STD risk. The mean score for the sample was .80 (SD=.26).
Scores ranged from 1 to 4.
Additionally, the perception of HIV/STD risk measure was found to be positively skewed.
Inverting the variable seemed to correct for this, somewhat, (prior to transformation the
skewness was 1.54, after transformation it was - 0.74). This variable was also mean centered
before the dataset was imported into MPlus.
STD Status
As mentioned above, participants were tested for gonorrhea and chlamydia. Positive
results to these two tests were combined into one variable which indicates presence of one or
both of the STDs. A total of 163 participants were tested for STDs out of these 8% (n=13) had a
positive result.

40

Control Variables and Preliminary Analyses
Preliminary analyses assessed the relationships between demographic characteristics and
outcome variables in order to determine the need to covary sample demographic characteristics
in further analyses. Specifically, consistent with previous literature, the following demographic
variables were examined: age and with whom the adolescent lived.
The need to control for childbearing attitudes was also assessed, given that previous
research has shown a link between positive childbearing attitudes and unprotected sexual activity
in adolescents (East, et al., 2006). Childbearing attitudes were assessed with one question: ―In
the next 3 months, do you want to get pregnant by [Main Partner] - would you say definitely no,
probably no, probably yes, or definitely yes‖ (for females) or ―In the next 3 months, do you want
to get [Main Partner] pregnant - would you say definitely no, probably no, probably yes,
definitely yes, or is she already pregnant?‖ (for males). This question was asked in relation to
the participants’ main partner, if one was identified. If a main partner was not identified, then
the question was asked in relation to the participants’ casual partners.
Finally, some research has found household income or socioeconomic status to be linked
to sexual risk and protection in African American youth (Millhausen, 2003). Given that income
data were not collected from the participants, census block groups (CBGs) were utilized to create
a proxy measure of household income. Median household income data for distinct census block
groups were obtained from the 2000 Census Database (United States Census Bureau, 2000).
The participants’ CBGs were then matched to the corresponding median household income to
create the ―median household income‖ variable. This median household variable ranged from
$12,000 per year to $102,000 per year. In order to facilitate analyses, the variable was rescaled
by dividing it by 1,000.

41

As expected, sexual risk taking was significantly correlated with both age (r=.265, p<.01)
and with whom the adolescent lived. Specifically, older youth were less likely to use condoms
than younger youth. Living alone or with a person who was not their guardian was positively
correlated with sexual risk taking (0.20, p<0.01); living with one parent or guardian was
negatively correlated with sexual risk taking (-0.16, p<0.05). Interestingly, living with two
parents or guardians was not significantly correlated with sexual risk taking. Attitudes toward
pregnancy were not correlated to any of the outcome variables or dependent variables. Median
household income was not correlated to the outcome variable. However, it was negatively
correlated with the measure for risky sex partners (r=-.148, p<.05). In other words, as median
income rose, the risky sex partner score decreased.
Based on these preliminary results, age, who the participant lived with, and median
income were entered in subsequent analyses as control variables.
Preliminary descriptive analyses were also conducted to verify how similar or dissimilar
adolescents who had friends who were interviewed were from adolescents who had no friends
who were interviewed. These analyses demonstrated that adolescents who had at least one friend
who was interviewed were similar in age and gender distribution to those who had no friends
who were interviewed. However, those participants who had at least one friend who was
interviewed had, on average, higher levels of sexual risk taking, riskier sex partners, and had
higher median household income levels (See Table 2). Independent samples t-tests were
conducted in order to ascertain whether these mean differences were statistically significant.
Only the differences in level of sexual risk taking emerged as statistically significant between the
two groups. (See Table 2 below.) In other words, participants who had friends who were

42

interviewed were significantly more likely to engage in higher levels of sexual risk taking than
their counterparts who did not have friends who were interviewed in the study.

43

Table 1. Participant Demographic Characteristics: Participants with 1 or 2 friends who were
interviewed versus participants with no friends who were interviewed
Type of Participant
Participants with 1 or 2

Participants with No

Friends who were

Friends who were

Interviewed

Interviewed

(n=102)

(n=97)

14-19

14-19

17.3, 1.3

17.0, 1.5

18.0

17.0

Male

41.2%

40.2%

Female

58.8%

59.8%

0–9

0–5

1.7, 1.8

0.9, 1.3

1.0

0.0

0 -5

0-3

0.8, 1.1

0.6, 0.8

0.0

0.0

T

df

1.5

197

3.0**

194

1.2

193

Age (in years)
Range
Mean, SD
Median
Gender

Sexual Risk
Taking
Range
Mean, SD
Median

Risky Sex Partners
Range
Mean, SD
Median

44

Table 1 (cont’d)
Type of Participant
Participants with 1 or 2

Participants with No

Friends who were

Friends who were

Interviewed

Interviewed

(n=102)

(n=97)

14.6 – 103.0

12.1 – 71.6

38.9, 16.3

35.5, 16.0

39.2

36.3

T

Df

1.0

183

Median Household
Income (Scaled by
dividing raw
income score by
1,000)
Range
Mean, SD
Median

* Significant at p<0.05, ** Significant at p<0.01.
Missing Data, Normality, and Outliers
Missing Data and Outliers
The statistical software, MPlus, was utilized in analyzing the data. This package can address
missing data through the full information maximum likelihood algorithm (FIML). FIML is able
to handle up to 25% missing data and produce accurate coefficient estimates and model fit
indices (Enders & Bandalos, 2001).
In order to accommodate the fact that not all participants had friend data, as mentioned
above, only the 102 participants who had close friend data were utilized in analyses pertaining to
45

peer influence. The full sample of 199 participants was used in analyses which did not pertain
to peer influence (i.e., exploratory analyses of sexual partner influence).
It is important to note that, in MPlus, FIML can only be used on dependent variables.
Therefore, cases that had missing data on the covariates (―age‖, ―who the adolescent lived with‖,
and ―median income‖ were allowed to covary in some of the models and constrained to 0 in
others) were excluded from particular model runs. Therefore, final models had fewer cases than
102 (for the peer risk model) and 199 (for the partner risk model).
In relation to outliers, descriptive statistics and histograms were utilized to identify
outliers at the univariate level. Individual scale scores which were three standard deviations from
the mean were considered and examined as outliers. Multivariate outliers were identified
utilizing Cook’s distance (measures the effect of deleting a single observation) and Mahalanobis
distance (calculates the distance of particular scores to the cluster of cases). (Note that these
multivariate outlier detection analyses were conducted separately for the peer model and partner
model.) Overall, 10 cases were identified as univariate and/or multivariate outliers. Analyses
were conducted excluding and including the outliers to determine changes in results and decide
whether to exclude outlying cases. Overall, results did not change significantly after excluding
these cases. This indicated that the outliers were not exerting undue influence on model results,
therefore, the cases were retained in the analyses.
Data Analytic Strategy
As mentioned above, structural equation modeling was utilized in assessing the fit of the
proposed conceptual model onto the observed data. As is the norm in conducting structural
equation modeling, analyses occurred in two main steps: estimation of the measurement model
(confirmatory factor analysis) and structural equation modeling (Kline, 1998). For each of these
46

two processes (CFA and path analysis), four basic steps were conducted: (1) defined a model
(specification), (2) MPlus software used observed correlations or covariances to estimate
parameters, (3) the model fit was tested and, where needed, (4) the model was modified.
As mentioned above, confirmatory factor analysis was utilized to test involved parenting
as a latent variable with 8 indicators (2 for support and 6 for monitoring). All other variables,
were treated as observed or measured variables
Model Fitting Process
MPlus software was used to estimate parameters for each of the proposed paths,
including control variables. This produced a baseline model. Several fit indices were examined
to determine the degree to which the conceptual model ―fit‖ the data. In other words, fit indices
provided the significance and magnitude of the hypothesized relationships. As recommended by
Kline (1998), the following fit indices were evaluated in determining model fit: Chi Square
Statistic (CMIN), Comparative Fit Index (CFI) and Root Mean Square of Error Approximation
(RMSEA). The CMIN compares the observed covariance structure against the hypothesized
model structure. Generally, a good fit is indicated when the CMIN is non-significant. This
indicates that the researcher hypothesized model and the model of the observed data are not
significantly different. However, it is important to note that the chi square statistic depends on
the assumption of multivariate normality, and is sensitive to sample size and Type II errors.
Therefore, the other fit indices were utilized in combination with the CMIN to determine model
fit before any decisions on modifications were made. The comparative fit index (CFI) indicates
the proportion of improvement in the fit of the hypothesized (researcher) model in comparison to
the null model (one in which the observed variables are assumed to be uncorrelated). The CFI
accounts for the number of parameters estimated. Importantly, the CFI is sensitive to magnitude
47

of correlations; if correlations are low it will be low as well. (Reasonable fit for CFI is .9 or
above.) Finally, the RMSEA was utilized to assess model fit. The RMSEA is based on the chi
square statistic. It accounts for the degrees of freedom in the model. Generally, an RMSEA of
.06 or less indicates good fit.
After examining model fit, pathways which were non-significant were set to zero in order
to increase model fit and attain a more parsimonious model. Next, modification indices were
assessed. Larger modification indices signal a better model fit if that particular parameter is
added to the model (Kline, 1998). Modifications that could potentially improve model fit were
considered. If theoretically possible, modifications were made. If modifications were made,
model fit indices were re-examined to determine the degree to which the model improved. This
process resulted in a final structural model.
It is important to note that in some cases it is not possible to attain model fit indices for
particular types of structural equation models. One such case is when one is estimating model fit
for a model involving an interaction with a latent variable. When estimating model fit for a
model involving a latent interaction variable, MPlus software (as well as other software
programs) is unable to provide traditional model fit statistics including Chi-Square statistic, CFI,
and RMSEA (Muthen, 2010). The reason why this is the case is that, in order to run this sort of
model, it is necessary to utilize model estimation (maximum likelihood estimation with robust
standard errors or MLR) that allows random slopes and intercepts (where the variance of the y
variable varies with the values on x). This precludes the estimation of chi square and related
statistics (Muthen, 2010). Instead, it is necessary to use a nested model technique where one
estimates model fit for a particular model, then sets one or more parameters to zero and tests

48

model fit again. After that, the resulting model statistics are utilized to conduct a log-likelihood
difference test, which is in fact chi-square distributed (Muthen, 2010).
Analysis of the traditional fit statistics was utilized in fitting the CFA model, as described
above. However, a nested model technique, utilizing log-likelihoods for the null and alternative
models were used in testing all of the path models, given that all included latent interaction
terms.

49

RESULTS
In summary, this study tested a conceptual model of parent and peer influences on sexual
risk taking and STD status of African American youth. Exploratory analyses were also
undertaken to test a conceptual model of parent and sexual partner influences on sexual risk
taking and STD status of African American youth. The following sections discuss the results of
this study. First, the sample characteristics in relation to the dependent variables are summarized.
Next, the model testing process is discussed, by research question, and the results of each set of
analyses are described.
Sample Characteristics
Among the 199 youth in the sample (all sexually active at time 2), the average age of first
sexual intercourse was 14.9 years of age. At Time 1 (a year before Time 2), 82.9% of that
sample had initiated sexual intercourse (n=165) and 66% of the sample (n=132) reported sexual
activity in the last 3 months. Additionally, at baseline, the mean number of partners in the last 3
months was 1.56 and the median was 1 (n=132). At Time 2, the mean number of partners in the
last 3 months was 1.75 and the median was 1 (n=199).
Overall, most youth reported some degree of sexual risk taking or non-condom use. At
Time 1, 39% of the participants who reported sexual activity in the past 3 months, reported that
they used condoms every time with all partners, while 60% reported that they had used condoms
inconsistently at some point in the last 3 months with at least 1 partner. At Time 2, 47% said that
they used condoms every time with all partners, while 50% had used condoms inconsistently at
some point in the last 3 months with at least 1 partner. Time 1 and 2 condom use behaviors were
also significantly correlated with one another (r = 0.329, p < 0.01).

50

Additionally, 75% (n=124) of sexually active participants at Time 1 reported that they
had been tested for an STD at some time in their lives. Out of those participants 124 participants
7.2% (n=9) said that they had a positive STD test the last time that they were tested.
A total of 36 participants (18%) reported that they had at least 1 partner who was 5 years
older than they were. A majority of the participants reporting older partners were girls, 29 out of
36 (81%).
Finally, on average, the index youth were somewhat ―riskier‖ than their friends. Index
participants had an average sexual risk score of 1.3 (SD = 1.6) with a median of 1.0 (score of 1=
used condoms most of the time). Meanwhile, friends had an average sexual risk score of 1.0 (SD
= 1.6) with a median of 0 (score of 0 = friend was a virgin or used condoms every time).
However, it is important to note that this is likely related to selection criteria. While all index
participants in this study had to be sexually active, their corresponding friends did not.
Confirmatory Factor Analysis: Involved Parenting as a Latent Construct
Before testing the hypothesized relationships amongst the variables, involved parenting
was tested as a latent variable with 8 indicators (2 for support and 6 for monitoring). Figure 2
shows the hypothesized model.

51

Figure 2. Confirmatory Factor Analysis of Involved Parenting as a Latent Construct
1
1
1
1
1
1
1
1

PM1
PM2
1

PM3
PM4

Involved
Parenting

PM5
PM6
PS1
PS2

The results of the confirmatory factor analysis for the hypothesized model demonstrated
that the model was an improvement over the MPlus generated baseline or independence model
(an unrestricted model with a free covariance matrix, in other words a model in which it is
assumed that there are no relationships between the variables). While the baseline model had a
chi-square of 361(28) (p<0.0000) the hypothesized model had a chi–square of 96.580 (20)
(p<0.0000). However, examination of the other fit indices (as suggested by the chi-square of the
hypothesized model) confirmed that the investigator model was not a close fit to the covariance
structure of the data. The results for the hypothesized model are included in Table 2.

52

Table 2. Confirmatory Factor Analysis: Model Fit Statistics
Model

Chi-square (df)

CFI

RMSEA

0.770

0.139

0.961

0.070

0.996

0.022

96.580 (20)
Hypothesized Model

p<0.000

Parental Monitoring ―try‖ Items
Allowed to Correlate to Each Other,
―really‖ Variables Allowed to
Correlate to Each Other and Support
Items Allowed to Correlate to Each

25.829 (13)

Other

P=0.0179

Parental Monitoring ―try‖ Items
Allowed to Correlate to Each Other,
―try‖ Items Allowed to Correlate with
Corresponding ―really‖ Items and
Support Items Allowed to Correlate to

14.216 (13)

Each Other

p=0.3588

The next step was examination of the modification indices. These revealed that allowing
some of the residual variances of the indicators to correlate could improve the model fit.
Therefore, from a theoretical perspective, it was decided the following modifications would be
made: (1) the residual variances of those items that related to ―trying‖ to monitor your children
would be allowed to correlate to each other, (2) the residual variances of those items that related

53

to ―really‖ monitoring your children would be allowed to correlate to each other and (3) the
residual variances of the two items that related to parental support would be allowed to correlate
to each other. These modifications seemed justified given that each set of variables related to a
distinct subconstruct of involved parenting (intention to monitor, actual monitoring, and
support). As is clear from the fit indices, this model was a much closer fit to the covariance
structure of the dataset. Once again, modification indices were examined. These revealed that
the correlations amongst the variables relating to ―trying‖ to monitor your children were indeed
significantly related to one another, however, the variables relating to ―really‖ monitoring your
children were not significantly related to one another. It was also noted that some of the parental
monitoring items which related to ―trying‖ to monitor were significantly correlated with the
corresponding items on ―really‖ monitoring (i.e., ―How much do your parents try to know about
what you do with your free time?‖ was significantly correlated with ―How much do your parents
really know about you do with yoru free time?‖). Therefore, a third model was tested with the
following modifications to the investigator model: (1) the residual variances of those items that
related to ―trying‖ to monitor your children were allowed to correlate to each other (2) the
residual variances for each of the ―try‖ parental monitoring variables were allowed to correlate to
their corresponding ―really‖ variables and (2) the residual variances to the two support variables
were allowed to correlate to each other. Once again, this model seemed theoretically possible
given that attempting to know about a particular situation that your child is involved in would
correlate to you actually knowing about it. As is evident from the fit indices included in Table 2
this produced the best fitting model. Figure 3 below illustrates all of the parameter estimates and
error covariances for this model.

54

0.444***

PM1

1

0.443***
0.334***

PM2
0.507***

-0.013

0.356***

Figure 3. Confirmatory Factor Analysis of Involved Parenting: Final Model (Standardized Parameter Estimates)

1
1
0.371***
PM3

0.238**

1

Involved
Parenting

0.741***
PM4
1

0.720***

0.123

a

0.275**
PM5

0.646***

0.220**

1
PS1

PS2

1

PM6
1

1

0.237**
55

Figure 3. (cont’d)
a

N=199. * Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001, Significant at p<0.10

56

Path Analysis Model Results by Research Question
In order to answer the research questions and exploratory questions of interest a series of
path analysis models were constructed and tested. For each set of analyses several steps were
undertaken. As a first step, before running the theoretical models for each analysis, a version of
the model that did not include the latent interaction term was run. This is recommended as a
preliminary step when running models for which it is not possible to attain traditional fit indices
(Muthen, 2010). If the model is a good fit, then one goes on to test whether the interaction in
question is significant. Secondly, parameter estimates were examined in order to assess the
direction and strengths of the relationships between the variables of interest. Note that in those
models including latent interaction variables it was not possible to attain standardized parameter
estimates, therefore, unstandardized parameter estimates were examined and reported (standard
errors are provided below to allow for interpretation). Given the greater ease of interpretability
of standardized parameter estimates, these were provided whenever available (i.e., in those
models that did not include the latent interaction terms).
Then, the paths from control variables to the dependent variable that did not reach
statistical significance were trimmed (by setting them to 0) one-by-one, in distinct iterations of
the model. (That is, if two control variable paths were found to be non-significant when the full
model was run, one of them would be trimmed, the model would be run again, and the
parameters re-examined. If warranted, the second non-significant path would be trimmed as well
in a next step. Trimming non-significant variables one-by-one allows the researcher to diagnose
accurately the cause of any changes in model fit and parameter estimates from model to model.)
Once a more parsimonious model was achieved through the trimming of non-significant paths
between control variables and the dependent variable, significance tests utilizing the log57

likelihood values of the new model and subsequent nested models (both the full and nested
models must be run with the same MLR estimator in order to conduct these) were conducted.
The purpose of these tests is to assess whether the full model (including the interaction) or the
simpler, nested model (excluding the interaction) should be chosen as a closer fit to the data
structure.
The formula to compute the TRd statistic is as follows (Satorra & Benteler, 1999):
cd = (p0 * c0 - p1 * c1)/(p0-p1)
TRd = -2*(L0 - L1)/cd
df = p1-p0
Where cd is the difference test scaling factor and TRd is the chi-squared distributed test
statistic. The loglikelihood values, scaling factors and number of free parameters are obtained
from the MPlus model outputs. Completed formulas for the model tests are included below.
RQ #1: Is involved parenting a significant buffer against high risk sex for adolescents whose
peers are sexually risky?
The main aim of this study was to test whether involved parenting could moderate the
relationship between sexually risky peers and sexual risk taking in this sample of African
American youth. Figure 4 depicts the first path model that was tested which, as described above,
excluded the latent interaction term between involved parenting and sexually risky peers. (Note
that this model also estimated the correlations amongst the residual variances of the observed
exogenous variables although these are not shown in Figure 4 below.)
It is important to note that FIML can only be used to handle missing data on dependent
variables, not on covariates. Therefore, cases that had missing data on the covariates (―age‖,
―who the adolescent lived with‖, and ―median income‖ were allowed to covary in this model)
58

were excluded from the model runs (even if the parameter was set to 0). The model below and
subsequent ones had 96 cases instead of 102.
Figure 4. Structural Equation Model of Involved Parenting and Peer Sexual Risk: Interaction
Term Excluded (Standardized Parameter Estimates)
1

1
Lives
with 1
parent or
guardian

1
1
Age
Involved
Parenting
(Monitoring
and Support)

-0.097

b

0.189

Sexually
Risky Peers

b

-0.042

0.038

0.081

Sexual
Risk
Taking
1

0.186

N=96. a p = 0.061 ,

Census
Block
Income

L

a

1

Lives
with 2
parents or
guardians

1

p = 0.067

Overall, this model had excellent model fit, χ2 (55) = 42.28 and was non-significant. The
CFI had a score of 1.00 and the RMSEA had a score of 0.00, both indicating good model fit.
However, the hypothesized relationship between higher levels of involved parenting and lower
levels of sexual risk taking did not reach statistical significance. On the other hand, the path
from sexually risky peers to sexual risk taking approached significance, B = 0.186, (p=0.061).
In other words, those youth who had sexually risky peers at time 1 were more likely to engage in
higher levels of sexual risk taking at time 2, while accounting for the effects of age, whether the
adolescent lived with one or two parents or guardians, and median census block income on
sexual risk taking. The path from age to sexual risk taking approached significance as well, B =
59

0.189 (p=0.067), suggesting that older youth were more likely to engage in sexual risk taking or
non condom use than younger youth. None of the paths from the other control variables to sexual
risk taking approached statistical significance. Table 3 below includes the unstandardized
parameter estimates and levels of significance.
After determining that the fit indices suggested that this model, excluding the latent
interaction term fit the data relatively well, a series of models was run (beginning with the
theoretical or investigator model) which included all of the hypothesized relationships, including
the latent interaction term. Table 3 below includes the unstandardized parameter estimates for
the different model runs. As mentioned above, it is not possible to attain traditional fit indices
and standardized estimates for these sorts of models. Therefore, unstandardized parameter
estimates and their p-values were examined in order to evaluate model fit.

60

Table 3. Peer Sexual Risk Model: Unstandardized Parameter Estimates and Significance Values
Path Models
Latent Interaction Term Omitted

Investigator Model

Non-significant

Nested Model

Control Variables
Trimmed
Est.

SE

Std. Est

Est.

SE

Est.

SE

Est.

SE

IP →SR

-0.086

0.135

-0.097

-0.139

0.256

-0.086

0.253

-0.075

0.164

SRP→SR

0.044

a

0.024

0.186

0.045

a

0.026

0.047

a

0.025

0.047

a

0.026

Na

Na

Na

-0.059

0.139

-0.011

0.136

set to 0

set to 0

Age→SR

0.042

0.023

0.189

0.044

a

0.024

0.045

a

0.024

0.045

LW1PorG→SR

-0.026

0.189

-0.042

-0.004

0.286

LW2PorG→SR

0.024

0.024

0.038

0.036

0.293

MedInc→SR

0.002

0.002

0.081

0.002

0.002

IPxSRP→SR

a

a

a

0.024

N=96. p < 0.10. IP = Involved Parenting, SR = Sexual Risk Taking, SRP = Sexually Risky Peers, LW1PorG = Living with one
parent or guardian, LW2PorG = Living with 2 parents or guardians, MedInc = Median Census Block Income

61

As can be seen in Table 3 above, none of the direct effects in the investigator model were
statistically significant. Two of the direct effects, however, approached significance. Those were
the paths from age to sexual risk taking (0.044, p = 0.070) and sexually risky peers to sexual risk
taking (0.045, p = 0.086). In other words, adolescents who were older were more likely to
engage in sexual risk taking and adolescents whose peers engaged in higher levels of sexual risk
taking at time 1 were more likely to engage in sexual risk taking at time 2.
Next, non-significant paths from control variables to sexual risk taking were trimmed
one-by-one starting with a model in which the two variables regarding who the adolescent lived
with were trimmed and then a model where the path from median income to sexual risk taking
was trimmed. Parameter estimates remained very similar to the original model in these iterations.
See the third column of Table 3 for the parameter estimates of the model in which the paths for
the non-significant control variables were trimmed. Once again, the path from age to sexual risk
taking approached significance (see Table 3) and was, therefore, left in the model.
After this, a version of this more parsimonious model was run where the interaction
between involved parenting and sexually risky peers had been set to 0. This new model, where
the interaction parameter is set to 0 is said to be nested within the previous model. See the last
column of Table 3 for the parameter estimates from this model. Similar to previous iterations,
the paths from age to sexual risk taking and sexually risky peers to sexual risk taking approached
statistical significance.
Once these models were run, a chi-square difference test utilizing the log-likelihoods
values of each model was estimated. Table 4 below summarizes this test.

62

Table 4. Chi-Square Test for Involved Parenting and Peer Risk Model: Testing significance of
interaction between involved parenting and sexually risky peers
Loglikelihood
Scaling
Number of Free
p*c
Value

Correction

Parameters (p)

Factor (c)
Nested Model

-712.962

1.089

36

39.204

Full Model

-712.956

1.106

37

40.922

Difference Test Scaling (cd) = 1.718
Chi-Square Difference Test (TRd) = 0.01 (df = 1)
As can be seen above, the TRd value obtained was 0.01. Given that for one degree of
freedom the TRd has to be at least 3.84 (at a p<0.5 level of significance) in order for the test to
be significant, it was determined that the test was not significant and, therefore, the simpler
model which did not contain the interaction, should be chosen. In other words, including the
interaction term in the structural equation model does not improve model fit indicating that
involved parenting did not moderate the relationship between sexually risky peers and sexual
risk taking in this sample.

2

RQ #2: Does gender moderate the relationship between peer sexual risk and sexual risk taking?
Another objective of this study was to test whether gender moderated the relationship
between peer sexual risk and sexual risk taking. Although specific hypotheses were not drawn,
2

The analyses interaction analyses described above (testing whether involved parenting could moderate
the effects of risky peers on sexual risk taking) were also conducted while controlling for the effects for
sexual risk taking at time 1. However, when sexual risk taking at time 1 was controlled for, model results
did not change significantly.

63

given the contradictory literature on this topic, it was expected that this relationship would differ
by the participants’ gender.
Given the small sample size and non-significant interaction terms, analyses focused on
testing whether gender moderated the relationships between the two direct effects in the model
(involved parenting and sexually risky peers) and sexual risk. Firstly, the only direct effect that
approached statistical significance (sexually risky peers) was tested for moderation. In order to
test this, an observed interaction term was constructed between the variables: ―peer sexual risk‖
and ―gender‖.
Next, the nested model technique outlined above was used to test the significance of this
interaction. Figure 5 below illustrates the model that was tested along with the corresponding
unstandardized parameter estimates.

64

Figure 5. Structural Equation Analysis of Involved Parenting and Peer Sexual Risk: Gender
Moderation Analyses (Sexually Risky Peers X Gender)
1

1
Involved
Parenting
(Monitoring
and Support)
1

Age
a

0.051
Sexually
Risky Peers

-0.013
-0.008

Involved
Parenting X
Sexually
Risky Peers

1

0.044 L
Sexual
Risk
Taking
1

-0.089

1

Sexually
Risky
Peers X
Gender

a

N = 96. p = 0.058
The path from the interaction to sexual risk taking did not emerge as significant as can be
seen in Figure 5. In fact, the only relationship that approached significance was age to sexual risk
taking. Unlike previous iterations of the model, the relationship between sexually risky peers
and sexual risk taking did not approach significance (p=0.104) when the interaction term was
included in the model.
After this, a model was run in which the interaction set had been set to 0. A chi-square
test using the loglikelihood values for the models was calculated afterwards. The test was not
significant, (see Table 5 below), meaning that the simpler model, excluding the interaction is a
better fit for the data. In other words, gender does not appear to moderate the relationship
between sexually risky peers and sexual risk taking in this particular sample.
65

Table 5. Chi-Square Test for Involved Parenting and Peer Risk Model: Testing significance of
interaction between sexually risky peers and gender
Loglikelihood
Scaling
Number of Free
p*c
Value

Correction

Parameters (p)

Factor (c)
Nested Model

-712.956

1.106

37

40.922

Full Model

-712.934

1.100

38

41.8

Difference Test Scaling (cd) = 0.878
Chi-Square Difference Test (TRd) = 0.05 (df = 1)
Secondly, model tests were run to determine whether gender moderated the relationship
between involved parenting and sexual risk taking in this sample. In order to test whether
participant gender moderated the relationship between involved parenting and sexual risk taking,
a latent interaction term was constructed between involved parenting and gender. That term was
then entered into the model and regressed onto sexual risk taking. As can be seen from Figure 6
below the interaction term was not significant. Only the paths for sexually risky peers to sexual
risk taking and age to sexual risk taking approached significance.

66

Figure 6. Structural Equation Analysis of Involved Parenting and Peer Sexual Risk: Gender
Moderation Analyses

1

1

Age

Involved
Parenting
(Monitoring
and Support)

a

-0.209
a

1

1

0.047

Sexually
Risky Peers

-0.018

Involved
Parenting X
Sexually
Risky Peers
1

0.044 L
Sexual
Risk
Taking
1

0.158

Involved
Parenting
X Gender

a

N=96. p = 0.058
Subsequently, a model was run where the interaction term between involved parenting
and gender had been set to 0. A chi-square difference test utilizing the log likelihood values of
each model were calculated subsequently. The results to that chi-square difference test are
included in Table 6 below. The test was not significant. In other words, gender does not appear
to moderate the relationship between involved parenting and sexual risk taking in this sample.

67

Table 6. Chi-Square Test for Involved Parenting and Peer Risk Model: Testing significance of
interaction between involved parenting and gender
Loglikelihood
Scaling
Number of Free
p*c
Value

Correction

Parameters (p)

Factor (c)
Nested Model

-781.279

1.076

39

41.964

Full Model

-781.146

1.078

40

43.12

Difference Test Scaling (cd) = 1.156
Chi-Square Difference Test (TRd) = 0.23 (df = 1)
RQ #3: Does higher risk taking predict higher incidence of STD diagnosis?
Another aim of this study was to test whether higher levels of sexual risk taking predicted
STD diagnosis. It was hypothesized that higher levels of sexual risk taking would predict STD
diagnosis (both measures were taken at time 2). In order to test this hypothesis, the model shown
below in Figure 7 was attempted utilizing maximum likelihood estimation. However, model
convergence issues, likely due to the small sample size (n=96) made it impossible to attain
reliable model estimates. Therefore a MonteCarlo estimation utilizing MLR was used in
estimating the model. (Monte Carlo studies are analyses in which a large number of samples are
drawn from the data. Each sample is then fit onto the hypothesized model (Muthen, 2002).)
This allowed for the estimation of parameter estimates and odds ratios, although model fit
statistics were not attained given that MLR (necessary when including an interaction in the
model) was used. Similar to previous models, several of the hypothesized paths approached

68

statistical significance (i.e., age and sexually risky peers to the sexual risk taking). Those are
shown in Figure 7.
Figure 7. Structural Equation Model of Involved Parenting and Sexually Risky Peers:
Mediational Analysis
1
1
Age
Involved
Parenting
(Monitoring
and Support)

3.324

b

0.044

0.381

Sexually
Risky Peers

4.349*
*

Sexual
Risk
Taking

a

0.008
1

-0.421

c

-0.076
0.045

1

L

1

1

-0.059

STD

Involved
Parenting X
Sexually
Risky Peers
a

b

c

N=96. Unstandardized regression coefficients are shown. p = 0.066, p = 0.060, p = 0.091.
* Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001.
(Note that the regression coefficients associated with the dichotomous dependent variable
provide the change in the z-score or probit index for a one unit change in the corresponding
predictor variable.) The path from sexual risk taking to STD status was significant at p<0.01.
However, upon examining the results of the logistic regressions, it became clear that there were
estimation problems with the model. According to these results, for every one unit increase in
sexual risk taking, participants were 77.438 times more likely to report an STD. This very large
odds ratio pointed to possible statistical separation problems. Separation is when the outcome
variable separates a predictor variable or a combination of predictor variables to certain degree
(Allison, 2008). In other words, a dependent variable is said to be separated when a particular
69

independent variable perfectly predicts a particular outcome. A cross-tabulation of sexual risk
scores against STD status confirmed quasi-complete separation. Nearly all participants who had
an STD at time 2 (8 out of 10) reported a score of 2 or higher in relation to condom use. (In
order to get a score of 2, a participant must have used condoms ―most of the time‖ with 2
partners or ―a few times‖ with 1 partner.) Therefore, STD status is said to separate (although not
perfectly) at a score 2> in the sexual risk taking scale. This lack of variability, in combination
with an unusual binary outcome variable and small sample size made the measurement of this
model unreliable. Given that a statistical treatment of separation issues is beyond the scope of
this dissertation, it was decided that hypotheses regarding mediation analyses would not be tested
at this time.
Exploratory Questions
EQ #1: Is involved parenting a significant buffer against high risk sex for adolescents whose
partners are typically sexually risky?
In addition to the research questions of interest, this dissertation study also sought to
answer whether involved parenting could buffer against high risk sex for adolescents whose
partners were typically sexually risky. In order to test this question, the same model-testing steps
outlined above were conducted (i.e., starting with a model that excluded the interaction terms,
proceeding with the investigator model, trimming non-significant control variables, and testing
of significance of interaction terms through a nested model technique and log-likelihood chisquare tests). Table 7 includes the parameter estimates and levels of significance for the path
models that were tested.

70

Table 7. Sexually Risky Partners Model: Unstandardized Parameter Estimates and Significance Values
Path Models
Interaction Terms Omitted

Latent Interaction Term Omitted

Investigator Model

Est.

SE

Std. Est

Est.

SE

Std. Est

Est.

SE

-0.058

0.098

-0.057

-0.055

0.100

-0.054

-0.055

0.110

RSP→SR

0.127***

0.032

0.290***

0.124***

0.032

0.284***

0.124***

0.035

PR→SR

0.112

0.078

0.104

0.108

0.080

0.100

0.109

0.81

IPxRSP→SR

Na

Na

Na

Na

Na

Na

0.023

0.149

RSPxPR→SR

Na

Na

Na

0.055

0.118

0.032

0.060

0.136

Age→SR

0.030*

0.014

0.150*

0.029*

0.014

0.148*

0.029*

0.14

LW1PorG→SR

-0.090

0.110

-0.151

-0.090

0.111

-0.150

-0.096

0.130

LW2PorG→SR

-0.013

0.114

-0.021

-0.014

0.115

-0.022

-0.020

0.133

MedInc→SR

0.001

0.001

0.079

0.001

0.001

0.081

0.001

0.001

IP →SR

N = 183. * Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001.
IP = Involved Parenting, SR = Sexual Risk Taking, PR = Perception of Risk, LW1PorG = Living with one parent or guardian,
LW2PorG = Living with 2 parents or guardians, MedInc = Median Census Block Income

71

Table 7 (cont’d)
Path Models
Non-Significant Control Variables

Nested Model 1

Nested Model 2

Trimmed
Est.

SE

Est.

SE

Est.

SE

-0.086

0.121

-0.074

0.112

-0.074

0.112

RSP→SR

0.120***

0.034

0.120***

0.034

0.120***

0.034

PR→SR

0.077

0.083

0.081

0.082

0.081

0.082

IPxRSP→SR

-0.103

0.199

set to 0

set to 0

set to 0

set to 0

RSPxPR→SR

0.066

0.140

0.088

0.130

0.088

0.130

Age→SR

0.031*

0.014

0.032*

0.014

0.032*

0.014

IP →SR

LW1PorG→SR
LW2PorG→SR
MedInc→SR
N = 183. * Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001.
IP = Involved Parenting, SR = Sexual Risk Taking, PR = Perception of Risk, LW1PorG = Living with one parent or guardian,
LW2PorG = Living with 2 parents or guardians, MedInc = Median Census Block Income

72

As a first step, a version of the model that excluded both interaction terms (i.e., the
observed interaction term between ―risky sexual partners‖ and ―perception of risk‖ and the latent
interaction term, a cross product of latent variable ―involved parenting‖ and observed variable
―risky sexual partners‖) was run in order to assess main effects and attain model fit statistics.
This model is shown in Figure 8 below. Once again, given that FIML can only be used to handle
missing data on dependent variables and not covariates, the models below have 183 cases
(instead of 199). The cases with missing data on the covariates were excluded (even when those
covariate parameters were trimmed by setting them to 0).
Figure 8. Structural Equation Model of Involved Parenting and Risky Sexual Partners:
Interaction Terms Excluded (Standardized Parameter Estimates)

1
Involved
Parenting
(Monitoring
and Support)
-0.057

1

1
Age

0.150*

1

Living
with 1
Parent or
Guardian

1

1

Living
with 2
Parents or
Guardians

Median
Census
Block
Income

-0.151
-0.021

Risky Sexual
Partners

0.290***
0.104

Perception of
Risk

Sexual
Risk
Taking

0.079

1

1

N=183. * Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001.
Overall, the fit statistics indicated that the model had a good fit to the data, χ=60.941 (62)
and was non-significant. The CFI was 1.0 and the RMSEA was 0.000, showing excellent model

73

fit as well. As can be seen in Figure 8, the relationship between involved parenting and sexual
risk taking did not emerge as significant. Additionally, the path from sexually risky partners to
sexual risk taking was significant at p< 0.001. In other words having sexual partners who were
higher risk at time 1 significantly predicted sexual risk taking at time 2. As with previous
models, the path from age to sexual risk taking was also significant (at p<0.05), indicating that
older youth were less likely to use condoms than younger youth. It is important to note that risky
partners and perception of risk were highly correlated (-0.367, p<0.000). (Their error terms were
allowed to correlate in this and subsequent model runs in order to reflect this correlation.)
Therefore, it is not surprising that the hypothesized relationship between ―perception of risk‖ and
―sexual risk taking‖ did not emerge as significant. In other words, when risky partners was taken
into account, perception of risk did not contribute to explaining levels of condom use because
these variables provided redundant information.
Next, a model was run that included the hypothesized interaction term between ―risky
partners‖ and ―perception of risk.‖ Although empirical findings from the previous model
suggested that finding a significant relationship was very unlikely (given the high correlation
amongst the two variables) the model was run given that it was of theoretical interest to this
study. As expected, the relationship between the interaction term and sexual risk taking did not
emerge as significant (see Table 7). Model fit statistics, however, did show good model fit,
χ=62.280 (69) and was non-significant. The CFI was 1.0 showing good model fit. The RMSEA,
however, was 0.068, which is slightly above the 0.06 conventional cut off point for good fit.
Next, a series of models starting with the investigator model was run. Similar to the
model in which the interaction term was omitted, the path from sexually risky partners to sexual
risk taking was significant at p<0.001 (B= 0.124) and the path from age to sexual risk taking was

74

significant at p<0.05 (B= 0.029). It is also important to note that this model included the
estimation of the correlations amongst the residual variances of: (1) the components of the
observed interaction term (risky sex partners and perception of risk), (2) the components of the
latent interaction term (involved parenting and risky sex partners), and (3) the correlation
between the residual variances of age and involved parenting. The first two parameters were
estimated to reflect collinearity. The third correlation (between the error terms of age and
involved parenting) was included because initial analyses demonstrated that, in this larger
sample (although not in the sample of 96 that was used for the peer sexual risk analyses), their
error terms were significantly correlated. All of these correlations amongst the residual
variances of these variables were statistically significant.
Non-significant control variables were trimmed in subsequent steps. As can be seen in
Table 7, model estimates shifted only slightly through the different iterations of the model.
Finally, the significance of the latent interaction (i.e., does involved parenting moderate
the effects of sexually risky partners on sexual risk taking?) was tested by calculating chi-square
difference test, utilizing their loglikelihood values. The values for this test are shown in Table 8
below.

75

Table 8. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing
significance of interaction between involved parenting and risky sexual partners
Loglikelihood

Scaling

Number of Free p * c

Value

Correction

Parameters (p)

Factor (c)
Nested Model

-1858.303

1.064

50

53.2

-1858.118

1.071

51

54.621

(Interaction set
to 0)
Full Model

Difference Test Scaling Factor (cd) = 1.421
Chi-Square Difference Test (TRd) = 0.26 (df=1)
The test was not significant, indicating that the simpler model, excluding the interaction
term is a better fit for the data. In other words, involved parenting did not appear to moderate the
effects of sexually risky partners on sexual risk taking in this sample.
Overall, the most important determinant of sexual risk taking in this set of models was
risky sexual partners. As noted above, this construct is often conceptualized as a dependent
variable, however, the above analyses demonstrate the important impact that this partner variable
can have on non-condom use when conceptualized as a predictor.
EQ #2: Does gender moderate the relationship between risky sexual partners and sexual risk
taking?
This study also sought to test whether gender moderated the relationship between risky
sexual partners and sexual risk taking. Although, specific hypotheses were not formulated given
76

the lack of literature on this topic, it was expected that this relationship would differ between
boys and girls. Once again, given the small sample size and non-significant interaction effects,
analyses focused on the two main effects (sexually risky partners and involved parenting). First,
analyses were conducted to test whether gender moderated the one significant main effect that
emerged from the initial analyses (the path from sexually risky partners to sexual risk). In order
to test whether gender moderated the relationship between risky sexual partners and sexual risk
taking (i.e., was the effect of peer risk on sexual risk taking different for boys versus girls?), a
new interaction term was created between the variables ―risky sexual partners‖ and ―gender‖.
The model that was tested and the corresponding parameter estimates are shown in Figure 9
below.

77

Figure 9. Structural Equation Model of Involved Parenting and Risky Sexual Partners: Gender
Moderation Analyses (Risky Sexual Partners X Gender)
1

Involved
Parenting
(Monitoring
and Support)

1

-0.087
Age

1
0.164***

Risky Sexual
Partners

0.032*
Sexual Risk
Taking

0.081
Perception
of Risk

0.067

1

1

-0.082
Risky Sexual
Partners X
Perception of
Risk
1

Risky
Sexual
Partners X
Gender

1

-0.130

Involved
Parenting X
Risky Sex
Partners

N=183. * Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001.
As can be seen in Figure 9, the relationship between the interaction term ―Risky Sexual
Partners X Gender‖ did not emerge as significant. Table 9 below summarizes the chi- square
difference test conducted to determine whether the full model or the nested model were a better
fit for the data.

78

Table 9. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing the
significance of the interaction between risky sexual partners and gender
Loglikelihood

Scaling

Number of Free p * c

Value

Correction

Parameters (p)

Factor (c)
Nested Model

-1987.119

1.051

53

55.703

1986.183

1.055

54

56.97

(Interaction set
to 0)
Full Model

Difference Test Scaling Factor (cd) = 1.267
Chi-Square Difference Test (TRd) = 1.48 (df=1)
Once again, the chi-square difference test was not significant indicating that the nested
model, not including the interaction, was a better fit for the variance covariance structure of this
dataset. In other words, in this sample, gender did not appear to moderate the relationship
between peer sexual risk taking and sexual risk.
Next, analyses were conducted to test whether the relationship between involved
parenting and sexual risk was moderated by the gender of the participant. In order to test this
relationship, a latent interaction term between involved parenting and gender was created and
entered in the path model. Figure 10 below summarizes the results of that analysis.

79

Figure 10. Structural Equation Model of Involved Parenting and Risky Sexual Partners: Gender
Moderation Analyses

1

Involved
Parenting
(Monitoring
and Support)

1
-0.191
Age

1
0.118**

Risky Sexual
Partners

0.030*
Sexual
Risk
Taking
1

0.081
Perception of
Risk

0.063

1

0.156
Risky Sexual
Partners X
Perception of
Risk
1
Involved
Parenting X
Gender

1

-0.141

Involved
Parenting X
Risky Sex
Partners

N=183. * Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001.
Finally, the same model was run with the interaction term set to 0 and a chi-square
difference test between the full model and the nested model was calculated. The results for that
test are summarized in Table 10 below.

80

Table 10. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing
the significance of the interaction between involved parenting and gender
Loglikelihood

Scaling

Number of Free p * c

Value

Correction

Parameters (p)

Factor (c)
Nested Model

-1987.110

1.051

53

55.703

-1986.842

1.044

54

56.376

(Interaction set
to 0)
Full Model

Difference Test Scaling Factor (cd) = 0.673
Chi-Square Difference Test (TRd) = 0.82 (df=1)
The test was not significant.
EQ #3: Does perception of risk moderate the relationship between risky sexual partners and
sexual risk taking?
Finally, this study sought to test whether perception of risk moderated the relationship
between risky sexual partners and sexual risk taking. It was expected that perception of risk
would moderate the relationship between risky partners and sexual risk taking such that
individuals with higher levels of perceived risk would be more likely to use condoms. As can
be seen below this did not emerge as significant.

81

Table 11. Chi-Square Difference Test for Involved Parenting and Partner Risk Model: Testing
significance of interaction between risky sexual partners and perception of risk
Loglikelihood

Scaling

Number of Free p * c

Value

Correction

Parameters (p)

Factor (c)
Nested Model

-1858.303

1.064

50

53.2

-1858.118

1.071

51

54.621

(Interaction set
to 0)
Full Model

Difference Test Scaling Factor (cd) = 1.421
Chi-Square Difference Test (TRd) = 0.26 (df=1)

82

DISCUSSION
This dissertation study represents an important step in understanding the influence of
parents, peers and sexual partners on adolescent sexual risk behaviors. This research assessed
whether involved parenting buffered the impact of sexually risky peers and sexual partners on
adolescents’ level of sexual risk or non-condom use. Several key findings emerged.
Firstly, similar to the findings of recent epidemiological studies (Centers for Disease
Control and Prevention, 2009), adolescents in this sample did not use condoms consistently.
More than half of the youth reported inconsistent use of condoms within the last 3 months at
both waves. Moreover, 30% of adolescents reported having more than one sex partner in the 3
months prior to the study. The fact that unsafe sex and multiple partners were reported by a
significant proportion of these adolescents is concerning in that it suggests that current efforts to
increase condom use among youth have not fully succeeded. The sexual risk taking these youth
report is particularly worrisome in light of the fact that 18% of the sample either reported a past
STD or had a positive result on the STD test that they took as part of the study.
In addition, being older was significantly related (or, in some models, approached being
significantly related) to more instances of unprotected sex. In other words, older adolescents
were more likely than younger adolescents to report unprotected sexual intercourse. This
confirms previous research which has also linked older age to increased sexual risk taking
(Reitman, 1996; Santelli, Robin, Brener, & Lowry, 2001). There are several plausible
explanations for why adolescents have an increased likelihood of engaging in sexual risk as they
grow older. First, as adolescents age, they become more likely to establish longer-lasting sexual
partnerships (Collins, 2003) and evidence suggests teens are less likely to remain safe in their
sexual practices with longer term partners (Manning, Flanigan, Giordano, & Longmore, 2009).
83

Second, as youth get older they move from engaging in sex episodically to more routine
engagement, resulting in more opportunities to engage in unsafe behaviors. Finally, it is possible
that, as adolescents get older, they become more likely to view condoms as detracting from
sexual pleasure or spontaneity (Grossman et al., 2008), therefore reducing their levels of
consistent condom usage.
Study analyses also showed that there was a significant correlation between condom use
behaviors at time 1 and time 2 (12 months apart), suggesting that habit or previous behaviors
could play a significant role in condom use.
Additionally, many more girls than boys reported that they had a sexual partner in the last 6
months who was older than they were. This is important given that empirical evidence has
demonstrated that girls who are in relationships with older male partners are less likely to use
condoms consistently than girls who are in relationships with partners that are similar to them in
age (DiClemente et al., 2002; Manlove, Terry-Humen, & Ikramullah, 2006). Some research
suggests that, during adolescence, some youth internalize gender power asymmetries that impart
more interpersonal power and authority to boys than girls (Tolman, Spencer, Rosen-Reynoso, &
Porche, 2003). This adoption of imbalanced gender norms can result in a relationship
environment in which males control sexual decision making and girls do not feel comfortable or
able to negotiate safe sex practices with their partners (Wingood & DiClemente, 2000). This
power disparity is intensified when there is an age discrepancy amongst partners. Younger girls
are more likely to be in a disempowered position in relation to their male partner. In addition,
empirical evidence has shown that girls whose partners are older are more likely to perceive that
those partners will have a negative reaction to the idea of condom use, making those adolescents
less likely to suggest it (Wingood & DiClemente, 2000).

84

Analyses also demonstrated that the relationship between having risky peers at time 1 and
sexual risk taking at time 2 approached statistical significance. This result appears to support the
hypothesis that peers can be an important influence on sexual risk taking for youth. This
apparent link between friends’ sexual risk and condom use also seems to align with two previous
studies that have linked actual peer risk to sexual risk taking in adolescents (Bearman &
Bruckner, 1999; Henry, et al., 2007). Further work is necessary to determine the pathways
through which this relationship takes place. It is possible that adolescents communicate with one
another about their sexual behaviors, thereby setting perceived norms around non-condom use
within their social network. Alternatively, it is also possible that youth self select into groups of
friends who are similarly risky. Regardless, this apparent connection between peer risk and
sexual risk taking is a potentially important finding in that it suggests that if peers are an
important determinant of condom use behaviors, perhaps healthy peer norms can have a
protective effect over condom use. Some empirical evidence does, in fact, demonstrate that
healthy peer norms can have a protective impact over sexual risk behaviors in youth (Bearman &
Bruckner, 1999). Specifically, Bearman and Bruckner (1999) found that girls with all low-risk
friends (defined as high on academic orientation and low on risk behaviors including drinking,
smoking, and skipping school etc.) were less likely to have initiated intercourse or experienced a
pregnancy than girls who had friends who were high risk (defined as low on academic
orientation and high on risk behaviors). Preventive intervention work should focus on
establishing healthy norms in relation to condom use within the context of adolescents’ social
peer network.
Additionally, having sexually risky partners (defined in this study as being 5 years or older
than the participant, having had an STD in the past, and having had other sexual partners in the

85

last 6 months) significantly predicted lower use of condoms. Prior research has also found that
having partners who are more sexually risky can result in higher levels of sexual risk taking
(Staras, et al., 2009). As one might expect, the relationship between sexually risky partners and
non-condom use was stronger than the relationship between sexually risky peers and noncondom use, suggesting that partners and partner risk (a function of partner selection) may be the
more important microsystemic level influence over condom use in adolescents. Although some
theoretical and empirical evidence suggests that more distal social influences (like involved
parenting) can serve as a protective influence over adolescent sexual risk taking, the most
proximal determinant of condom use is the sexual relationship between the adolescent and his or
her sexual partner. In the end, this is largely what determines whether or not sexual risk or
protection takes place within the context of any particular sexual encounter. Evidence suggests
that there are multiple characteristics and processes involved in adolescent relationships that can
affect whether or not condom use occurs. For instance, controlling behavior and conflict in the
context of a sexual relationship has been linked to decreased condom use (Manning, et al., 2009).
A partner may want to use condoms but feel disempowered to suggest it, especially within a
high-conflict relationship. Interestingly, some evidence also suggests that positive relationship
characteristics and dynamics (e.g., love and enmeshment) can result in non-condom use
(Manning, et al., 2009). Partners who feel like they are in a committed relationship may not feel
the need to protect themselves with condoms. Overall, condom use occurs within a complicated
constellation of social forces. However, out of all of those, sexual partnership dynamics seem to
be the most important. Therefore, it is crucial that additional research focus on attempting to
understand the ways in which distinct combinations of partner and relationship characteristics
result in risky or protective sexual behavior trajectories for youth.

86

Additionally, the importance of partner characteristics in predicting condom use behaviors
implies that preventive interventions that teach adolescents how to make healthy partner choices
might be an effective way to decrease sexual risk behaviors, as well as STD and HIV infection in
youth. Partner selection may be an especially important point of intervention given that
unprotected sex with a non-infected partner does not result in a positive STD or HIV diagnosis.
A significant relationship also emerged between having risky sexual partners and perceiving
HIV/STD risk (signaling that youth were aware that their partners were sexually risky).
Therefore, not surprisingly, when ―risky partners‖ was accounted for in the model, perceiving
that one’s partner may put one at risk for HIV or an STD did not predict condom use (above and
beyond the variability accounted for by the ―risky partners‖ variable). This significant
relationship between having risky partners and perceiving STD/HIV risk highlights the fact that
youth seem to know that they are engaging in sexual risk, however, other aspects of the partner
relationship (such as attitudes towards condoms and relationship dynamics) may simply play
more important roles in whether the adolescent chooses to or is able to protect herself or himself
with condoms.
Contrary to expectations, involved parenting did not act as a buffer in the relationship
between sexually risky peers and risky sexual behavior. Indeed, there was no main effect of
involved parenting on sexual risk in these data. This result contradicted one study which found
that involved parenting protected youth from negative sexual health outcomes (East, et al.,
2006). However, East et al. (2006) examined the effects of protective parenting on pregnancy
(not condom use) in a sample of African American and Latina girls whose friends were sexually
risky. It is likely that the difference in outcome variable accounts for why this effect was not
found in this study. That is, pregnancy is a more general sexual health outcome that can be

87

avoided in multiple ways (such as not having vaginal intercourse and using birth control)
whereas condom use or requesting condom use of one’s partner is a very specific behavior that
can be deeply intertwined within relationship dynamics. It seems possible that, for youth who
are having sex, the mechanisms that drive condom use might be different than those that protect
girls from pregnancy. Additionally, it seems likely that those mechanisms that protect girls from
pregnancy may be more easily influenced by parents than are those that drive condom use. For
instance, involved parenting may protect young women from pregnancy by promoting delayed
sexual initiation. This mechanism is simply not a relevant protective factor for condom use for
youth who are already sexually active. It is also important to note that a close relationship to a
parent may not always be as protective as suggested by some of the literature. Low income
African American communities are disproportionately affected by STDs and HIV (Centers for
Disease Control and Prevention, 2007, March). It is possible that, within this sample, involved
parents may not have been strong models for condom use behaviors themselves.
The way in which parenting is measured may also bear on whether or not the impact of
parenting on children’s condom use behaviors can be detected. Although the full ―involved
parenting‖ latent construct did not have the expected relationship to condom use, some of the
items that were included in the construct were related to it. Specifically, one of the parental
monitoring items had a significant negative relationship to sexual risk taking (How much do your
parent(s) or guardian(s) really know about where you go at night?) and a second item had a
negative relationship that approached significance (How much do your parent(s) or guardians
really know about what you do with your free time?). Interestingly, when broken down by
gender, there was a significant relationship between the item ―How much do your parent(s) or
guardian(s) really know about where you go at night?‖ and sexual risk taking for boys but there

88

was no significant relationship between this item and sexual risk taking for girls. Similarly,
when examined separately by gender, ―How much do your parent(s) or guardians really know
what you do with your free time?‖ had a nearly significant negative relationship with sexual risk
taking for girls but not for boys. These findings suggest that, perhaps, parenting does in fact
affect boys’ and girls’ behaviors differently. However, these differences may be nuanced and,
therefore, undetectable when involved parenting was examined as a complete construct. (It is
possible that being aware of what ones’ adolescents are doing at night might have been a more
relevant item for boys given that adolescent boys are often given more liberty to go out
unsupervised during evening hours.)
It is interesting to note that both of the items that showed significant or nearly significant
relationships had to do with ―really‖ monitoring your children, not ―trying‖ to monitor them. It
is possible that being aware of what your adolescents are doing at night and with their free time
is more important for this sample of youth than simply ―trying‖ to know what they do at night
and with their free time. Some researchers have proposed that the protective effects of
monitoring are a function of whether the adolescent feels cared for (Longmore, et al., 2001).
Longmore and colleagues (2001) imply that ―trying‖ to monitor your adolescents might be as
protective as actually monitoring your adolescents in that both convey that parents care about
their kids. However, findings from the current study suggest that ―really‖ monitoring your kids
is what is protective. Huebner and Howell (2003) propose that the mechanism through which
parental monitoring has a protective effect over adolescent behavior is actually adolescent
disclosure of what they do with their unsupervised time. It is possible that the items on ―really‖
monitoring were better able to capture the degree to which adolescents disclosed their

89

whereabouts to their parents than the items on ―trying‖ to monitor your kids which related more
to parents’ intent to monitor.
Additionally, it is important to note that the parental monitoring items were measured on a 3item scale (response scale: not at all, a little and a lot), with most adolescents reporting high
scores on all variables. (For example, for all but one of the items, more than half of the sample
reported that they were monitored ―a lot‖, score of 3.) In other words, a majority of the sample
was very similar to each other on this scale. This is likely the combined result of the 3-item
response scale and a reflection of the fact that these particular youth were indeed similar to each
other in relation to parental monitoring. Regardless, this lack of variability and the restricted
range, can result in an inability to detect relationships that may in fact exist in a population. It is
possible, that some of the non significant findings (i.e., lack of relationship between involved
parenting construct and sexual risk taking) in this study were due to this phenomenon.
In relation to the parental support variables, neither of the two items was significantly related
to the outcome of interest (i.e., sexual risk defined as non-condom use). However, there was a
significant negative relationship between having a close relationship with your father and having
a lower score on the ―risky sex partners‖ scale. (―Parental support from female guardian‖ did not
have a statistically significant relationship with any of the sexual risk-related outcomes that were
examined.) Once again, when examined separately by gender, this relationship was only
significant for boys and not for girls. It is possible that male parental figures may be especially
well positioned to influence boys’ selection of sexual partners. A close relationship with a male
parental figure may open pathways of communication for parents to intervene if an adolescent
becomes involved with a potentially risky partner (e.g., someone who is considerably older).

90

In addition to parent-peer moderation analyses undertaken, this is the first study, to this
author’s knowledge, to assess whether involved parenting could buffer the effect of sexually
risky partners on sexual risk taking. Similar to findings on peer influences, parents had no
impact on the effects of risky sexual partners on youths’ own risk taking. As mentioned above,
involved parenting may be too distal of a force on condom use, in relation to risky sex partners (a
very proximal influence) in order to be an effective buffer against high risk sex. The protective
cascading effects proposed by Ecodevelopmental Theory between more distal forces (parents)
and proximal forces (partners) may simply be too diluted to make a significant difference by the
time that youth reach middle adolescence. Even if parents are very close with their children and
are generally aware of their whereabouts, condom negotiation is an intimate dyadic interaction
amongst sexual partners. Therefore, parents may not be able to impact this outcome.
Additionally, past work has found that a youth’s gender may make him or her more
vulnerable to negative peer influences in relation to sexual risk taking. Among these youth,
however, no differences were found in susceptibility to peer or partner influence. In contrast to
the findings of Henry et al., (2007), gender did not moderate the relationship between peer risk
and sexual risk taking in this sample. That is, the impact of peers’ sexual risk on youth’s
individual sexual risk taking was no different for boys than it was for girls. Henry and
colleagues observed that girls were more highly impacted by sexually risky peers than were boys
(Henry, et al., 2007). However, Henry et al. (2007) examined the effects of actual peer condom
use and ―peers’ perceptions of the costs of sex‖ (defined as the degree to which the adolescent
perceived that sex might have negative consequences such as becoming pregnant or losing
respect from your partner or family members) on the condom use behaviors of adolescents and it
was the ―peers’ perceptions of cost of sex‖ that impacted whether girls used condoms.

91

Therefore, it is possible that this gender moderated effect is of importance only when examining
the impact of peer attitudes (not behaviors) on adolescents. Perhaps peers ―teach‖ each other,
through their behaviors and communications, that they see no cost in sex and therefore sex
should not warrant a condom.
Additionally, it is important to note that despite recent declines in pregnancy rates, teenage
pregnancy remains relatively high in African American youth, especially in low-income urban
youth (Hamilton, Martin, & Ventura, 2010). In 2009, the birth rate for non-Hispanic Blacks was
59.0 per 1,000 (2.3 times higher than that of White teens) (Hamilton, et al., 2010). In
communities with high pregnancy rates, adolescent pregnancy may be seen as normative.
Therefore, preventing pregnancy through the use of condoms might not be as salient of a priority
to some African American youth living in low-income urban environments.
Finally, one might expect for boys to be more likely to utilize condoms with riskier sex
partners given the gender dynamics which can play out in condom negotiation. However, this did
not emerge as a significant relationship. This implies that gender may not be as salient an
influence over condom use as other factors in this sample. Perhaps it is the case that having a
greater degree of empowerment to enact condom use (as may be the case for boys) is not
sufficient for that behavior to take place. Instead, other variables (such as the adolescents’
attitudes toward condoms or motivation to use condoms) may be more important for condom use
to occur. It is important that potential influences to sexual risk taking and protection are further
explored. Understanding what factors make adolescents more likely to use condoms is a crucial
first step toward effective intervention strategies.

92

Theoretical Implications
In summary, this research sought to test a model of microsystemic level influences on the
condom use behaviors in a sample of African American adolescents. As described above, the
proposed model was elaborated on the basis of the extant literature and the Ecodevelopmental
theory and tested whether involved parenting was a significant buffer against high risk sex for
adolescents with sexually risky peer groups and partners. As can be derived from the results
outlined in the previous section, some aspects of the model were confirmed by the study analyses
while others were not. For instance, as suggested by Ecodevelopmental theory, the proposed
model postulated that influences pertaining to distinct microsystems would have an effect on
condom use in this population. While parenting did not have the hypothesized protective effect,
the study did find that both peers and partners appeared to influence condom use in this
population, thereby providing support for the importance of examining multiple systems in
trying to understand sexual risk in adolescents.
However, other aspects of the proposed theoretical model were not supported by the
empirical evidence. For instance, Ecodevelopmental theory and some of the extant literature
suggest that family is the most proximal and important microsystem in relation to adolescent
development, and therefore, the most significant microsystem in relation to risk and protective
behaviors. This was not corroborated by the study’s findings. Instead, partners emerged as the
most important influences followed by peers. It is possible that this parental influence over
adolescent sexual behaviors diminishes toward middle adolescence when a majority of youth is
already sexually active and outcomes of interest shift from sexual initiation to condom use.
Results suggest that theoretical models that have been utilized in the past to predict sexual

93

initiation may not be adequate to reliably explain and predict condom use and other
contraceptive behaviors.
It is also important to note that there are some key assumptions about familial characteristics
that underlie Ecodevelopmental theory that may or may not be applicable to this sample. That is,
the framework presumes that parental forces will always have a protective influence over
adolescents’ behaviors. However, not all parents display ―safe‖ sexual behaviors themselves.
Therefore, regardless of how close one is to one’s parental figure, if he or she is not modeling
protective sexual behaviors, it is unlikely that the expected protective effect will be found. As
mentioned above, it is likely that many of the youth in this sample could have had parents who
were quite young themselves and, therefore, not sexually safe or particularly skilled at
socializing their children in a way that might be protective against high risk sex. This could, in
turn, explain why the relationship between involved parenting and sexual risk was not found.
Additionally, it is important to note that low income African American families in urban
environments often face contextual challenges that may impact future orientation and therefore
diminish families’ abilities to be protective in relation to certain outcomes. (For example, some
research has found a significant relationship between increased racial discrimination and
decreased future orientation (Herrera, 2009).) The desire to use condoms in order to avoid
pregnancy and STDs presumes a future orientation. It is plausible, that some of the youth in the
sample might have had a decreased sense of hope for the future and, therefore, have been less
likely to be concerned about the potential negative consequences of unsafe sexual behaviors such
as pregnancy and STDs. Additionally, even for youth with a stronger future orientation, this
might not be the most salient force within the context of a sexual encounter. Oftentimes,
adolescents cite that ―it just happened‖ as the reason why they had unprotected sex. It is

94

important that theoretical models have the flexibility to account for these potential influences
over sexual behavior.
The hypothesized theoretical model also proposed that microsystemic influences would
work in concert, with involved parenting, buffering the effects of risky peers and partners.
However, this was not supported by the study results. Perhaps, it is not always the case that
these influences interact with one another. It is possible that, on some occasions and for certain
types of sexual risk outcomes, they actually work in a sequential manner. That is, parents may
start out as the most influential force in adolescents’ sexual risk and protective trajectory prior to
sexual initiation, however, after some period of early sexual experience, parents may become
less important (as protective forces) as partners and peers become the primary microsystemic
level factors that drive sexual risk and protection.
Overall, it would seem that Ecodevelopmental theory’s focus on understanding multiple
systems, how those systems interact with one another, and how the risk and protective factors
(and interactions amongst them) imbedded in each of those systems play out over an adolescent’s
development is indeed a useful way to organize risk and protective influences for condom use
behaviors in youth. However, this research suggests that the familial environment (in relation to
monitoring and support) may not be as central as suggested by this framework for this sample of
youth. It is necessary to expand current theories and create new theoretical frameworks that can
explain microsystemic influences on sexual behavior in middle and late adolescents.
As social scientists, it is important that we continue to map specific aspects of social
environments onto specific sexual risk outcomes. Eventually, this work can result in a more
complete picture of risk and protective influences in distinct populations that can account for
ecodevelopment (development within the individual and across ecosystems that surround her or

95

him) over time. This effort, in turn, will lead to more effective preventive intervention efforts for
different populations.
Implications
Overall, the apparent link between peer sexual risk and sexual risk taking observed in this
study suggests that prevention efforts that intervene at the level of the peer group may be
effective for African American adolescents in this age range. Teenagers between the ages of 14
and 19 are in a developmental stage when they are expanding their social networks and forging
closer relationships with their peers. This time period lends itself to greater influence from one’s
friends (some of whom are likely also to be sexual partners). Preventive work with these
populations should aim to set safe sex norms within adolescents’ social networks early on in
their development, before they have had the opportunity to develop risky habitual behaviors.
Programs should focus on the dissemination of healthy peer norms around sex. Additionally,
peer-to-peer and opinion-leader interventions should capitalize on the impact of friends’
behaviors on youth. Increasing healthy peer models for contraceptive practices, within the
context of adolescents’ social networks can be a promising way to harness peer influence for
health instead of risk.
Additionally, the importance of sexually risky partners on sexual risk taking speaks to the
potential benefit of a focus on sexual relationship dynamics in intervention work. That is,
preventive work could focus more on teaching adolescents how to negotiate for safer behaviors
with all partners. Although this can be a delicate prevention focus, given the complex trust
dynamics that can surround condom use in romantic relationships, it is crucial that young men
and women are able to negotiate safe sex practices effectively with their partners.

96

It is also important to note that the findings to this study seem to suggest that the manner in
which protective parenting is conceptualized is an important factor in whether or not research
shows that it is significantly related to sexual risk outcomes. For example, some studies suggest
that perhaps it is the ―support‖ aspect of involved parenting that is most important in influencing
condom use in youth who have initiated sex (Romer, et al., 1994). For instance, Romer et al.
(1994) found that while parental monitoring protected youth from initiating sexual intercourse at
an early age, parental closeness actually impacted whether or not they utilized condoms
consistently at a later time. This implies the existence of a threshold effect for the potential
protective effects of monitoring on sexual behaviors. It is also possible that a close parental
relationship may make the adolescent feel ―cared for‖ more than what is conveyed by close
monitoring, thereby promoting safer sex behaviors.
In this study, some of the individual items related to parental monitoring and parental support
showed significant relationships to the outcomes of interest while involved parenting as a whole
was not significantly related to sexual risk. This suggests that, for this particular sample, the
unique variance of those items was more important for sexual risk taking than their shared
variance. Specifically, it was those items that related to parents really knowing what adolescents
did with their free time and at night that were related to decreased condom use. As mentioned
above, simply trying to know what your adolescent does when unsupervised may not be
sufficient to protect him or her from risk. It is important that parents foster strong relationships
with their teenagers where they feel safe disclosing their whereabouts when they are not being
monitored.
Other literature suggests that different aspects of parenting, such as perceived parental
disapproval of teen sex and parenting (East, et al., 2006) or open parent-child communication

97

about safe sex (Crosby, DiClemente, et al., 2001a; Huebner & Howell, 2003), can be important
determinants of safe sexual behaviors in adolescents in this age group. McBride Murry’s (2007)
work suggests that protective factors such as parenting in relation to ethnic pride can have
important effects on African American adolescents’ intent to engage in sexual activity by
decreasing their susceptibility to negative peer influences. Finally, Belgrave and colleagues
(2000) found that higher levels of ethnic affiliation, belongingness and self esteem were related
to less risky sexual attitudes in urban African American girls (Belgrave, Van Oss Marin, &
Chambers, 2000). It is important that future research explores the potential protective influences
of these factors in African American youth.
Overall, it is crucial that the field develop clear and consistent definitions and
operationalizations of parenting concepts. This will help to improve our understanding of how
different risk and protective factors can impact distinct sexual behaviors (from sexual initiation
to STD status and pregnancy) throughout adolescents’ developmental paths. This increased
understanding could lead to improved intervention efforts that can be appropriately tailored to
adolescents’ developmental trajectories.
In addition to the ways in which parenting concepts are conceptualized, this study in
combination with the extant literature suggests that the ways in which protective factors, risk
factors, and outcomes of interest are defined and organized within contextual models plays an
important role in whether effects are found. For instance, ―risky sexual partners‖ was
conceptualized as a predictor of sexual risk taking in this model given that this study was
interested in trying to understand the ways in which partner characteristics (a function of partner
selection) could impact whether or not condom use took place. However, other studies have
examined partner constructs as outcome variables, in trying to understand the ways in which risk

98

and protective factors can lead to specific partner choices such as having older partners or more
partners (Miller, et al., 2000; Rodgers, 1999; Sturdivant, 2007). This brings up the possibility of
reciprocal relationships and the ways in which, as researchers, we tend to understand these
relationships as linear. However, they may in fact include feedback loops, whereby, for example,
having risky partners can lead to sexually risky behaviors, which in turn leads to selection of
further risky partners. It is important that future work examines the potential reciprocal effects
of risk factors, protective factors and outcomes of interest. It is also important that theoretical
frameworks are able to capture the dynamic interplay of risk and protective factors along
developmental trajectories.
Finally, it is important to note that few studies have utilized randomly attained communitybased samples of low-income, African American youth. Therefore, it is possible that the
findings of this study may be more representative of this population than more traditional
samples, drawn from schools for instance. In fact, nearly a quarter of the sample in this study
was not enrolled in school. Those youth would have gone unaccounted in other studies. Perhaps
youth who are not in school are riskier and less likely to be highly influenced by their parents.
This could explain why the expected protective effect of parenting was not found. Understanding
the ways in which these microsystemic influences impact (or do not impact) sexual risk taking in
more at risk populations, like adolescents who are not in school, is important in creating contextspecific intervention programming. Researchers and practitioners must be careful to not
generalize the results from ―safer‖ populations to more at risk populations. It is crucial that
future work continue to utilize innovative sampling designs like the Bayview Network Study.
Randomly attained, community based samples of this type can lead to effective intervention
work aimed at youth who are most at risk.

99

Limitations
The implications of this study should, of course, be understood within the context of its
limitations. Firstly, the limited sample size (especially in the peer risk model analyses) in this
study presented the possibility that some effects might have been present but simply too small to
be identified in this sample. Although all models converged and provided the expected
parameters of interest, it is possible that some relationships simply went undetected. It is
important that future studies using social network data, utilize sufficiently large samples that
allow for the exploration of microsystemic network influences.
Additionally, the small sample size and lack of variability in relation to STD status, made
it impossible to conduct accurate mediation analyses that could predict STD status. It is
important that future, large-scale, longitudinal studies also examine biological markers in order
to further explore these relationships over time and how they result in STD or HIV status.
Another limitation of this study is that only two time-points were examined. As
suggested by the literature summarized above, it seems that parenting, peers and partners may
have differing impacts at distinct stages of adolescent development. Having the opportunity to
track youth from very early on in their development through young adulthood might elucidate
these pathways and provide important implications for intervention.
Future Research
Overall, this research raises several important questions in relation to parent, peer, and
partner influences on condom use. Firstly, study findings, in combination with the extant
literature, seem to suggest that perhaps there is an age threshold at which parenting no longer
seems to have a significant effect on sexual risk outcomes in youth. Future work should explore
the ways in which distinct aspects of parenting are linked to sexual protection practices and
100

whether there is indeed a threshold for impact. A more nuanced developmental understanding of
parents as a microsystemic influence over sexual behaviors can lead to better-tailored
intervention efforts.
It is also important that future studies continue to examine the mechanisms through
which peers and partners can have negative sexual risk influences on youth. Knowing that these
microsystems are important is a first step toward effective prevention, but certainly not enough
to build effective interventions strategies. The joint use of qualitative and quantitative
methodologies may be an effective way to elucidate the ways in which these influences impact
sexual risk taking behaviors. Untangling the complicated constellation of micro and
mesosystemic influences over sexual risk behaviors in youth is a challenging endeavor. As
researchers, it is important that we stand back from our work and ask ourselves whether our
methodologies are truly able to capture those complexities. More open-format qualitative
methodologies may provide the richness necessary to uncover, not only the strength and
direction of certain relationships, but the actual mechanisms that drive them.
Future work should also examine adolescents’ complete social networks. The ability to
understand the ways in which peer risk can influence sexual risk taking would be greatly
enhanced with data collection strategies that could capture the complex nature of youths’ peer
groups. Specifically, learning what peer group characteristics and processes are associated with
greater degrees of sexual risk and protection can be an important step toward effective
prevention strategies.
Finally, the results of this study suggest that partner dynamics, especially with high risk
partners, can be very important in predicting condom use. It is important that future work
examine the ways in which these relationship processes play out between sexual partners and

101

how they result in either risk or protective behaviors. Achieving a better understanding of these
mechanisms can result in improved preventive programming for adolescents.
Conclusion
Overall, the findings from this study suggest the importance of microsystemic level
influences over condom use. It is crucial that future research continues to develop a deeper
understanding of the ways in which parents, peers and partners influence sexual risk and
protection at distinct phases of development.
HIV and STDs disproportionately affect African American adolescents. It is imperative
that research efforts continue to identify those factors that can contribute to risk and protective
practices in this population, especially for middle adolescents. This type of work is the first step
toward building effective intervention efforts that can result in decreased HIV and STD infection
in this population.

102

APPENDICES

103

APPENDIX A
Items in Scales

104

BAYVIEW NETWORK STUDY
ITEMS IN SCALES

A. PARENTAL MONITORING
These questions ask you about how much your parent(s) or guardian(s) know about how you
spend your time when you are not with them.
ASKED IF FEMALE OR MALE GUARDIAN IS IDENTIFIED
Monitoring
1. How much do your parent(s) or guardian(s) TRY to know about your whereabouts at night?
Would you say not at all, a little, or a lot?
Not at all
A little
A lot
Don't know
Refused
2. How much do your parent(s) or guardian(s) TRY to know about what you do with your free
time? Would you say not at all, a little, or a lot?
Not at all
A little
A lot
Don't know
Refused
NOTE: THE WORDS "after school" ARE ADDED IF IN SCHOOL.
3. How much do your parent(s) or guardian(s) TRY to know about where you are most typical
afternoons [after school]? Would you say not at all, a little, or a lot?
Not at all
A little
A lot
Don't know
Refused
4. How much do your parent(s) or guardian(s) REALLY know about where you go at night?
Would you say not at all, a little, or a lot?
Not at all
105

A little
A lot
Don't know
Refused
5. How much do your parent(s) or guardian(s) REALLY know about what you do with your
free time? Would you say not at all, a little, or a lot?
Not at all
A little
A lot
Don't know
Refused
6. How much do your parent(s) or guardian(s) REALLY know about where you are most
typical afternoons [after school]? Would you say not at all, a little, or a lot?
Not at all
A little
A lot
Don't know
Refused
B. PARENTAL/FAMILY SUPPORT
ASKED IF FEMALE GUARDIAN IDENTIFIED
These questions ask about your family and how close you feel to them. How close do you feel to
your [female guardian]? By close I mean, feeling like you can count on her to talk with you
about your personal problems and to provide you with emotional support.
1. Would you say you feel very close, somewhat close, not too close, or not close at all?
Very close
Somewhat close
Not too close
Not close at all
Refused
ASKED IF MALE GUARDIAN IDENTIFIED
2. How close do you feel to your [male guardian]. By close I mean, feeling like you can count
on him to talk with you about your personal problems and to provide you with emotional
support. Would you say you feel very close, somewhat close, not too close, or not close at
all?

106

Very close
Somewhat close
Not too close
Not close at all
Refused

NOT ASKED IF DOESN'T KNOW BIOLOGICAL MOM OR IF BIOLOGICAL MOM IS
DECEASED OR IF CURRENTLY LIVING WITH BIO MOM.
3. How close do you feel to your biological mother? By close I mean, feeling like you can
count on her to talk with you about your personal problems and to provide you with
emotional support. Would you say you feel very close, somewhat close, not too close, or not
close at all?
Very close
Somewhat close
Not too close
Not close at all
Refused
NOT ASKED IF DOESN'T KNOW BIOLOGICAL DAD OR IF BIOLOGICAL DAD IS
DECEASED OR IF CURRENTLY LIVING WITH BIO DAD.
4. How close do you feel to your biological father? By close I mean, feeling like you can count
on him to talk with you about your personal problems and to provide you with emotional
support. Would you say you feel very close, somewhat close, not too close, or not close at
all?
Very close
Somewhat close
Not too close
Not close at all
Refused
C. SEXUAL RISK/CONDOM USE AND SEXUALLY RISKY PEERS
1. How often have you and [partner 1] used condoms in the last 6 months
Every time
Most of the time
A few times
Never
2. Did you two use a condom the last time you had sexual intercourse?

107

Yes
No
D. SEXUALLY RISKY PARTNERS
1. Did [he/she] ever have a sexually transmitted disease?
Yes
No
2. Did [he/she] have any other sex partners in the last 6 months
Yes
No
D. PERCEPTION OF HIV/STD
1. How likely do you think it is that [he/she] (could/did) give you a sexually transmitted
disease? Would you say
Very likely
Somewhat likely
Somewhat unlikely
Very unlikely
2. How likely do you think it is that [he/she] (could/did) give you HIV? Would you say
Very likely
Somewhat likely
Somewhat unlikely
Very unlikely

108

APPENDIX B
IRB Documentation

109

IRB <IRB@ora.msu.edu>
To: morale33@msu.edu

Thu, Mar 18, 2010 at 2:29 PM

Dear Mercedes:
Please note, because you are only analyzing pre-existing deidentified data with no indirect
identifiers, and Dr. Jonathan Ellen of Johns Hopkins University has written a letter stating
that you will never be given any identifiers, you do not need to seek IRB approval. Good
luck with your research.
Thank you for erring on the side of caution and checking with us before beginning your
research study.
Good luck!
Becky Gore, 355-2181 (direct line)
SIRB Administrator
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Human Research Protection Programs
Biomedical and Health Institutional Review Board (BIRB)
Community Research Institutional Review Board (CRIRB)
Social Science/Behavioral/Education Institutional Review Board (SIRB)
Office of Regulatory Affairs
207 Olds Hall
Michigan State University
East Lansing, MI 48824-1046
Phone: (517) 355-2180
Fax: (517) 432-4503
Email: IRB@msu.edu

110

APPENDIX C
Correlations Tables

111

Table 13. Bivariate Correlations for all Study Variables
Age
Gender Lalone L1PorG L2PorG
Age

PM1

PM2

PM3

PM4

PM5

PM6

PS1

1

Gender

0.032

1

Lalone

.243**

.156*

1

L1PorG

-0.096

-0.058

-.336**

1

L2PorG

-0.022

-0.018

-.153*

-.879**

1

PM1

-0.131

0.014

.a

-0.088

0.088

1

PM2

-.160*

0.116

.a

0.08

-0.08

.512**

1

PM3

-.199**

0.09

.a

0.091

-0.091

.459**

.546**

1

PM4

-.234**

0.031

.a

-0.002

0.002

.312**

.226**

.309**

1

PM5

-0.072

-0.085

.a

-0.005

0.005

.317**

.351**

.183*

.531**

1

PM6

-0.135

0.001

.a

0.039

-0.039

.297**

.321**

.365**

.477**

.464**

1

PS1

-0.022

-.180*

-0.127

-0.014

0.076

0.128

0.113

0.13

.148*

.217**

.196**

1

PS2

0.037

-.253**

-0.118

-0.115

.178*

0.09

.166*

0.128

0.147

.195*

0.128

.281**

SR(T2)

.265**

0.133

.195**

-.155*

0.064

0.001

-0.043

-0.04

-.148*

-0.128

-0.029

-0.095

Lalone=Living alone, L1PorG=Living with one parent or guardian, L2PorG= Living with 2 parents or guardians,
112

Table 13. (cont’d)
PM=Parental monitoring (items 1-6), PS1=Parental support female parent or guardian, PS2=Parental support male parent or guardian,
SR(T2) = Sexual risk taking at time 2.
* Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001, .a= Unable to compute correlation because one of the
values is a constant (Adolescents who lived alone did not have parental monitoring scores.)

113

Table 13. (cont’d)
PS2

SR (T2)

PeerSR

RiskySP

PercRisk

PregAtt

MedInc

STD

SR (T1)

0.037

.265**

0.188

.210**

-.166*

-0.033

-0.083

0.02

-.383**

Gender

-.253**

0.133

0.119

0.074

-0.048

-0.187

-0.01

0.04

-0.097

Lalone

-0.118

.195**

0.111

.167*

-0.063

-.235*

-0.086

0.101

-0.132

L1PorG

-0.115

-.155*

-0.134

-0.001

-0.041

0.058

-.260**

-0.112

0.065

L2PorG

.178*

0.064

0.083

-0.085

0.075

0.09

.312**

0.063

-0.001

PM1

0.09

0.001

0.044

0.074

0.076

-0.115

-0.104

0.021

-0.002

PM2

.166*

-0.043

0.081

0.051

0.027

-0.041

-0.141

-0.01

-0.093

PM3

0.128

-0.04

0.127

-0.021

0.086

-0.079

-0.088

-0.054

0.063

PM4

0.147

-.148*

-0.019

-0.118

.148*

0.124

-0.007

-0.053

.165*

PM5

.195*

-0.128

-0.017

-0.029

.150*

0.131

-0.039

0.076

0.01

PM6

0.128

-0.029

0.115

-0.027

0.09

-0.039

-0.067

0.06

0.076

PS1

.281**

-0.095

-0.09

-0.086

0.107

0.054

-0.056

0.085

0.075

PS2

1

-0.063

0.016

-.196**

0.051

-0.006

0.027

0.103

0.002

Age

PS2=Parental support male parent or guardian, SR(T2) = Sexual risk taking at time 2, PeerSR=Peer sexual risk, PercRisk=Perception
of risk, PregAtt=Pregnancy attitudes, MedInc=Median census block income, STD=STD status, SR(T1)= Sexual risk taking at time 1,
114

Table 13. (cont’d)
Lalone=Living alone, L1PorG=Living with one parent or guardian, PM=Parental monitoring (items 1-6), PS1=Parental support female
parent or guardian. * Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001

115

Table 13 (cont’d)
PS2

SR (T2)

PeerSR

RiskySP

PercRisk

PregAtt

MedInc

STD

SR (T2)

-0.063

1

PeerSR

0.016

.226*

1

RiskySP

-.196**

.359**

0.021

1

PercRisk

0.051

-.212**

.206*

-.376**

1

PregAtt

-0.006

-0.119

-0.139

-0.132

0.006

1

MedInc

0.027

0.041

0.015

-.148*

0.121

0.181

1

STD

0.103

.192*

0.181

.164*

0.005

-0.1

0.017

1

SR (T1)

-.152*

.412**

0.038

.246**

-0.06

-0.203

0.026

0.105

SR (T1)

1

PS2=Parental support male parent or guardian, SR(T2)=Sexual risk taking at time 2, PeerSR=Peer sexual risk, PercRisk=Perception of
risk, PregAtt=Pregnancy attitudes, MedInc=Median census block income, STD=STD status, SR(T1)=Sexual risk at time 1.
* Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001

116

Table 13. (cont’d)
Age

Gender

Lalone

L1PorG L2PorG

PM1

PM2

PM3

PM4

PM5

PM6

PS1

PeerSR

0.188

0.119

0.111

-0.134

0.083

0.044

0.081

0.127

-0.019

-0.017

0.115

-0.09

RiskySP

.210**

0.074

.167*

-0.001

-0.085

0.074

0.051

-0.021

-0.118

-0.029

-0.027

-0.086

PercRisk

-.166*

-0.048

-0.063

-0.041

0.075

0.076

0.027

0.086

.148*

.150*

0.09

0.107

PregAtt

-0.033

-0.187

-.235*

0.058

0.09

-0.115

-0.041

-0.079

0.124

0.131

-0.039

0.054

MedInc

-0.083

-0.01

-0.086

-.260**

.312**

-0.104

-0.141

-0.088

-0.007

-0.039

-0.067

-0.056

0.02

0.04

0.101

-0.112

0.063

0.021

-0.01

-0.054

-0.053

0.076

0.06

0.085

.318**

.200**

.157*

-0.003

-0.076

-.144*

-.163*

-.165*

-.246**

-.195**

-.195**

-.164*

STD
SR (T1)

Lalone= Living alone, L1PorG=Living with one parent or guardian, PM=Parental monitoring (items 1-6), PS1=Parental support
female parent or guardian, PeerSR=Peer sexual risk, PercRisk=Perception of risk, PregAtt=Pregnancy attitudes, MedInc=Median
census block income, STD=STD status, SR(T1)= Sexual risk at time 1.
* Significant at p<0.05, ** Significant at p<0.01, *** Significant at p<0.001

117

REFERENCES

118

REFERENCES
Allison, P. D. (2008). Convergence Failures in Logistic Regression. SAS Global Forum:
Statistics and Data Analysis, 360.
Auerswald, C. L., Muth, S. Q., Brown, B., Padian, N., & Ellen, J. (2006). Does partner selection
contribute to sex differences in sexually transmitted infection rates among African
American adolescents in San Francisco? Sex Transm Dis, 33(8), 480-484.
Baltes, P. B., Reese, H. W., & Nesselroade, J. R. (1977). Life-span developpmental psychology:
Introduction to research methods. Monterrey, CA: Brooks/Cole.
Bandura, A. (1977). Social Learning Theory. Englewood Cliffs, NJ: Prentice Hall.
Baptiste, D. R., Tolou-Shams, M., Miller, S. R., McBride, C. K., & Paikoff, R. L. (2007).
Determinants of parental monitoring and preadolescent sexual risk situations among
African American families living in urban public housing. Journal of Child and Family
Studies, 16(2), 261-274.
Barnes, G. M., Reifman, A. S., Farrell, M. P., & Dintcheff, B. A. (2000). The effects of parenting
on the development of adolescent alcohol misuse: A six-wave latent growth model.
Journal of Marriage & the Family, 62(1), 175-186.
Bearman, P., & Bruckner, S. (1999). Power in numbers: Peer effects on adolescent girls' sexual
debut and pregnancy.
Belgrave, F. Z., Van Oss Marin, B., & Chambers, D. B. (2000). Cultural, contextual, and
intrapersonal predictors of risky sexual attitudes among urban African American
adolescent girls in early adolescence. Cultural Diversity and Ethnic Minority Psychology,
6(3), 309-322.
Borawski, E. A., Ievers-Landis, C. E., Lovegreen, L. D., & Trapl, E. S. (2003). Parental
monitoring, negotiated unsupervised time, and parental trust: The role of perceived
parenting practices in adolescent health risk behaviors. Journal of Adolescent Health,
33(2), 60-70.
Bronfenbrenner, U. (1979). Ecology of Human Development: Experiments by Nature and
Design. Cambridge, MA: Harvard University Press.
Browning, C. R., Leventhal, T., & Brooks-Gunn, J. (2005). Sexual initiation in early
adolescence: The nexus of parental and community control. American Sociological
Review, 70(5), 758-778.
Cates, J. R., Herndon, N. L., Schulz, S. L., & Darroch, J. E. (2004). Our Voices, Our Lives, Our
Futures: Youth and Sexually Transmitted Diseases. Chapel Hill, NC: School of
Journalism and Mass Communication, University of North Carolina at Chapel Hill.
119

Centers for Disease Control and Prevention. (2006). Youth Risk Behavior Surveillance-United
States, 2005. Morbidity and Mortality Weekly Report, 55(SS-5), 1-108.
Centers for Disease Control and Prevention. (2007, March). HIV/AIDS Surveillance Report.
Centers for Disease Control and Prevention. (2008a). HIV in the United States, CDC HIV/AIDS
Facts. Retrieved July 17, 1008, from
http://www.cdc.gov/hiv/resources/factsheets/PDF/us.pdf
Centers for Disease Control and Prevention. (2008b). HIV/AIDS Surveillance in Adolescents
and Youth Adults. Slide Sets. Retrieved July 14, 2008, from
http://www.cdc.gov/hiv/topics/surveillance/resources/slides/adolescents/index.htm
Centers for Disease Control and Prevention. (2008, March). Sexually Transmitted Disease
Surveillance, 2006.
Centers for Disease Control and Prevention. (2009). Sexual and Reproductive Health of Persons
Aged 10–24 Years — United States, 2002–2007. Morbidity and Mortality Weekly Report,
58(SS-6).
Collins, W. A. (2003). More than myth: The developmental significance of romantic
relationships during adolescence. Journal of Research on Adoelscence, 13(1), 1-24.
Crosby, R. A., DiClemente, R. J., Wingood, G. M., Cobb, B. K., Harrington, K., Davies, S. L., et
al. (2001). HIV/STD-protective benefits of living with mothers in perceived supportive
families: A study of high-risk African American female teens. Preventive Medicine: An
International Journal Devoted to Practice and Theory, 33(3), 175-178.
Crosby, R. A., DiClemente, R. J., Wingood, G. M., Sionean, C., Cobb, B. K., Harrington, K., et
al. (2001a). Correlates of using dual methods for sexually transmitted diseases and
pregnancy prevention among high-risk African-American female teens. J Adolesc Health,
28(5), 410-414.
Crosby, R. A., DiClemente, R. J., Wingood, G. M., Sionean, C., Cobb, B. K., Harrington, K. F.,
et al. (2001b). Correlates of casual sex among African-American female teens. Journal of
HIV/AIDS Prevention & Education for Adolescents & Children, 4(4), 55-67.
DiClemente, R. J., Salazar, L. F., & Crosby, R. A. (2007). A review of STD/HIV preventive
interventions for adolescents: Sustaining effects using an ecological approach. [Review].
Journal of Pediatric Psychology, 32(8), 888-906.
DiClemente, R. J., Wingood, G. M., Crosby, R., Sionean, C., Cobb, B. K., Harrington, K., et al.
(2001). Parental monitoring: association with adolescents' risk behaviors. Pediatrics,
107(6), 1363-1368.

120

DiClemente, R. J., Wingood, G. M., Crosby, R. A., Sionean, C., Cobb, B. K., & Harrington, K.
(2002). Sexual risk behaviors associated with having older parnters: A study of Black
adolescent females. Sexually Transmitted Diseases, 29, 20-24.
DiLorio, C., Dudley, W. N., Soet, J. E., & McCarty, F. (2004). Sexual possibility situations and
sexual behaviors among young adolescents: the moderating role of protective factors. J
Adolesc Health, 35(6), 528.e511-520.
Donenberg, G. R., Wilson, H. W., Emerson, E., & Bryant, F. B. (2002). Holding the line with a
watchful eye: The impact of perceived parental monitoring on risky sexual behavior
among adolescents in psychiatric care. AIDS Education and Prevention, 14(2), 138-157.
East, Khoo, S. K., & Reyes, B. T. (2006). Risk and Protective Factors Predictive of Adolescent
Pregnancy: A Longitudinal, Prospective Study. Applied Developmental Science, 10(4),
188-199.
East, & Reyes, B. T. (2007). Association between adolescent pregnancy and a family history of
teenage births. Perspectives on Sexual and Reproductive Health, 39(2), 108-115.

Ellen, J. M., Brown, B. A., Chung, S. E., Potterat, J. J., Muth, S. Q., Valente, T. W., et al. (2005).
Impact of sexual networks on risk for gonorrhea and chlamydia among low-income urban
African American adolescents. J Pediatr, 146(4), 518-522.
Enders, C. K., & Bandalos. (2001). The relative performance of full information maximum
likelihood estimation for missing data in structural equation models Structural Equation
Modeling: A Multidisciplinary Journal, 8(3), 430-457.
Grossman, C., Hadley, W., Brown, L. K., Christopher, D. H., Peters, A., Tolou-Shams, M., et al.
(2008). Adolescent sexual risk: Factors predicting condom use across the stages of
change. AIDS and Behavior, 12, 913-922.
Hamilton, B. E., Martin, J. A., & Ventura, S. J. (2010). Preliminary data for 2009. National Vital
Statistics Reports, 59(3).
Henggeler, S. W., & Bourdin, C. M. (1990). Family therapy and beyond: A multisystemic
approach to treating the behavior problems of children and adolescents. Pacific Grove,
CA: Brooks/Cole.
Henrich, C. C., Brookmeyer, K. A., Shrier, L. A., & Shahar, G. (2006). Supportive Relationships
and Sexual Risk Behavior in Adolescence: An Ecological--Transactional Approach.
Journal of Pediatric Psychology, 31(3), 286-297.
Henry, D. B., Schoeny, M. E., Deptula, D. P., & Slavick, J. T. (2007). Peer Selection and
Socialization Effects on Adolescent Intercourse Without a Condom and Attitudes About
the Costs of Sex. Child Development, 78(3), 825-838.
121

Herrera, D. E. (2009). Perceived racial/ethnic discrimination, hope, and social connectedness:
Examining the predictors of future orientation in emerging adults. University of Texas,
Austin.
Hobfoll, S. E. (1998). Ecology, community, and AIDS prevention. American Journal of
Community Psychology, 26(1), 133.
Huebner, A. J., & Howell, L. W. (2003). Examining the relationship between adolescent sexual
risk-taking and perceptions of monitoring, communication, and parenting styles.
[Empirical]. Journal of Adolescent Health, 33(2), 71-78.
Kirby, D., & Lepore, G. (2007). Factors Affecting Teen Sexual Behavior, Pregnancy,
Childbearing and Sexually Transmitted Disease: Which are Important? Which can you
Change? : ETR Associates.
Kline, R. B. (1998). Principles and Practice of Structural Equation Modeling. New York: The
Guilford Press.
Li, X., Feigelman, S., & Stanton, B. (2000). Perceived parental monitoring and health risk
behaviors among urban low-income African-American children and adolescents. Journal
of Adolescent Health, 27(1), 43-48.
Li, X., Stanton, B., & Feigelman, S. (2000). Impact of perceived parental monitoring on
adolescent risk behavior over 4 years. Journal of Adolescent Health, 27(1), 49-56.
Longmore, M. A., Manning, W. D., & Giordano, P. C. (2001). Preadolescent parenting strategies
and teens' dating and sexual initiation: A longitudinal analysis. Journal of Marriage &
the Family, 63(2), 322-335.
Luke, D. A. (2005). Getting the big picture in community science: Methods that capture context
American Journal of Community Psychology, 35(3/4), 185-200.
Lyons, H., Giordano, P., Manning, W., & Longmore, M. (2009, April 30 - May 2). Identity, Peer
Context, and Adolescent Girls' Sexual Behavior. Paper presented at the Annual Meeting
of the Population Association of American, Detroit, Michigan.
Manlove, J., Terry-Humen, E., & Ikramullah, E. (2006). Young teenagers and older sexual
partneres: Correlates and consequences for males nad females. Perspectives on Sexual
and Reproductive Health, 38, 197-207.
Manning, W. D., Flanigan, C. M., Giordano, P. C., & Longmore, M. A. (2009). Relationship
dynamics and condom use among adolescents. Perspectives on Sexual and Reproductive
Health, 41(3), 181-190.

122

McBride Murry, V., Berkel, C., Brody, G., Gibbons, M., & Gibbons, F. X. (2007). The Strong
African American Families Program: Longitidinal pathways to sexual risk reduction.
Journal of Adolescent Health, 41(4), 333-342.
Miller, K. S., Forehand, R., & Kotchick, B. a. (2000). Adolescent sexual behavior in two ethnic
minority groups: A multisystem perspective. Adolescence, 35(138), 313-333.
Millhausen, R. R. (2003). Rural and nonrural African American high school students and
STD/HIV sexual behaviors. American Journal of Health Behavior, 27(4), 373-379.
Minuchin, S. (1974). Families and family therapy. Cambridge, MA: Harvard University Press.
Muthen. (2002). Using MPlus Monte Carlo simulations in practice: A note on assessing
estimation quality and power in latent variable models. MPlus Web Notes, 1(2).
Muthen. (2010). MPlus Support Discussion Forum: Standardized Coefficients and Fit Indices.
Retrieved March 15, 2011, from http://www.statmodel.com/cgibin/discus/discus.cgi?pg=prev&topic=11&page=4464
New Strategist Publications. (2007). American Sexual Behavior: Demographics of Sexual
Activity, Fertility, and Childbearing.
O'Donnell, L., Myint, U. A., O'Donnell, C. R., & Stueve, A. (2003). Long-term influence of
sexual norms and attitudes on timing of sexual initiation among urban minority youth. J
Sch Health, 73(2), 68-75.
O'Donnell, L., O'Donnell, C. R., & Stueve, A. (2001). Early sexual initiation and subsequent sexrelated risks among urban minority youth: the reach for health study. Fam Plann
Perspect, 33(6), 268-275.
Paikoff, R. L. (1995). Early heterosexual debut: situations of sexual possibility during the
transition to adolescence. Amer J Orlhopsychial, 63(3), 389-401.
Paikoff, R. L. (1996). Adapting developmental research to intervention design: Applying
developmental psychology to an AIDS prevention model for urban African American
youth. [Empirical, Theoretical]. Journal of Negro Education. Special Focus: Coordinated
Children Services, 65(1), 44-59.
Perrino, T., Gonzalez-Soldevilla, A., Pantin, H., & Szapocznik, J. (2000). The role of families in
adolescent HIV prevention: A review. Clinical Child and Family Psychology Review,
3(2), 81-96.
Rai, A. A., Stanton, B., Wu, Y., Li, X., Galbraith, J., Cottrell, L., et al. (2003). Relative
influences of perceived parental monitoring and perceived peer involvement on
adolescent risk behaviors: An analysis of six cross-sectional data sets. Journal of
Adolescent Health, 33(2), 108-118.
123

Regnerus, M. D., & Luchies, L. B. (2006). The Parent-Child Relationship and Opportunities for
Adolescents' First Sex. Journal of Family Issues, 27(2), 159-183.
Reitman, D. (1996). Predictors of African American adolescents' condom use and HIV risk
behavior. AIDS Education and Prevention, 8(6), 499-515.
Resnick, M. D., Bearman, P. S., Blum, R. W., Bauman, K. E., Harris, K. M., Jones, J., et al.
(1997). Protecting adolescents from harm. Findings from the National Longitudinal Study
on Adolescent Health. JAMA, 278(10), 823-832.
Rink, E., Tricker, R., & Harvey, S. M. (2007). Onset of sexual intercourse among female
adolescents: The influence of perceptions, depression, and ecological factors. Journal of
Adolescent Health, 41(4), 398-406.
Roche, K. M., Ahmed, S., & Blum, R. W. (2008). Enduring consequences of parenting for risk
behaviors from adolescence into early adulthood. Social Science & Medicine, vol, 66(9),
2023-2034.
Rodgers, K. B. (1999). Parenting processes related to sexual risk-taking behaviors of adolescent
males and females. Journal of Marriage & the Family, 61(1), 99-109.
Romer, D., Black, M., Ricardo, I., Feigelman, S., & et al. (1994). Social influences on the sexual
behavior of youth at risk for HIV exposure. American Journal of Public Health, 84(6),
977-985.
Rose, A., Koo, H. P., Bhaskar, B., Anderson, K., White, G., & Jenkins, R. R. (2005). The
influence of primary caregivers on the sexual behavior of early adolescents. Journal of
Adolescent Health, 37(2), 135-144.
Santelli, J. S., Robin, L., Brener, N. D., & Lowry, R. (2001). Timing of alcohol and other drug
use and sexual risk behaviors among unmarried adolescents and young adults. Family
Planning Perspectives, 33(5), 200-205.
Satorra, A., & Benteler, P. (1999). A scaled difference chi-square test statistic for moment
structure analysis. University of California Los Angeles: Statistics Series, 260, 1-13.
Scaramella, L. V., Conger, R. D., Simons, R. L., & Whitbeck, L. B. (1998). Predicting risk for
pregnancy by late adolescence: A social contextual perspective. Developmental
Psychology, 34(6), 1233-1245.
Stanton, B., Li, X., Pack, R., Cottrell, L., Harris, C., & Burns, J. M. (2002). Longitudinal
influence of perceptions of peer and parental factors on African American adolescent risk
involvement. J Urban Health, 79(4), 536-548.

124

Staras, S. A., Cook, R. L., & Clark, D. B. (2009). Sexual partner characteristics and sexually
transmitted diseases among adolescents and young adults. [Journal Article; Research
Support, N.I.H., Extramural]. Sex Transm Dis, 36(4), 232-238.
Sturdivant, A. D. (2007). Peer pressure and risky sexual behavior in African-American
dolescents: Moderating effects of parental communication about sex, parental
monitoring, and best friend's sexual behavior. Dissertation Abstracts International:
Section B: The Sciences and Engineering Vol, 67(9-B), 5425.
Szapocznik, J., & Coatsworth, D. (1999). An ecodevelopmental framework for organizing the
influences on drug abuse: A dvelopmental model of risk and protection. In M. D. Glatz
& C. R. Hartel (Eds.), Drug Abuse: Origins and Interventions (pp. 331-366).
Washington, DC: American Psychological Association.
Tabachnick, B. G., & Fidell, L. S. (Eds.). (2007). Using Multivariate Statistics (5th ed.): Pearson,
Boston.
Teitelman, A. M., Ratcliffe, S. J., Morales-Aleman, M. M., & Sullivan, C. (2008). Sexual
relationship power, intimate partner violence, and condom use among minority urban
girls, Journal of Interpersonal Violence, 23, 1694-1711.
Tolman, D. L., Spencer, R., Rosen-Reynoso, M., & Porche, M. V. (2003). Sowing the seeds of
violence in heterosexual relationships: Early adolescents narrate compulsory
heterosexuality. Journal of Social Issues, 59, 159-178.
United States Census Bureau. (2000). San Francisco County, California by Block Group. Median
Household Income in 1999: 2000, from
http://factfinder.census.gov/servlet/ThematicMapFramesetServlet?_bm=y&_MapEvent=zoom&-context=tm&-errMsg=&-_useSS=N&-_dBy=150&redoLog=false&-_zoomLevel=5&-tm_name=DEC_2000_SF3_U_M00024&tm_config=|b=50|l=en|t=403|zf=0.0|ms=thm_def|dw=0.127654448550011|dh=0.0757647
5132991932|dt=gov.census.aff.domain.map.EnglishMapExtent|if=gif|cx=122.4943948387799|cy=37.76111587275979|zl=4|pz=4|bo=|bl=|ft=350:349:335:389:388:
332:331|fl=403:381:204:380:369:379:368|g=05000US06075|ds=DEC_2000_SF3_U|sb=
50|tud=false|db=150|mn=0|mx=200001|cc=1|cm=1|cn=5|cb=|um=Dollars|pr=0|th=DEC_2
000_SF3_U_M00024|sf=N|sg=&-PANEL_ID=tm_result&-_pageY=&-_lang=en&geo_id=05000US06075&-_pageX=&-_mapY=&-_mapX=&-_lati#?255,14
Voisin, D. R., DiClemente, R. J., Salazar, L. F., Crosby, R. A., & Yarber, W. L. (2006).
Ecological Factors Associated with STD Risk Behaviors among Detained Female
Adolescents. [Empirical]. Social Work, 51(1), 71-79.
Wingood, G. M., & DiClemente, R. J. (2000). Application of the theory of gender and power to
examine HIV-related exposures, risk factors, and effective interventions for women.
Health Education & Behavior, 27(5), 539-565.

125