Three Essays on the Cultural Context of Adolescent Romantic Relationships and Sexual Behavior DISSERTATION Presented in Partial

Home , Abstinence pledge, Sexual partner

Three Essays on the Cultural Context of Adolescent Romantic Relationships and Sexual Behavior

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Brian Soller, M.A.

Graduate Program in Sociology

The Ohio State University

2013

Dissertation Committee:

Christopher R. Browning, Co-Advisor

Dana L. Haynie, Co-Advisor

Hui Zheng

Copyrighted by

Brian Soller

2013

Abstract

The transition into adolescence brings increased involvement in romantic and sexual relationships for most youth. However, sociologists have only recently begun to examine the developmental consequences of early romantic involvement. And while much research has focused on adolescent sexual intercourse, sex is most often narrowly conceptualized as a form of risk-taking in these studies. While informative, conceptualizing adolescent sexual behavior purely in terms of risk limits the understanding of the wider impact of adolescents’ sexual activity on their development and personal well-being. Integrating theoretical insights from perspectives on culture, gender, social networks, and identity, this dissertation examines adolescent romantic relationships and sexual behavior using data from the first two waves of the National

Longitudinal Study of Adolescent Health (Add Health). I focus on both the dynamics of early romances as well as the mental health consequences of romantic involvement and sexual intercourse among adolescents.

First, I consider whether school-based sexual double standards—differing standards of sexual permissiveness among boys and girls—alter the association between sexual intercourse, gender, and adolescent mental health. I measure one particular aspect of the sexual double standard by quantifying within-school differences in boys’ and girls’ perceptions of the social benefits of sexual intercourse. I find that girls who had sexual intercourse are more likely to report severe depression as the sexual double standard in a ii

school increases. Conversely, boys who engaged in sexual intercourse with one or more non-romantic partners are more likely to report high self-esteem as the sexual double standard increases.

Second, I integrate insights from cultural sociology and differential association/social learning theories to explain how cultural and structural features of friendship groups influence adolescent romantic relationship inauthenticity—the extent of incongruence between one’s thoughts/feelings and actions within romantic contexts. I use sequence analysis and linear regression to test whether adolescents experience greater romantic relationship inauthenticity when the ordering of events within their ideal romantic relationship scripts (e.g., holding hands, saying “I love you,” having sexual intercourse) diverges from sequencing of events within their friends’ ideal romantic relationship scripts. I also test whether this association varies according to adolescents’ level of interaction with friends and the overall popularity of their friendship groups.

Results indicate romantic relationship inauthenticity increases as one’s ideal script diverges from the scripts of one’s friends. I also find that being attached to popular friends accentuates this association.

Finally, integrating insights from cultural sociology and identity theory, I explore the mental health consequences of adolescent romantic relationship inauthenticity by measuring its association with numerous mental health outcomes (e.g., severe depression, suicide ideation, etc.). I find high levels of romantic relationship inauthenticity increase the risk of poor mental health, but only among girls.

iii

This dissertation provides novel insights regarding: (1) the importance of culture in determining how sexual activity affects subsequent well-being; (2) how cultural reinforcement and social network processes shape the link between culture and action; and (3) the role gender and culture play in determining how early romantic involvement influences psychological well-being.

Acknowledgments

I first wish to thank the Department of Sociology at Ohio State University for supporting over the years. I also acknowledge the staff members and faculty affiliates of the Criminal Justice Research Center and the Institute for Population Research. The scholarly and collegial environments and financial backing of these centers have greatly enhanced my experience at Ohio State.

I have had a number of collaborators at OSU, including Trevon Logan, Kate

Calder, Leigh Fine, Nate Doogan, and Chris Keenan. I have learned much from you all and I am glad to call you all friends and colleagues. I also wish to thank Hui Zheng for serving on my dissertation committee and Ruth Peterson for serving on my general exam committee. I think of Ruth as the prototypical scholar and I hope to emulate her throughout my career.

I have had the great pleasure of working closely with Chris Browning and Dana

Haynie. They have made an indelible impact on my own approach to sociological inquiry and I will be forever grateful for their guidance. I look forward to our future collaborations and continuing friendships.

I also wish to thank my friends in Columbus and those back in the Bay Area. I have always been able to count on them for support and to make me laugh. My parents,

Al and Kathy Soller, my sister and brother-in-law, Kim and Brent Tanimoto, and my

grandparents, Nancy and Tony Potts, have been some of my most vocal supporters throughout my graduate career and I cannot thank them enough.

Finally, I wish to thank Aubrey Jackson. Her presence has made working countless hours (on a graduate student stipend) something I can easily bear. She is a great scholar, friend, and partner and I hope one day to repay her kindness.

Dedicated to Tony Potts

vii

Vita

June 2001 ...... Bishop O’Dowd High School

August 2005 ...... B.A. California State University, East Bay

July 2008 ...... M.A. Sociology, California State University,

East Bay

2005 to 2010 ...... Research Associate, Prevention Research

Center, Berkeley, CA

2008 to 2009 ...... Graduate Fellow, The Ohio State University

2009 to present ...... Graduate Research Associate, Department

of Sociology, The Ohio State University

Publications

In Press Browning, Christopher R., Brian Soller, Margo Gardner, and Jeanne Brooks- Gunn. “‘Feeling Disorder’ as a Comparative and Contingent Process: Gender, Neighborhood Conditions, and Adolescent Mental Health” Journal of Health and Social Behavior.

In Press Haynie, Dana L., Brian Soller, and Kristi Williams “Anticipating early fatality: The role of individual, peer, and school level fatality on adolescent risky behaviors.” Journal of Youth and Adolescence.

viii

In Press Haynie, Dana L. and Brian Soller. “Social network analysis and the measurement of peer effects.” In Bruinsma and Weisburd (Eds), The Encyclopedia of Criminology and Criminal Justice.

In Press Soller, Brian, and Christopher R. Browning. “Neighborhood effects and social networks.” In Bruinsma and Weisburd (Eds), The Encyclopedia of Criminology and Criminal Justice.

2012 Cagney, Kathleen A., Christopher R. Browning, Aubrey L. Jackson and Brian Soller. “Social network, neighborhood, and institutional effects in aging research: An integrated ‘activity space’ approach to examining social context.” Pp. 60-80 in Perspectives on the Future of the Sociology of Aging. Linda J. Waite, Editor. Washington, DC: The National Academies Press.

2011 Lee, Juliet P., Robynn Battle, Brian Soller, and Naomi Brandes. “Thizzin’— Ecstasy use contexts and emergent social meanings” Addiction Research and Theory 19:528-541.

2010 Lee, Juliet P., Robynn Battle, Rob Lipton, and Brian Soller. “Smoking: Use of cigarettes, cigars, and blunts among Southeast Asian youth and young adults” Health Education Research 25:83-96.

2010 Soller, Brian and Juliet P. Lee. “Drug intake methods and social identity: The use of marijuana in blunts among Southeast Asian adolescents and emerging adults” Journal of Adolescent Research 25:783-806.

Fields of Study

Major Field: Sociology

Table of Contents

Abstract ...... ii

Acknowledgments...... v

Vita ...... viii

List of Tables ...... xii

List of Figures ...... xxiv

Chapter 1: Introduction ...... 1

Chapter 2: The Sexual Double Standard, Sexual Intercourse, and Adolescent Mental

Health ...... 15

Chapter 3: “I Did it My Way”: The Peer Context of Adolescent Romantic Relationship

Inauthenticity ...... 56

Chapter 4: Caught in a Bad Romance: Adolescent Romantic Relationships and Mental

Health ...... 93

Chapter 5: Conclusion...... 121

References ...... 133

Appendix A: Tables from Chapter 2 ...... 144

Appendix B: Figures from Chapter 2...... 150 x

Appendix C: Tables from Chapter 3 ...... 153

Appendix D: Figures from Chapter 3 ...... 161

Appendix E: Tables from Chapter 4 ...... 164

Appendix F: Figures from Chapter 4 ...... 173

List of Tables

Table A.1. Descriptive Statistics ...... 145

Table A.2. Multilevel Logistic Regression Models of Severe Depression Regressed on

School Sexual Double Standards and Sexual Behavior Relationship Context ...... 146

Table A.3. Multilevel Logistic Regression Models of High Self-Esteem Regressed on

School Sexual Double Standards and Sexual Behavior Relationship Context...... 147

Table A.4. Coefficients for Control Variables Omitted from Table A.2...... 148

Table A.5. Coefficients for Control Variables Omitted from Table A.3...... 149

Table C.1. Substitution Cost Matrix for Ideal Relationship Script Items...... 154

Table C.2. Descriptive Information on Additional Variables from Table C.4...... 155

Table C.3. Descriptive Statistics...... 156

Table C.4. Probit Regression of the Hazard of New Romantic Relationship Between

Waves 1 and 2...... 157

Table C.5. Linear Regression Models of Romantic Relationship Inauthenticity Regressed on Script Discordance...... 158

xii

Table C.6. Linear Regression Models of Romantic Relationship Inauthenticity Regressed on Script Discordance, Friend Involvement, and Friends’ Popularity...... 159

Table C.7. Sensitivity Analyses...... 160

Table E.1. Descriptions of Control Variables...... 165

Table E.2. Probit Regression of the Hazard of New Romantic Relationship Between

Waves 1 and 2...... 167

Table E.3. Descriptive Statistics by Gender...... 168

Table E.4. Logistic Regressions of Girls’ Severe Depression and High Somatic

Symptoms...... 169

Table E.5. Logistic Regressions of Girls’ Suicide Ideation and Suicide Attempt...... 170

Table E.6. Logistic Regressions of Boys’ Severe Depression and High Somatic

Symptoms...... 171

Table E.7. Logistic Regressions of Boys’ Suicide Ideation and Suicide Attempt...... 172

xiii

List of Figures

Figure B.1. Predicted Probability of Severe Depression Among Girls by the Sexual

Double Standard and Sexual Relationship Categories ...... 151

Figure B.2. Predicted Probability of High Self-Esteem Among Boys by the Sexual

Double Standard and Sexual Relationship Categories ...... 152

Figure D.1. Illustration of High Script Discordance and Low Script Heterogeneity...... 162

Figure D.2. Predicted Values of Relationship Inauthenticity by Script Discordance and

High and Low Friend Popularity...... 163

Figure F.1. Predicted Probability of Girls’ Poor Mental Health by Romantic Relationship

Inauthenticity Quartiles...... 174

xiv

Chapter 1: Introduction

Increased involvement in romantic and sexual relationships is a hallmark of adolescence (Giordano, Manning, and Longmore 2006; Tolman and McClelland 2011).

Despite this fact, social scientists have only recently taken to romantic relationship dynamics as an imperative subject of empirical study. As a result, how early romances factor into adolescent development is not well understood. And although much research has focused on adolescent sexual behavior, most studies on the topic are narrow in theoretical and empirical scope. Until the end of the 20th century, social scientists have for the most part equated adolescent sexuality with risk-taking and danger (Moran 2000).

Accordingly, research on adolescent sexual behavior has tended to approach the subject from a risk-based theoretical framework, focusing on the predictors of outcomes such as teenage pregnancy, sexually transmitted infection, condom non-use, and sexual violence

(Tolman and McClelland 2011). While these outcomes are important to consider, conceptualizing early sexual behavior purely in terms of risk limits the understanding of sexual development, as well as the wider impact of adolescents’ sexual activity on their development and personal well-being.

In this dissertation I advance the understanding of adolescent sexual behavior and romantic relationships by focusing in internal dynamics of early romances and the effects of romantic involvement and sexual behavior on mental health. I develop a more 1

comprehensive approach to understanding adolescent sexual behavior and romantic relationships by integrating complementary theories centered on gender, identity, culture, and social network processes. Incorporating key aspects of these perspectives brings novel insights into the understanding of how cultural processes operating at multiple levels (i.e., individual, peer groups, and schools) shape the inner workings of early romances condition the impact of romantic and sexual relationships on the mental well- being of youth in the United States. In this this introductory chapter I detail how my dissertation synthesizes insights from multiple theoretical perspectives in order to provide novel insights into adolescent romantic relationships and sexual development.

The Significance of Adolescent Sexual Behavior and Romantic Involvement

Adolescence is typically a time during which youth first form serious romantic relationships and engage in sexual activity (Connolly et al. 2004; Giordano 2003).

Despite increasing involvement in romantic relationships throughout adolescence, little sociological research focuses on the dynamics of early romances (Brown, Feiring, and

Furman 1999), especially when compared to the large number of studies on adolescent friendships and sexual relationships (Giordano, Longmore, and Manning 2006; Giordano,

Manning, et al. 2006). In addition, little research centers on the developmental consequences of early romances.

However a limited number of studies on adolescent romantic relationships attest to their developmental significance. Romantic involvement is linked to depressive symptoms among youth and under certain conditions may contribute to delinquent

behavior (Joyner and Udry 2000; Kreager and Haynie 2011; McCarthy and Casey 2008).

In addition, psychological and physical aggression also occurs within significant portions of adolescent romantic relationships (Halpern et al. 2001). However, supportive youth romantic relationships may promote adolescent well-being (e.g., Simons and Barr forthcoming) and usher more successful union formations during adulthood (Collins

2003).

Adolescent romantic relationships are also developmentally significant because sexual activity most often occurs between romantic partners (Manning, Longmore, and

Giordano 2005). Although sexual activity becomes increasingly normative throughout adolescence, sexual intercourse is linked to a number of adverse outcomes among youth.

Apart from increasing the risk of unwanted pregnancies and sexually-transmitted infections, sexual intercourse is linked to poor mental health among youth (Hallfors et al.

2005). The effects of sexual intercourse on mental health however, appear to depend upon a myriad of individual and contextual factors (McCarthy and Casey 2008; Meier

2007; Spriggs and Halpern 2008). One major objective of this dissertation is to illustrate how one particular cultural feature of adolescent school contexts—the sexual double standard—alters the association between sexual intercourse and subsequent mental health. In so doing, I further clarify how the developmental significance of sexual intercourse among adolescents is contingent upon the cultural context in which it occurs.

The Sexual Double Standard, Identity, and Mental Health

The effect of cultural features of adolescents’ interactional contexts (e.g., peer groups, schools, neighborhoods, etc.) on their behavior is a withstanding concern among sociologists. For instance, early cultural scholars assert subcultural norms and values typically concentrated in disadvantaged and socially-disorganized neighborhoods promote delinquency and other risky behavior among youth (Cloward and Ohlin 1960;

Shaw and McKay 1942; Sutherland 1947). More recently, cultural theories have been extended to explain adolescent sexual behavior and romantic relationship outcomes

(Harding 2007, 2010; Warner et al. 2011). Scholars also provide convincing arguments concerning how context-based cultural norms regarding adolescent sexual behavior affect the link between early sexual activity and subsequent mental health (Meier 2007).

Some have suggested the presence of sexual double standards—i.e., differing standards of sexual permissiveness across gender groups—help explain gender variation in sexual activity (Crawford and Popp 2003). For instance, greater acceptance of sexual intercourse among males may lead to increased involvement in sexual intercourse among boys when compared to girls. Also implied in this argument is that girls embedded with contexts bereft of sexual double standards may experience adverse social and personal consequences (e.g., poor mental health, social stigma) upon engaging in sexual intercourse. However, no research to date has focused on how sexual double standards factor into the link between sexual involvement and mental health among adolescents.

In Chapter 2, I synthesize perspectives on gender, identity, and culture and examine whether the sexual double standard alters the association between sexual intercourse and boys’ and girls’ mental health. Drawing from cultural sociology (Harding

2010; Kirk and Papachristos 2011) and symbolic interactionism (Giordano et al. 2009;

Matsueda 1992; Mead 1934), I propose school-based sexual double standards function as cultural “frames” (Lamont and Small 2008) pertaining to gender differences in the social consequences of sexual activity. As a cultural frame, sexual double standards help adolescents conceptualize sexual behavior and identify how it fits into various social roles within school contexts.

Among girls, sexual double standards contribute to a perception that sexual activity involves taking stigmatized or devalued roles from within the larger school context (e.g., sluts). Upon engaging in sexual activity, girls may experience higher likelihoods of severe depression and lower likelihoods of high self-esteem as the sexual double standard in a school becomes more salient. On the other hand, sexual double standards may promote a perception that sexual activity is integral to esteemed male roles

(e.g., “player”). If so, sexually-active boys may be less likely to experience severe depression and have higher self-esteem upon appraising their actions through the sexual double standard frame.

Chapter 2 addresses an emerging concern regarding healthy sexuality development among adolescents (Tolman and McClelland 2011). Rather than conceptualizing early sexual activity as necessarily entailing undue risk or being accompanied by adverse outcomes, researchers and policy scholars are increasingly recognizing that romantic and sexual involvement are normative throughout adolescence.

Acknowledging this fact has led to a burgeoning theoretical and empirical approach to understanding adolescent health and sexual development. To that end, researchers are

offering more attention to important contingencies in the association between sexual and romantic involvement and adolescent developmental outcomes. In addition, researchers are offering more attention to the role of social contexts in shaping adolescent romantic and sexual relationship formation and dynamics. In Chapter 2, I address how school- based sexual double standards alter the association between sexual intercourse and subsequent mental health. In so doing, I contribute to the understanding of the social context of adolescent romantic relationships and sexual development.

Friendship Networks and Romantic Relationship Inauthenticity

Chapter 3 focuses on the peer context of romantic relationship inauthenticity, which refers to the extent of incongruence between one’s ideas/thoughts and one’s actions within romantic settings. According to developmental psychologists, relationship inauthenticity represents a key mechanism through which early romances influence adolescent development and well-being (Impett et al. 2008). Adolescents often forgo their emotional needs to avoid conflict and maintain intimate relationships, thereby increasing the sollrisk of experiencing relationship inauthenticity (Impett et al. 2008).

Identifying the conditions that lead to inauthentic romances is important as compromised relationship authenticity is associated with low self-esteem and depression among adolescents (Impett et al. 2008; Theran 2011; Tolman et al. 2006) and diminished sexual self-efficacy among girls (Impett, Schooler, and Tolman 2006).

Most research on adolescent relationship inauthenticity comes from developmental psychology. As a result, few have explored the social factors that

contribute to inauthentic romantic relationships. Integrating insights from cultural sociology, social network perspectives, and social learning theory, I focus on romantic relationship “scripts”—cultural templates for ordering behavior within romantic contexts—to measure adolescents’ idealized relationships and conceptualize how friends influence relationship inauthenticity. The extent of behavioral divergence from one’s ideal romantic relationship script captures an important aspect of relationship inauthenticity, namely the degree to which actions and events within relationships unfold in a manner that is inconsistent with how they would ideally occur.

Chapter 3 focuses on the association between “script discordance,” among one’s peers and relationship inauthenticity. Script discordance refers to the extent to which the sequencing of events within close friends’ ideal relationship scripts diverges from one’s ideal romantic relationship script. I suggest one’s ideal script is more strongly reinforced within one’s peer group when there is greater agreement between one’s personal ideal script and the ideal scripts of close friends. Conversely, one’s ideal script is less strongly reinforced when it diverges from the scripts of one’s close friends. Adolescents may thus be increasingly likely to enact inauthentic romances when their ideal script lacks reinforcement within their peer groups.

I also explore whether important network structural characteristics of friendship groups alter the association between script discordance and romantic relationship inauthenticity. Drawing insights from social network perspectives (Haynie 2001) and differential association/social learning theory (Akers 2009; Sutherland 1947), I suggest important features of friendship networks alter the association between script discordance

and relationship inauthenticity. High levels of friend involvement provide opportunities for individuals to articulate and reinforce their own ideal ordering of romantic relationships to friendship group members. Thus, the extent of interaction with friends may accentuate the association between script discordance and relationship inauthenticity by providing opportunities for group members to reinforce certain scripts.

I also examine whether attachment to peers who maintain high status within school-based social hierarchies alters the association between script discordance and relationship inauthenticity. Differential association theory suggests intense relations are consequential for peer influence processes because they determine the probability that cultural features of peer groups will be reinforced on individual members. Akers

(2009:65) suggests intensity refers to the “significance, salience, or importance of the association to the individual.” I argue friends with high social status within larger peer settings represent particularly intense associations because these friends potentially yield more social rewards and help bolster one’s own peer status (Dijkstra et al. 2010).

Accordingly, scripts within a friendship group are more likely to be reinforced within high status peer groups than within low status peer groups. This leads to the expectation that the positive association between script discordance and relationship inauthenticity will be stronger among those embedded in high status peer groups.

Results from Chapter 3 suggest peer contexts play an important role in shaping adolescents’ experiences with romantic relationship inauthenticity. The study in Chapter

3 also contributes to the sociology of culture by highlighting how network structure

relates to cultural processes and by demonstrating the importance of script discordance in shaping the association between culture and action at the individual-level.

The Mental Health Consequences of Relationship Inauthenticity

Chapter 4 focuses on the mental health consequences of romantic relationship inauthenticity. Here I integrate insights from gender perspectives and identity theory to explore the relationship between gender, relationship inauthenticity, and multiple markers of mental health, including depression, somatic symptoms, suicide ideation, and suicide attempt. In this chapter I suggest romantic relationship inauthenticity contributes to poor mental health by disrupting the process of self-verification. Self-verification occurs when meanings within a situation match or confirm meanings of a particular identity (Cast and

Burke 2002). Ideal relationship scripts are meaningful components of identity that reflect idealized romantic selves. Enacting meaningful ideal scripts within relationships verifies components of the self and enhances mental health. Conversely, experiencing relationships that diverge from one’s ideal romantic relationship likely contributes to poor mental health through failure in the self-verification process.

I also examine whether gender alters the association between relationship inauthenticity and mental health. While some suggest romantic involvement equally enhances boys’ and girls’ well-being (La Greca and Harrison 2005), past research indicates romantic dynamics have stronger effects on girls’ mental health (Joyner and

Udry 2000). Gender variation in the association between romantic relationship outcomes and mental health may be attributable to the increased salience of interpersonal

relationships in girls’ self-concepts (Rosenfield and Mouzon 2013). Accordingly, relationship inauthenticity may have a particularly strong association with romantically- involved girls’ mental health.

The study presented in Chapter 4 demonstrates the importance of gender and inauthenticity in shaping the effect of romantic involvement on adolescent mental health.

I also contribute to theories of culture and action by demonstrating the importance of gender in determining how relationship inauthenticity affects adolescent mental health.

This study suggests a holistic approach to adolescent romantic relationships—one attuned to culture and relationship progressions—provides novel insights into how romantic involvement influences adolescent mental health.

DATA AND METHODS

All data used in this dissertation come from the first two waves National

Longitudinal Study of Adolescent Health (Add Health). Add Health began as a nationally representative longitudinal school-based study that explores the etiology of health behaviors and outcomes among youth in the United States. A total of 4 waves of data collection have gathered information on respondents’ health behaviors and outcomes from adolescence and into young adulthood. Importantly, the first two waves of the Add

Health data include a number of variables that can be used to operationalize key network structural and cultural features of adolescent school and peer contexts. As such, the Add

Health data provide an unrivaled data source for testing the theoretical perspective presented in this dissertation.

The Add Health study is based on a cluster-based sample for which high schools served as the primary sampling unit. All high schools in the United States that included an 11th grade and had 30 or more enrollees were eligible for the study. The Add Health research compiled a random sample of 80 high schools that was stratified by region, urbanicity, school type, ethnic makeup, and size. The largest feeder school for each sampled high school was also recruited when available, which resulted in a sample of more than 130 schools.

All Add Health respondents were originally nested in high schools and feeder schools during the 1994-1995 school year. More than 90,000 adolescents completed an in-school survey that took place between 1994 and 1995. Roughly 20,000 adolescents completed the first in-home interview, which gathered detailed information on respondents’ sexual behavior, romantic relationships, and mental health. Nearly 15,000 respondents completed the wave 2 interview, which took place approximately 1 year after the first in-home interview. Although the Add Health research team has collected data from respondents well into adulthood, I restrict my analysis to the first two waves of Add

Health in this dissertation because of my focus on adolescent romantic relationships.

Analytic Samples

The analytic samples used in this dissertation vary across study analyses.

Chapter 2. Chapter 2 focuses on how school-level sexual double standards

(measured at wave 1) shape the impact of sexual intercourse (measured at wave 2) on subsequent mental health among adolescents (also measured at wave 2). I restrict my

analysis to respondents over the age of 15 because key questions used to measure school sexual double standards were only asked of respondents aged 15 and older. I drop 63 middle schools (and 4,239 respondents) from the analysis because most respondents from feeder schools were younger than age 15 at the first in-home interview. Accordingly, I am unable to reliably measure the sexual double standard within these schools. I exclude

7 schools (and 30 respondents) with insufficient data for multilevel modeling of both boys’ and girls’ mental health across waves 1 and 2. Finally, I exclude respondents with missing survey weights and respondents with missing data on the dependent variables.

My final sample for the study presented in Chapter 2 includes 8,513 respondents who were nested in 75 schools.

Chapter 3. The study presented in Chapter 3—“I Did it My Way”: The Peer

Context of Adolescent Romantic Relationship Inauthenticity —focuses on the peer context of adolescent romantic relationship inauthenticity. In this study I test whether script discordance within friendship groups is associated with adolescent romantic relationship inauthenticity and whether key network structural characteristics alter the association between script discordance and romantic relationship inauthenticity. Testing these associations requires complete data on both school-based social networks as well as information on friends’ ideal romantic relationship scripts. Complete information on both friendship ties and ideal romantic relationship scripts are only available for a subset of respondents from a “saturated” sample of respondents.

As part of the original design of Add Health, the research team attempted to interview every respondent from 16 schools (including 2 large and 14 small) during the

first two in-home interviews. As population samples were collected from these schools, this “saturated” sample contains information on a wide variety of attributes for every friend—including romantic relationship scripts. (Conversely, information on every friend’s ideal relationship script was not collected for most respondents from the non- saturated sample). Accordingly, the analytic sample used in Chapter 3 is restricted to this saturated sample. I exclude one of the schools from the saturated schools because its low response rate precludes the measurement of school network characteristics. I also exclude respondents who were lost to follow up or were not interviewed at wave 2. Finally, I exclude respondents who did not form a new romantic relationship between interview waves or have missing data on their ideal relationship scripts or their first subsequent romantic relationship. Finally, I exclude respondents with missing survey weights. The analytic sample from Chapter 3 includes 1,013 respondents nested in 15 schools.

Chapter 4. The study presented in Chapter 4—Caught in a Bad Romance:

Adolescent Romantic Relationships and Mental Health—focuses on the influence of romantic relationship inauthenticity on adolescents’ mental health. In this study I test whether the discrepancy between respondents’ ideal relationship (measured at wave 1) and their actual relationships (measured at wave 2) is associated with multiple measures of mental health (assessed at wave 2). For this study I am able to draw from the entire

Add Health sample, however I restrict my analytic sample as follows. Of the nearly

15,000 respondents who participated in both waves 1 and 2, my sample includes 6,173 who formed a new romantic relationship between waves. Among those respondents I exclude 550 individuals who did not provide information on ideal romantic scripts or

actual romantic relationships. In addition, I exclude 12 respondents with missing data on dependent variables and 295 respondents who were missing survey weights. The final analytic sample for Chapter 2 includes 5,316 respondents (2,905 girls and 2,411 boys).

Chapter 2: The Sexual Double Standard, Sexual Intercourse, and Adolescent Mental Health

Gender differences in the personal consequences of sexual activity are drawing increasing attention from journalists and “pop cultural feminists” (Armstrong, Hamilton, and England 2010). For example, instances of “slut shaming” among adolescents (i.e., publicly degrading girls on the basis of their actual or perceived sexual behavior) suggest sexual activity continues to increase girls’ risk of peer-based stigma (Snow and Hagan

2009). At the same time, boys are often rewarded for their sexual escapades by their peers. For instance, several male athletes in one Northern California high school participated in a “Fantasy Slut League” in which participants “drafted” female students and earned points for engaging in sexual activity with draftees (Huet 2012). Although the frequency of slut shaming and prevalence of slut leagues across the U.S. remains unknown, accounts of gender variation in the social consequences of sexual activity are strongly suggestive of a continued sexual double standard within certain adolescent contexts (i.e., differing standards of sexual permissiveness across gender groups). These accounts also suggest schools and adolescent peer contexts fundamentally shape the personal consequences of sexual activity.

To date, few sociologists have examined the effects of sexual behavior on adolescents’ psychological well-being. A few notable studies have found that sexual

activity is associated with emotional and psychological distress, however the relationship depends on a number of individual factors (e.g., gender, age of sexual onset, relationship context Armour and Haynie 2007; McCarthy and Casey 2008; Meier 2007). Others have found inconsistent associations between sexual activity and subsequent self-esteem

(Goodson, Buhi, and Dunsmore 2006; Lyons et al. 2010). These studies demonstrate the personal consequences of sexual behavior are complex and depend on a myriad of individual and social factors.

Even less research has examined how social contexts (e.g., family, neighborhoods, schools, etc.) condition the association between sexual activity and subsequent mental health. For instance, it is suggested sexual double standards determine how sexual behavior affects social and psychological well-being (Lyons et al. 2010).

However the mechanisms through which sexual double standards alter the association between gender, sexual behavior, and mental health are under-theorized and lack empirical support. These limitations are in part due to the fact that no study has directly measured the severity of sexual double standards across several school contexts and assessed how it bears on the association between gender, sexual behavior, and psychological well-being.

In this study I extend recent research on the association between sexual activity and adolescent psychological well-being by considering whether this association depends upon gender and the salience of sexual double standards within schools. Building upon insights from cultural sociology (Harding 2010; Kirk and Papachristos 2011) and symbolic interactionism (Giordano et al. 2009; Matsueda 1992; Mead 1934), I propose

the sexual double standard in part functions as a cultural “frame” (Lamont and Small

2008) that pertains to gender differences in the social consequences of sexual activity.

This frame in turn helps adolescents encode sexual behavior with meaning and identify how it fits into various social roles within school contexts.

Individuals draw upon shared conceptions of sexual behavior to understand how certain behavior relates to one’s identity (Giordano et al. 2009). Among girls, sexual double standards may lead to the perception that partaking in sexual activity (particularly sexual intercourse) entails taking roles that are stigmatized or devalued within the larger context (e.g., sluts). Compared to girls who abstain from sexual intercourse, sexually- active girls may be increasingly more likely to experience severe depression and less likely to have high self-esteem as the sexual double standard in a school becomes more salient. Conversely, sexual double standards foster the perception that sexual activity is an integral component of esteemed male roles (e.g., “player”). As a result, boys who engage in sexual intercourse may experience less depression and have higher self-esteem upon appraising their sexual behavior through the sexual double standard frame.

I use data from the National Longitudinal Study of Adolescent Health (hereafter

Add Health) to test these hypotheses. As Add Health includes a nationally-representative sample for which schools served as the primary sampling unit, it provides a unique opportunity to both measure the sexual double standard across several school contexts and assess how it factors into the association between gender, sexual behavior, and mental health. I first construct a measure of the sexual double standard across 75 high schools in the U.S. that captures the extent to which boys in a school perceive there to be

more social rewards (e.g., respect among peers) associated with sexual intercourse than do girls. I then test whether this school-based sexual double standard frame alters the association between gender, sexual behavior, and subsequent severe depression and high self-esteem. Apart from identifying social conditions under which sexual behavior adversely affects adolescent psychological well-being, this study helps understand how sexual double standards hinder young people’s healthy sexual development (Tolman and

McClelland 2011).

BACKGROUND AND SIGNIFICANCE

Sexual Behavior and Adolescent Mental Health

A number of studies indicate the prevalence of depression and other mood disorders (e.g., anxiety, low self-esteem) is higher among adolescents than children

(Costello, Erkanli, and Angold 2006). Substantial social and physiological changes experienced throughout adolescence appear to account for increases in depression and lowered self-esteem throughout this life-course period. According to “storm and stress” perspectives (Arnett 1999; Hall 1904), adolescence exposes youth to stressors that place them at increased risk of mental and behavioral health problems. Relational stressors rooted in schools and peer groups are thought to be particularly important in contributing to higher rates of depression among adolescents.

For most, adolescence brings increasing romantic involvement (Connolly et al.

2004; Giordano 2003). Romantic relationships entail unique stressors that may negatively impact psychological health. For instance, romantic involvement may negatively alter

existing relationships with parents and friends (Joyner and Udry 2000). Physical and psychological aggression also occurs within significant proportions of adolescent romantic relationships (Halpern et al. 2001). Rejection and relationship dissolution are also often emotionally-taxing experiences (Larson and Sweeten 2012). Perhaps it is no surprise romantic involvement is positively associated with depression and low self- esteem among some adolescents (Joyner and Udry 2000).

Romantic involvement may also increase depressive symptoms because adolescent sexual activity most often occurs between romantic partners (Manning et al.

2005). Sexual activity in turn has been linked to depression past research (Hallfors et al.

2005). However, the association between sexual activity and adolescent depression depends on a number of individual and social factors. For instance, Meier (2007) found that age-graded norms regarding the timing of sexual intercourse in large part determine the nature of the association between sexual activity and mental health. Violating age- graded sexual norms may cause distress because early entrants may not be developmentally or socially mature enough to handle the emotional and physical responsibilities this life course transition poses (Meier 2007). At the same time, the relationship context in which sexual activity occurs (e.g., romantic versus non-romantic), the status of the partnership (e.g., ongoing versus dissolved), and the extent of emotional commitment between partners also appear to alter the association between sexual behavior and subsequent mental health (Meier 2007). Other studies further suggest the association between sexual behavior and subsequent mental and behavioral health depend

upon these and other factors (Armour and Haynie 2007; McCarthy and Casey 2008;

Spriggs and Halpern 2008).

While individual characteristics alter the association between sexual activity and mental health, few have considered (apart from age-graded norms) how characteristics of larger social contexts in which sexual activity occurs factor into the association between sexual behavior and adolescent psychological well-being. As a result, how culture affects the association between sexual behavior and subsequent mental health remains unknown.

A few qualitative studies however highlight the potential for cultural climates of adolescent environments (e.g., schools, peer groups, neighborhoods) to determine how sexual behavior affects adolescent depression and self-esteem (Anderson 1989; Eder,

Evans, and Parker 1995; Harding 2010; Lyons et al. 2010; Pascoe 2007). In the following sections, I focus on particularly salient cultural components of adolescent school contexts—sexual double standards—and specify the mechanisms through which they potentially alter the association between sexual activity and mental health.

Conceptualizing and Measuring the Sexual Double Standard

Most research on sexual double standards focuses on within-person variation in evaluations of men and women on the basis of sexual behavior. For instance, surveys have been used to assess differences in personal assessments of hypothetical female and male “targets” who engage in similar sexual behavior (Reiss 1964). Other studies have used vignettes in which respondents evaluate male and female targets on the basis of the targets’ sexual histories (Crawford and Popp 2003). The fact that female targets are more-

often-than-not judged more harshly than their male counterparts for similar sexual behavior across study designs suggests sexual double standards persist (Crawford and

Popp 2003).

While sexual double standards most likely remain, research on sexual double standards is limited in a number of ways. First, most studies rely on assessments of others to measure sexual double standards among respondents within a single setting such as a college campus. An important yet overlooked aspect of sexual double standards pertains to how individuals conceptualize the personal consequences of sexual activity. For instance, sexual double standards may be conceptualized as differences in perceived social benefits of sexual behavior among boys and girls within a particular context (e.g., schools, neighborhoods, etc.) (Hamilton and Armstrong 2009). In schools where strong sexual double standards exist, sexual activity is perceived to result in more social rewards

(e.g., increased peer status) among boys than girls.

Quantifying within-school differences in boys’ and girls’ perceptions of the consequences of sexual activity allows for a more thorough examination of the mechanisms through which sexual double standards alter the association between sexual behavior and adolescent mental health. For instance, respondents in Hamilton and

Armstrong’s (2009) qualitative study on women’s experiences with “hooking up” (i.e., sexual activity that occurs outside of traditional dating contexts) in college campuses highlights a double standard surrounding the personal consequences of casual sex. As one respondent noted, “Guys can have sex with all the girls and it makes them more of a man, but if a girl does then all of a sudden she’s a ho, and she’s not as quality of a person”

(Hamilton and Armstrong 2009:589 emphasis added). Hamilton and Armstrong suggest a sexual double standard surrounding the informal labeling of men and women who engage in hook-ups generated a fear of sexual stigma that severely limited their respondents’ sexual agency (2009:598) and contributed to feelings of personal shame after engaging in hookups (2009:606). Attending to individuals’ conceptions of the gender-specific consequences of sexual behavior across contexts may help understand the mechanisms through which sexual double standards shape the association between sexual activity and mental health.

Another limitation of past research on sexual double standards is that no study has taken a contextual approach to examine sexual double standards and measured their severity across several settings. Accordingly, little is known as to whether the sexual double standards vary in their severity across contexts and whether such variation alters the association between sexual activity and psychological well-being. It may be the case that sexual activity has negligible effects on girls’ and women’s psychological well-being among those embedded in contexts with low sexual double standards. Conversely, compared to girls who abstain from sex, sexually-active girls may be increasingly more likely to experience poor mental health when they attend schools with strong sexual double standards.

I address limitations of past research on sexual double standards by testing whether one particular type of sexual double standard alters the association between sexual behavior and boys’ and girls’ mental health. I suggest sexual double standards represent cultural frames that encode sexual behavior with meaning and helps actors

identify how sexual behavior fits various roles within school contexts. In this study I measure one particular sexual double standard at the school-level by quantifying the extent to which boys and girls in the same school differ in the perception that sexual intercourse promotes social esteem among their peers (e.g., high peer status, appearing sexually/romantically desirable). Drawing upon insights from cultural sociology (Harding

2010; Kirk and Papachristos 2011) and symbolic interactionism (Giordano et al. 2009), I specify the mechanisms through which one sexual double standard alters the association between sexual behavior, gender, and mental health.

Cultural Frames and Self-Appraisal Processes

An emerging sociological theory suggests culture informs behavior in large part through the provision of “frames” (Lamont and Small 2008). Snow and Benford

(1992:137) define a frame as “an interpretive [schema] that simplifies and condenses the

‘world out there’ by selectively punctuating and encoding objects, situations, events, experiences, and sequences of action within one’s present or past environment.” Frames help individuals arrive at shared understandings of the personal and social consequences of behavior (Kirk and Papachristos 2011). Most relevant to the present study, actors rely upon frames pertaining to gender and sexual behavior to conceptualize how sexual behavior fits into certain roles. Frames facilitate the classification of one’s self and others on the basis of personal traits (e.g., gender) and sexual behavior.1

1See Harding’s (2010) discussion of frames and the categorization of girls on the basis of girls’ actual or perceived behavior (including sexual activity). 23

Frames are implicated throughout the “role-taking” process, which is a mostly unconscious practice in which individuals project themselves into the roles of others, and appraise from others’ standpoints, one’s self in the situation and possible lines of action

(Matsueda 1992). According to Soller et al. (2012), cultural frames inform appraisals by helping actors encode certain behavior with meaning. Upon appraising actions through frames infused with social understandings of behavior, individuals identify how actions constitute certain social roles. This process allows actors to classify activities and expectations that distinguish certain roles from others. Individuals thus rely upon frames throughout the “role-taking” process in order to encode behavior with meaning and associate certain with certain roles according to their behavior and personal traits (e.g., gender).

As a cultural frame, a sexual double standard pertaining to the social benefits of sexual activity may lead actors to codify sexual permissiveness as integral to esteemed and degraded gendered roles. Adolescents may be increasingly likely to associate female sexual permissiveness as being closely tied to “sluts,” (Eder et al. 1995), “stunts”

(Harding 2010), or other stigmatized female gender roles as the sexual double standard in a school becomes more salient. Conversely, boys may be more likely to conceptualize male sexual permissiveness as being intimately tied to esteemed male roles, such as

“pimps,” (Staiger 2005) or “players” (Giordano et al. 2009) as the sexual double standard frame in a school becomes more salient. Importantly, within-school gender differences in the perceived consequences of sexual behavior help ensure that sexual behavior is fundamental to adolescents’ roles within certain school contexts.

I test whether one sexual double standard frame alters the association between gender, sexual behavior, and mental health. A sexual double standard pertaining to gender differences in the social benefits of sexual intercourse may modify the association between sexual behavior and subsequent depression/self-esteem among boys and girls because it represents a frame that helps one identify how sexual behavior informs one’s own particular role(s). Upon evaluating their sexual activity through the sexual double standard frame, sexually-active girls may be increasingly likely to self-categorize themselves into stigmatized or low-status female roles (e.g., slut). Perceiving they express negative attributes of such roles, sexually-active girls may experience lower feelings of self-worth, thus contributing to higher odds of experiencing severe depression and lower odds of high self-esteem. This leads me to my first set of hypotheses:

H.1a. Sexual activity increases the risk of severe depression among girls as the sexual double standard in a school increases, and

H.1b. Sexual activity decreases the odds of having high self-esteem among girls as the sexual double standard in a school increases.

Among boys, sexual activity may lead to lower levels of depression and higher self-esteem within schools with strong sexual double standards. Through the same self- appraisal process, sexually-active boys may be increasingly likely to perceive that they occupy esteemed male roles (e.g., player) upon evaluating their sexual activity through a frame imbued with the sexual double standard. Perceiving they embody the most valued attributes of respected male roles, sexually-active boys may have lower risks of severe depression and higher odds of high self-esteem through amplified feelings of self-worth.

Anderson’s (1989) study of disadvantaged black male youth highlights potential mechanisms through which sexual double standards reinforce barometers of self-worth for boys. Anderson found that among young black males in disadvantage inner-city contexts, the male peer group “places a high value on sex, especially what many middle- class people call casual sex. Although the sex may be casual in terms of commitment to the partner, it is usually taken quite seriously as a measure of a boy’s worth. Thus a primary goal of the young man is to find as many willing females as possible. The more

‘pussy’ he gets, the more esteem accrues to him” (Anderson 1989:61 emphasis added).

The same increase in peer status is not afforded to girls and women upon engaging in frequent sexual exploits, as evidenced by the necessity for boys’ need to first develop

“game” (i.e., stylized presentations of self) in order to “get over” girls’ and women’s sexual defenses and engage in sexual activity to accrue social status and self-esteem.

Furthermore, Anderson suggests inner-city black girls maintain dreams of “having a boyfriend, a fiancé, and husband, and the fairy-tale prospect of living happily ever after in a nice house in a neighborhood with one’s children—essentially the dream of the middle-class American life-style” (Anderson 1989:62). The context in which Anderson conducted his field work appears to be replete with a sexual double standard pertaining to differences in the extent to which sexual activity can be used to promulgate esteemed social identities among boys and girls.2

2Importantly, Anderson hypothesized that his respondents establish alternative means for achieving positive masculine identities because conventional markers of manhood (e.g., heading an economically self-sufficient household) are perceived to be mostly unattainable for those within socially-isolated and disadvantaged contexts. However, Giordano and colleagues’ (2009) test of Anderson’s hypothesis suggests that the “player” 26

For girls, there are fewer social advantages for engaging in sexual activity as the sexual double standard increases. Conversely, for boys casual sexual activity is a fundamental component of positive social roles within contexts characterized by strong sexual double standards. Sexually-active boys may perceive they occupy esteemed social roles and may be increasingly likely to have positive self-evaluations upon appraising one’s self through a frame imbued with the sexual double standard. This leads me to my second set of hypotheses:

H.2a. Sexual activity decreases the risk of severe depression among boys as the sexual double standard in a school increases, and

H.2b. Sexual activity increases the odds of having high self-esteem among boys as the sexual double standard in a school increases.

DATA AND METHODS

Add Health is a longitudinal nationally-representative survey that focuses on health-related behaviors and outcomes among adolescents and young adults in the United

States. Study respondents were initially nested within randomly selected high schools and feeder schools. The Add Health research team compiled a school-based clustered random sample of 80 high schools that was stratified by school population size, ethnic composition, public/private status, geographic region, and urbanicity. All high schools that included at least 30 enrollees and an eleventh grade were eligible for participation.

The largest feeder school for each high school was recruited when available. Add Health

persona resonates with both black and white boys. These findings highlight the importance of sexual behavior in achieving esteemed male roles among different race/ethnicities and socioeconomic groups. 27

included more than 130 schools ranging in size from fewer than 100 students to more than 3,000.

The current study uses data from the first two waves of Add Health. All respondents included in my study completed initial in-home interviews in 1995 and a second interview that took place roughly a year after the first interview. The in-home interviews gathered detailed information concerning respondents’ sexual behavior and experiences with depression, as well as demographic characteristics and a number of other health-related outcomes.

Analytic Sample

In total, 14,736 respondents (from 145 schools) participated the first two in-home interviews. Of these respondents, I exclude 7 schools (and 30 respondents) with insufficient data for multilevel modeling of both male and female depression across the first two study waves. Questions measuring the perceived benefits of sexual activity were asked only of those aged 15 years or older. Because most respondents from feeder schools were younger age 15 at the first in-home interview, I drop 63 middle schools (and

4,239 respondents) from the analysis as the sexual double standard cannot be reliably measured within these schools. I also exclude an additional 1,513 respondents who were younger than 15 at the time of the first interview. Finally, I exclude 431 respondents with missing survey weights and 10 respondents with missing data on the dependent variables.

My final sample includes 8,513 respondents who were nested in 75 schools.

Dependent Variables

Severe Depression. Following Perreria and colleagues (2005), I measure depression with a modified version of the CES-D scale (Radloff 1977) that uses a subset of 5 items from the original 20-item scale.3 Respondents assessed the frequency of depressive symptoms including having the blues, feeling that life was not worth living, feeling depressed, feeling sad, and feeling happy throughout the 7 days leading up to the wave 2 interview. Responses were ordinal and ranged from 0 (“never or rarely)” to 3

(“most of the time or all of the time”).

To measure severe depression, I first summed the items, with the responses for

“felt happy” reverse-coded to indicate higher depression (α=.782). Scores potentially ranged from 0 to 15. I then identified adolescents with severe depression by setting gender-specific cut-points on the summed scale. Past research using the full version of

CES-D determined cut-points for identifying adolescents with major depressive disorders to be 24 for girls and 22 for boys (Roberts, Lewinsohn, and Seeley 1991). Following others who have examined severe depressive symptoms with CES-D subscales (Warren,

Harvey, and Henderson 2010), I account for the reduced number of items in the reduced

CES-D scale by lowering the cut-point for major depressive disorders to 6 for girls and boys. Severe depression, is binary and indicates whether the respondent’s depressive

3Compared to the full set of items, Perreria et al.’s (2005) abbreviated version of the CES-D scale provides a superior measurement of depressive symptoms with improved psychometric properties across race/ethnic groups. As an empirical check, I ran analysis with measures of major depression that are based on the measurement strategies of Joyner and Udry (2000) as well as Shields and Beaver (2011), who include more of the original items from the original CES-D scale. Results from those models were nearly identical to those presented in this chapter. 29

symptomology score was greater than or equal to the severe depression cut-point score

(0=no, 1=yes).

High Self-Esteem. I use an abridged version of the Rosenberg Self-Esteem Scale

(Rosenberg 1965) to measure self-esteem. At both interview waves, respondents were asked their level of agreement with 6 statements, including “you have many good qualities,” “you have a lot to be proud of,” “you like yourself just the way you are,” “you feel you are doing things just about right,” “you feel socially accepted,” and “you feel loved and wanted.” Responses were ordinal and ranged from 1 (“strongly agree”) to 5

(“strongly disagree”). To measure high self-esteem, I first reverse-coded the responses so that higher values indicated higher self-esteem and then summed the items (α=.856).

Unfortunately, there is no established cut-point associated with the Rosenberg Self-

Esteem Scale for identifying individuals with high versus low self-esteem. In this study, respondents were identified as having high self-esteem if their score was greater than or equal to one standard deviation above the mean of high self-esteem (0=no, 1=yes).4

Key Independent Variables

Relationship Contexts of Sexual Intercourse. I use responses to a series of questions regarding sexual activity within up to 3 “special” romantic and 3 non-romantic relationships between study waves to identify the relationship contexts of sexual behavior between waves. The first measure, sex with non-romantic partner, indicates whether

4I re-ran analyses with the cut-point for high self-esteem set to the 75th percentile as an empirical check. Those results were nearly identical to those presented in this study.

respondents had vaginal or anal intercourse with someone not identified as a romantic partner after the wave 1 interview (0=no, 1=yes). Respondents who indicated that they had non-romantic sexual intercourse with someone other than the three non-romantic partners were also coded as 1 for this measure. The second measure, sex with romantic partner only, indicates whether the respondent only had vaginal or anal intercourse within romantic relationships between waves (0=no, 1=yes). Abstinence between waves is the reference category and indicates that the respondent did not have sexual intercourse after the date of the first in-home interview (0=no, 1=yes).5

School sexual double standards. Following Soller and Haynie (2012a), I measure the severity of the sexual double standard within schools (at wave 1) with responses from three questions assessing respondents’ perceptions of the social benefits of sexual intercourse. These items, each prefaced by “If you had sexual intercourse,” include “it would make you feel less lonely,” “it would make you more attractive to men/women,” and “your friends would respect you more” (α=.699).6 Responses were ordinal ranging

5Interview sections pertaining to sexual behavior were conducted using Computer- Assisted Self-Interview (CASI), where questions were heard through headphones and displayed on a screen and responses were entered into a computer by the subject. This method helps increase the accuracy of answers by limiting interviewer-induced biases (Turner et al. 1998).

6Ideally, I would rely on questions capturing whether respondents perceive that sexual activity leads to occupying specific stigmatized or esteemed roles to measure the sexual double standard. Unfortunately such questions are not available in Add Health. However, the meaning and significance of role terms, such as “slut” (Eder, Evans, and Parker 1995) versus “stunt” (Harding 2010), vary across, and even within, social contexts (see Staiger 2005). As a result, my measurement strategy, which focuses on sexual activity and social esteem, is best able to capture the extent to which adolescents perceive that sexual activity leads to occupying respected roles among one’s peers across several school contexts. 31

from 1 (“strongly disagree”) to 5 (“strongly agree”). I use a hierarchal item-response theory (IRT) model to scale the sexual double standard. This model captures school-level differences in the perceived benefits of intercourse among boys and girls. More detail on the IRT measurement approach is provided in the “Modeling Strategy” section.

Control Variables

I control for a number of factors that are either associated with depression and self-esteem or potentially confound the relationship between sexual activity, the sexual double standard, and boys’ and girls’ mental health.

Changes in peer attachment following sexual activity may account for the interactive association between the sexual double standard, sexual behavior, and boys’ and girls’ mental health. Accordingly, I control for baseline levels of school peer attachment and problems with school peers (measured at wave 1) as well as changes in these measures from wave 1 to wave 2.

At both study waves, respondents were asked how often they had trouble with other students since the start of the current school year. Responses were ordinal and ranged from 0 (“never”) to 4 (“everyday”). I use information on the frequency problems with school peers to create a measure of change in peer problems, which indicates the change in the frequency with which respondents had problems with school peers between waves. Negative values indicate decreased problems with peers, while positive values designate increased problems with peers (with 0 indicating no change). I also control for

prior peer problems which indicates the frequency with which respondents experienced problems with school peers prior to wave 1.

Changes in peer attachments between waves may also account for the interactive association between school sexual double standards, sexual behavior, and mental health.

During both study waves, respondents assessed their agreement with the following statements: “you feel close to people at your school,” “you feel like you are part of your school,” and “you are happy to be at your school.” Responses ranged from 1 (“strongly disagree”) to 5 (“strongly disagree”). To measure change in peer attachment, I first reverse-recode the items for these measures so that higher values indicate stronger attachment. I then generate two measures of peer attachment by taking the means of the items for each respective wave (αw1=.779, αw2=.790). I then subtract the value of peer attachment at wave 1 from the value at wave 2 to measure increases in peer attachment between waves. As with changes in peer problems, negative values for this measure indicate decreased peer attachment, while positive values capture incrased peer attachment (with 0 indicating no change). I also control for baseline peer attachment (i.e., peer attachment at wave 1).

I control for religiosity because it is known to be associated with sexual activity among adolescents (Rostosky et al. 2004). Religiosity is measured with four-items

(assessed at wave 1) that capture the frequency of prayer (responses ranged from 1 “At least once a day” to 5 “Never”), religious service attendance, and religious youth-group participation (responses ranged from 1 “Once a week or more” to 4 “Never”) and a variable indicating how important religion is to the respondent (responses ranged from

1=“Very Important” to 4=“Not important at all”). My measure of religiosity consists of the mean of the reverse-coded and standardized items, with higher values indicating greater religiosity (α=.845). I also include a variable indicating whether the respondent took a pledge to remain a virgin until marriage prior to wave 1 (0=no, 1=yes) (Bearman and Brückner 2001).

Strong parent-child attachments are protective against adolescent depressive symptoms (Brumariu and Kerns 2010). Accordingly, I control for parental attachment, which consists of a five-item scale that captures bonding between adolescents and their parents according to responses to questions such as “how close do you feel to your mother?” and “how much do you think your father cares about you?” Each question

(assessed at wave 1) was asked in reference to the mother and then the father, for a potential total of 10 questions. To account for respondents in single-parent households, I took the maximum value from each paired response and construct a five-item scale representing the mean of the items (α=.845).

I also control for respondents’ perceived benefits of sexual intercourse with a variable consisting of the three items used to measure the sexual double standard. This scale represents the individual-level empirical Bayes (EB)–adjusted intercept of the perceived consequences of intercourse from the IRT model used to measure school sexual double standards. I also control for age, gender, race (binary indicators for black,

Latino and other, with white as reference), single-parent household (0=no, 1=yes), and parental socioeconomic status (consisting of the standardized values of parents’ highest occupational status, income, and education level), which were all measured at wave 1. I

include a binary variable that captures prior sexual intercourse, which indicates whether the respondent had sexual intercourse prior to wave 1 (0=no, 1=yes). In all models, I also include a lagged dependent variable, which indicates either severe depression or high self-esteem.

I include a number of school-level measures to account for regional and other compositional characteristics. First, I control for school size, which represents the number of students on the school roster (at wave 1). I also include binary variables indicating whether the school is urban or rural (suburban as reference) and its region (West,

Midwest, and Northeast; South as reference). Finally, I include a measure of the proportion of students who had sexual intercourse (prior to wave 1). Table A.1 presents descriptive statistics for all variables in my analyses.

Modeling Strategy

Measuring School Sexual Double Standards. I use a multilevel Item-Response

Theory (IRT) model to measure gender differences in the perceived benefits of sexual intercourse across schools. The model consists of scale items at level 1, respondents at level 2 and schools at level 3. At level 1, respondents’ perceived benefits of intercourse randomly vary as follows:

2 2 Yijk   0 jk   pjk X pijk  eijk eijk ~ N0,σ  p1

where Yijk is the ith response to the sexual double standard scale by respondent j in school k and Xpijk is a dummy variable that equals 1 if response item i is to item p and 0 if

otherwise. Because I center each Xpijk around its respective grand mean, π0jk is the adjusted perceived benefits of sexual intercouse for respondent j in school k (i.e., the empirical Bayes adjusted intercept). Coefficient πpjk is the severity of item p, and eijk are measurement errors that are normally distributed, with variance σ2 and mean 0. Level-2 considers gender when estimating person-level intercepts and captures respondents’ perceived benefits of intercourse. It is expreessed as:

 0 jk  00k  01k Male jk  r0 jk r0 jk ~ N0, 

where β00k is the mean level of perceived benefits of sexual intercourse in school k

(adjusted for gender) and coefficient β01k represents the effect of male on the individual- level intercept. Because I center Malejk around it’s school (i.e., “group”) mean and allow the effect of maleness on the slope of the benefits of sexual intercourse to vary randomly across schools, β01k represents the within-school differnce between boys and girls in their perceived benefits of sexual intercourse in school k (i.e., school k’s sexual double standard). r0jk is an individual-level error term for the perceived benefits of intercourse that has mean 0 and variance τπ. Level-3 captures gender variation in the perceived benefits of sexual intercourse across schools and is expressed as:

00k   000  u00k u00k ~ N0,  00

01k   010  u01k u01k ~ N0,  01

where γ000 is the grand mean of the benefits to girls of intercourse and γ010 represents the mean difference between boys’ and girls’ anticipations of the benefits of sexual intercourse across the entire sample (i.e., the sexual double standard across all schools). 36

u00k is an error term capturing the degree to which the mean perceived benefits of intercourse among girls within school k diverges from the overall mean among the girls, and u01k is an error term indicating the degree to which the boys’ mean difference in the perceived benefits of intercourse in school k diverges from the overall difference between boys and girls. Error terms have means of 0 and assume covariance matrix:

   00k   01k ,  00k    .   00k ,  01k   01k 

A school in which β01k equals 0 would indicate boys and girls within the school do not differ with regard to their perceptions of the social benefits of intercourse (on average).

Values greater than 0 indicate boys perceive intercourse to be more socially beneficial than girls do (i.e., a sexual double standard), and values less than 0 indicate girls perceive intercourse to be more socially beneficial than boys do (i.e., a reverse sexual double standard).

The Sexual Double Standard and Psychological Well-Being. I construct two-level interacted hierarchical logit models to measure the association between school sexual double standards, sexual behavior, gender, and psychological well-being. Using severe depression as an example, the individual-level model is expressed as:

ij  01j Femaleij  02j Female * Romanticij  03j Female * Non  Romanticij  P 04j Maleij  05j Male * Romanticij  06j Male * Non  Romanticij    pj X pij p7 where ηij is the log-odds of severe depression for respondent i in school j; β01j and β04j are the gender-specific school intercepts, which represent adjusted mean log-odds of severe depression among girls and boys (respectively) in school j, and Femaleij are Maleij 37

are binary variables indicating whether respondent i in school j is either female or male

(0=no, 1=yes). The effects of only having sex with a romantic partner on the log-odds of severe depression among girls and boys in school j are captured by β02j and β05j

(respectively), while Female*Romanticij and Male*Romanticij are dummy variables that respectively indicate whether respondent i in school j is 1) female and only had sex with one or more romantic partners, or is 2) male and had romantic-only sexual intercourse

(0=no, 1=yes). Note Female*Romanticij is 0 if respondent i in school j is either male, abstained from sexual intercourse, or had non-romantic intercourse, and Male*Romanticij is 0 if respondent i in school j is either female, abstained from sexual intercourse, or had non-romantic intercourse. β03j and β06j capture the effects of non-romantic sexual intercourse on the severe depression among girls and boys (respectively) in school j.

Female*Non-Romanticij and Male*Non-Romanticij are binary and respectively indicate whether respondent i in school j is 1) female and had sexual relations with a non- romantic partner, or is 2) male sexual relations with a non-romantic partner (0=no,

1=yes). It should be noted that as abstained between waves serves as the reference category, β01j and β04j represent the log-odds of sexual intercourse for girls and boys

(respectively) who abstained from sex, adjusted for individual and school covariates in this specific model. Finally, βpj is a coefficient representing the effect of individual-level control variable p=7,…,P on depression in school j and Xpij represents the value of p=7,…,P for respondent i in school j.

I model gender-specific intercepts as randomly varying functions of the sexual double standard at level-2. The first two school-level equations may be written as:

Q 01j   100   110Double Standard j   120Intercourse j   qjWqj  u01j u01j ~ N 0,  01 q3 Q 04j   400   410Double Standard j   420Intercourse j   qjWqj  u04j u04j ~ N0,  04  q3 where γ100 and γ400 are the grand-mean intercepts, or the adjusted grand-mean log-odds of severe depression among abstaining girls and boys (respectively). Coefficients γ110 and

γ410 are the effects of the sexual double standard girls’ and boys’ intercepts (respectively), and Double Standardj represents the value of the sexual double standard within school j.

Coefficients γ120 and γ420 are the effects of the effects of the school-level of sexual activity on girls and boys intercepts (respectively) and Intercoursej is a continuous measure indicating the proportion of students in school j who had sexual intercourse prior to wave 1. γqj is a coefficient representing the effect of school control variable q=3,…,Q on the log likelihood of severe depression and Wqj is the value of variable q=3,…,Q for school j. Because gender variation in the effects of school-level control variables Q are not of theoretical interest in this study, I constrain the effects of all school-level control variables (except the proportion of students who have had sexual intercourse) on the gender-specific intercepts to be equal among boys and girls. Finally, u01j and u04j are school-level gender-specific error terms with means of 0 and variances of τβ01 and τβ04, respectively.

I also model the slopes of sexual behavior as non-randomly varying functions of the sexual double standard. The final level-2 equations capture the interactive effects of gender, sexual behavior, and the sexual double standard on the outcome and are expressed as follows:

02j   200   210Double Standard j

03j   300   310Double Standard j

05j   500   510Double Standard j

06j   600   610Double Standard j

where γ200 and γ500 represent the proportions of female and male respondents

(respectively) who only had sex with one or more romantic partners, and γ300 and γ600 are the proportions of girls and boys (respectively) who had sex with at least one non- romantic partner. Finally, γ210 and γ510 represent the effects of the sexual double standard on the slopes of romantic-only sex among girls and boys (respectively), and γ310 and γ610 represent the effects of the sexual double standard on the slopes of non-romantic sex among girls and boys (respectively).

Missing Data and Survey Weights. I used Imputation through Chained Equations

(ICE) to multiply impute missing values for independent variables with the ICE command in Stata 12 (Royston 2004). I estimated hierarchical logistic regression models with 10 imputed data sets, using the multiple imputation procedure in HLM7. Individual- and school-level survey weights that account for Add Health’s complex survey design were applied (Chantala and Tabor 2010). To aid the interpretation of the results, all independent variables not pertaining to gender or sexual behavior are centered at their respective grand-means.

RESULTS

Variation in the Sexual Double Standard

I first examine the results from the multilevel IRT model used to measure the sexual double standard and assess the extent of its variation across the schools in our sample (results not displayed, available upon request). Results indicate on average, boys perceive there to be greater social benefits associated with sexual intercourse than girls

(γ010=.597, p<.01), suggesting a sexual double standard exists across all respondents in this study. The variance component τβ01 quantifies the extent of variation in the differences by gender of the perceived benefits of sexual intercourse across schools. The significant variance component (τβ01=.017, p<.01) indicates that within-school differences in perceived benefits of sexual behavior vary across contexts, suggesting the severity of the sexual double standard fluctuates across high schools in Add Health.7

The Sexual Double Standard, Sexual Intercourse and Mental Health

I now turn to results from models that test the association between the sexual double standard, gender, sexual behavior, and mental health. Across all models, I control for school and individual-level measures previously discussed.

Relationship Context of Sexual Behavior and Severe Depression. Table A.2 displays coefficients from multilevel logistic regression models measuring the association between the sexual double standard, gender, relationship contexts of sexual behavior, and severe depression. Model 1 tests the association between the sexual double standard and

7Boys on average anticipated greater benefits associated with sexual intercourse than girls in all but 2 schools in the sample. To ensure that my results were not driven by outliers, I re-ran my analyses with these two schools excluded. Those analyses resulted in associations between the sexual double standard, sexual behavior, and the outcomes that were statistically significant and stronger in magnitude than those presented in this chapter. 41

severe depression, controlling for individual and school factors. The positive and significant coefficient for Sexual Double Standard*Female indicates the salience of the school sexual double standard frame is positively associated with the likelihood of severe depression among girls (b=1.11, p<.05). The coefficient for Sexual Double

Standard*Male is positive but not statistically-significant (b=1.24, n.s.). Model 2 introduces variables capturing the relationship context of sexual activity. The non- significant coefficients for the sexual activity variables indicate romantic and non- romantic sexual intercourse between waves are not associated with severe depression at wave 2 after controlling for prior severe depression. In addition, introducing the sexual behavior variables renders the coefficient for Sexual Double Standard*Female non- significant.

Model 3 includes cross-level interactions between the sexual double standard and sexual relationship context categories. Doing so changes the interpretations of the coefficients for Sexual Double Standard*Female and Sexual Double Standard*Male such that they now capture the association between the sexual double standard and severe depression among girls and boys (respectively) who abstained from sexual intercourse between waves. Adding these cross-level interactions to the model reverses the direction of the coefficient for Sexual Double Standard*Female from Model 2, although the coefficient is non-significant (b=-.50, n.s.).

Positive and significant values for the coefficients on the cross-level interactions introduced in Model 3 would suggest that sexual behavior is more likely to lead to severe depression as the school sexual double standard increases. Among girls, I find a positive

and significant coefficient for the Sexual Double Standard*Female*Romantic Sexual

Intercourse interaction (b=2.08, p<.05) and a positive and significant coefficient for the

Sexual Double Standard*Female*Non-Romantic Sexual Intercourse interaction (b=4.42, p<.01). Together these results suggest the sexual double standard is not associated with the likelihood of experiencing severe depression among girls who abstain from sex between waves. However, compared to their abstaining counterparts, girls who had sexual intercourse with either a romantic or non-romantic partner have higher likelihoods of severe depression as the sexual double standard increases. Importantly, having intercourse with a non-romantic partner appears to result in an even larger increase in the likelihood of severe depression as the sexual double standard becomes more salient. I find no evidence that the sexual double standard is associated with severe depression across the three sexual behavior categories among boys.

I display the interactive association between sexual relationship categories, the sexual double standard, and girls’ severe depression in Figure B.1. This figure displays the relationship between predicted probabilities of girls’ severe depression, the sexual double standard, and sexual behavior and is based on estimates from Model 3 in Table

A.2. with all other variables held at their means. Compared to girls who abstained from sexual intercourse between waves, girls reporting sex with a non-romantic partner actually have a lower predicted probability of reporting severe depression at wave 2 at lower ends of the school sexual double standard. In addition, there is little difference in the predicted probabilities of severe depression among girls who abstained or only had intercourse with a romantic partner at lower ends of the sexual double standard. However

girls who had sexual intercourse with a non-romantic partner are increasingly more likely to report severe depression as the school sexual double standard increases. The upward slope for girls who only reported sexual intercourse with romantic partners is more modest when compared to the line representing non-romantic partners, however the figure illustrates the positive and significant association between romantic sexual intercourse and severe depression as the sexual double standard in a school increases.

Relationship Context of Sexual Behavior and High Self-Esteem. I now turn to models of high self-esteem. Table A.3. displays coefficients from interacted multilevel logistic regression models that test hypotheses 2.A. and 2.B. Model 1 measures the association between the sexual double standard and high self-esteem, controlling for prior high self-esteem and other individual- and school-level factors. Results indicate the sexual double standard is positively associated with boys’ likelihood of reporting high self-esteem (b=1.12, p<.05). The coefficient for Sexual Double Standard*Female is negative, although not statistically significant (b=-.21, n.s.). Model 2 introduces measures capturing the relationship contexts of sexual intercourse. None of the measures introduced in this model are statistically significant, suggesting on average, sexual intercourse is not associated with boys’ or girls’ self-esteem after controlling individual and school-factors.

Interestingly, the Sexual Double Standard*Male coefficient remains positive and significant in Model 2, suggesting boys are more likely to report high self-esteem as the sexual double standard increases (b=1.18, p<.05). The magnitude of the coefficient is such that when the sexual double standard is 1 standard deviation below the mean, the

probability of reporting high self-esteem for an average boy (i.e., with all control variables held at their means) who abstained from sex between waves is .17. Conversely, the probability of reporting high self-esteem for an otherwise average boy who abstained from sex between waves is .22 when the school sexual double standard is at 1 standard deviation above the mean. The differences in these probabilities suggest all boys experience a modest boost in self-esteem as the sexual double standard in a school increases. I find no evidence that the sexual double standard is associated with high self- esteem among girls.

Model 3 (Table A.3) introduces cross-level interactions between the sexual double standard and sexual relationship context variables. Including these interactions means the coefficient for Sexual Double Standard*Male now captures the association between the sexual double standard and high self-esteem among boys who abstained from sexual intercourse between waves. Introducing the interactions decreases the Sexual

Double Standard*Male coefficient from the Model 1 (Table A.3.) by roughly 15%, however the sexual double standard remains positively and significantly associated with high self-esteem among boys reporting no sexual partners between waves (b=1.00, p<.05). Including the cross-level interactions drastically increases the magnitude of the coefficient for Male*Non-Romantic Sexual Partner (b=-.40, p<.01), indicating that when the sexual double standard is at or below the mean, boys who had a non-romantic sexual intercourse have lower odds of reporting high self-esteem compared to boys who abstained from sexual activity. Introducing the cross level interactions in Model 2 does

not drastically alter the coefficients pertaining to the sexual activity categories among girls or romantic only sexual partners among boys observed in the previous model.

The positive and significant coefficient for Sexual Double Standard*Male*Non-

Romantic Sexual Intercourse in Model 3 (b=3.11, p<.05) indicates boys who had sexual intercourse in non-romantic relationships were more likely to report high self-esteem as the sexual double standard in their schools increased. The coefficient for Sexual Double

Standard*Male*Romantic Sexual Partner is non-significant (b=-.63, n.s.), suggesting that having romantic-only intercourse does not increase the likelihood of high self-esteem as the sexual double standard, compared to boys reporting no sexual intercourse. Finally, the interactions between the sexual double standard and the sexual relationship context categories for girls are both non-significant, suggesting among girls, sexual behavior is no more or less likely to lead to high self-esteem at different levels of the sexual double standard.

Figure B.2. graphically-illustrates the interactive association between the sexual relationship context categories, the sexual double standard, and high self-esteem among boys. Displayed are the predicted probabilities of high self-esteem across different levels of the sexual double standard for boys who abstained from sexual intercourse, only had intercourse with romantic partners, and had sex intercourse one or more non-romantic partners. This figure is based on estimates from Model 2 in Table A.3. with all other independent variables held at their means. Across the sexual intercourse categories, boys who reported having sexual intercourse with a non-romantic sexual partner have the lowest probabilities of reporting high self-esteem when the sexual double standard in

their schools is low. It is notable that there is little difference in the probabilities of high self-esteem among boys who reported having no partners and those reporting romantic- only intercourse when school sexual double standards are low. However, as the sexual double standard increases, boys who reported having sexual intercourse with a non- romantic partner are increasingly more likely to report high self-esteem. The upward slopes for boys who abstained or had romantic-only sexual intercourse are much more modest when compared to the slope for the line representing boys who had non-romantic intercourse, indicating that the positive association between the sexual double standard and high self-esteem is stronger in magnitude for boys who have sex outside of romantic relationships.

DISCUSSION AND CONCLUSION

Relying on data from a nationally-representative sample of adolescents, this study focused on the potential for a sexual double standard pertaining to gender differences in the social benefits that follow sexual intercourse to alter the association between sexual behavior and subsequent depression and self-esteem. I proposed cultural frames imbued with sexual double standards inform self-appraisals and help adolescents identify how their sexual behavior fits into certain esteemed or devalued roles. Building on a recent study by Soller and Haynie (2012a), I measured sexual double standards across 75 schools in Add Health by quantifying the extent to which boys perceive there to be more social rewards associated with sexual intercourse than do girls in a school. I then tested

whether school sexual double standards alter the association between sexual behavior and boys’ and girls’ depression and self-esteem.

Results indicated among girls, having sexual intercourse (with either romantic or non-romantic partners) was associated with higher likelihoods of experiencing severe depression as the sexual double standard in their school increased. Conversely, school sexual double standards did not alter the association between either measure of sexual intercourse and depression for boys. However, sexual behavior was associated with high self-esteem among boys as school sexual double standards increased. In particular, boys who reported having sexual intercourse with at least one non-romantic partner were more likely to have high self-esteem as the sexual double standard in a school increased.

School-based sexual double standards did not alter the association between sexual intercourse and high self-esteem among girls. Importantly, changes in peer attachments and problems with other students following sexual activity did not account for the interactive associations observed in the present study.

It is noteworthy that sexual intercourse in either romantic or non-romantic relationships was not on average associated with girls’ severe depression or high self- esteem after controlling for a number of individual factors, including the lagged dependent variable. In addition, coefficients capturing the partial effects of the romantic and non-romantic sexual intercourse were non-significant upon interacting sexual intercourse categories with the school sexual double standard. Together these results suggest sexual activity has a negligible effect on girls’ psychological well-being within

schools in which the sexual double standard is less severe. Among boys, sexual behavior on average was not associated with high self-esteem.

Results from this study reflect those from past research focusing on how sexual behavior informs boys’ identities. Anderson (1989) demonstrates boys who successfully convey (hetero)sexual prowess to their peers are revered within the socially-isolated, black urban ghetto he studied. Anderson suggests the premium placed on sexual activity among his male respondents is rooted in the neighborhood’s socioeconomic characteristics. Perceiving a low likelihood of achieving conventional markers of manhood (e.g., heading an economically self-sufficient family) due to concentrated disadvantage and racial/social isolation, Anderson suggests males develop alternative barometers for masculine-identity construction, in which having sexual relations with as many girls and women as possible is central for achieving esteemed social identities. In support of this assertion, a recent study by Giordano and colleagues (2009) found that roughly one-third of black males residing in disadvantaged neighborhoods themselves self-identified as “players.” However the majority of the boys who self-identified as

“players” were not black males living in disadvantaged contexts. Results from Giordano and colleagues’ study suggest the link between sexual activity and esteemed male roles transcends race and potentially socioeconomic status. Findings from the present study suggest this process is borne out in contexts characterized by sexual double standards regarding differences in the social benefits of sexual activity among boys and girls.

Findings from this study also reflect those from recent work that suggest sexual double standards condition the association between sexual activity and girls’ and

women’s subsequent mental and emotional health. Hamilton and Anderson (2009) suggest sexual double standards regarding the social consequences of sexual activity contribute to sexual stigma among women who engage in hook-ups within university settings. In another study, Armstrong, England, and Fogarty (2012) suggest a sexual double standard regarding the gender-differences in the entitlement of sexual pleasure in hookups lead to a lower likelihood that women achieve orgasms within hookups

(compared to intercourse within romantic relationships). While the authors do not focus on the mental health consequences of casual sexual encounters, results from these studies suggest sexual double standards limit the sexual pleasure and emotional payoffs girls and women may derive from sexual activity.

Adding to the emerging literature focusing on differences in the personal consequences of sexual activity, I suggested sexual activity most likely contributes to poor mental health among girls who are embedded in contexts characterized by strong sexual double standards. However, I proposed this relationship is largely due to processes related to how cultural frames inform self-appraisal and role-taking processes.

Adolescents rely upon cultural frames to help make sense of sexual behavior and identify how it fits into certain social roles (Soller et al. 2012). As sexual double standards become more salient, sexually-active girls may be increasingly likely to perceive they occupy stigmatized roles within school contexts. Conversely, I suggested sexual activity enhances boys’ mental health as the sexual double standard increased. This is because the sexual double standard frame may lead sexually-active boys to perceive they fulfill esteemed roles within the school context.

Apart from contributing to the theoretical understanding the sexual double standard, this study has a number of policy implications. Abstinence-only education programs—sex education that teaches sexual abstinence without providing other types of sexual and reproductive health education such as birth control—still receive public funding. One major justification of abstinence education programs is that early and nonmarital sexual activity harms adolescents’ psychological well-being (Social Security Act

1996). However as Meier (2007) notes, little sociological research has focused on the detrimental effects of sexual activity on adolescent psychological well-being. As a result, a major impetus for abstinence education programs remains under-scrutinized. Even less is known about how social contexts factor into the association between sexual behavior and subsequent mental health, despite increased attention of “slut shaming” in the popular press.

Findings from this study indicate the association between adolescent sexual activity and subsequent mental health is in no small part a function of sexual double standards within school peer contexts. Among girls, the adverse effects of sexual behavior on their mental health appear to be fundamentally shaped by the cultural contextd in which they are embedded. Importantly, sexual intercourse appears to have negligible effects on subsequent depression among girls when enacted in schools with less severe sexual double standards. Thus a primary goal for abstinence education—to enhance mental health through abstinence—appears to be relevant to contexts replete with sexual double standards regarding the personal consequences of sexual intercourse.

Perhaps a more fruitful approach to alleviating the harmful consequences of sexual

activity among girls is to address the structural factors that contribute to sexual double standards within schools and other contexts.

At the same time, non-romantic sexual intercourse enhanced boys’ self-esteem as the sexual double standard increased. On the surface, this association suggests sexual double standards benefit boys’ psychological well-being by enhancing the self-esteem of those who engage in casual sex. However, the positive association between sexual behavior and boys’ self-esteem may be dependent upon an alternative barometer for achieving positive masculine roles that reinforce the sexual double standard examined in this study. As Anderson (1989) notes, sexual behavior is such a salient marker of positive masculine identities in the community he studied because males perceive they are unable to achieve more conventional markers of manhood (e.g., heading economically self- sufficient household). Regarding the present study, the link between casual sexual intercourse and high self-esteem among boys attending schools with strong sexual double standards may occur because markers of conventional masculine identity are less salient aspects of esteemed male roles within such school contexts. Most importantly, the boost in self-esteem among males who have non-romantic sexual intercourse is contingent upon a cultural frame that likely contributes to poor mental health for sexually-active girls. Further research focusing on the emergence of esteemed sexualized male roles (e.g., the player) may help identify the processes that determine the salience of sexual double standards within schools.

While this study advances the understanding of the association between sexual double standards, sexual behavior, and subsequent mental health, it is not without

limitations. First, I was unable to measure the extent to which adolescents perceive that they occupy esteemed or stigmatized roles following sexual behavior because Add Health did not gather information regarding self-appraisals and role-taking processes. As a result, the significant associations found in this analysis may be due to alternative mechanisms, such as increased stigma (among girls) or peer esteem (among boys) from school peers following sexual activity. However, although I was unable to directly measure peer stigma and acceptance that followed sexual activity, changes in problems with peers and peer acceptance between waves did not account for the interactive association between school sexual double standards, sexual activity, and girls’ depression. Thus overt peer reactions to sexual activity (e.g., increased/decreased peer acceptance) most likely do not account for the interactive association between gender, sexual behavior, and the sexual double standard. Rather, the association between sexual behavior and adolescents’ mental health appear in large part to depend on how adolescents themselves conceptualize how their own sexual behavior fits into their roles within the larger school context. Further research that focuses on how sexual behavior informs self-appraisal and role-taking processes may help clarify the mechanisms through which sexual double standards alter the association between sexual behavior, gender, and subsequent mental health.

In addition, my study only examined one particular aspect of sexual double standards, namely gender differences in the perceived social benefits that follow sexual activity. As noted by others (Armstrong et al. 2012; Hamilton and Armstrong 2009), double standards regarding gender differences in expectations for reaching orgasm in

hookups, the acceptability of casual sex among men versus women, and other aspects of sexuality abound. Thus different components of sexual double standards may differentially alter the association between sexual behavior, gender, and psychological and social well-being. Further research that attends to the multidimensional nature of the sexual double standard may shed more light onto the understanding of how sexual double standards affect the association between sexual behavior and subsequent well-being.

As Meier (2007) notes, early pregnancy, sexually-transmitted infections, and poor mental are potential consequences of adolescent sexual behavior. However, the vast majority of sexually-active teens do not contract sexually-transmitted infections and do not become, or get someone else, pregnant. In addition, the association between adolescent sexual behavior and low self-esteem and depression appear to be contingent upon a number of individual factors. Findings from this study suggest cultural contexts are also key in determining how sexual activity impacts adolescents’ subsequent health and well-being.

I wish to caution against interpreting this study as an attempt to identify conditions under which adolescent sexual activity entails positive or neutral effects on psychological well-being. Rather, this study represents an effort to understand the mechanisms through which gendered cultural structures interact with gender and sexual behavior to shape adolescent mental health. That said, for better or worse, sexuality is an integral aspect of adolescent identities (Tolman and McClelland 2011; Welsh, Rostosky, and Kawaguchi 2000). Accordingly, we must not shy away from the fact that understanding the contextual factors that promote or hinder the development of

normative and healthy sexuality during adolescence is a worthy endeavor. Future exploration into the association between culture and teens’ sexuality may promote the development of healthy sexual practices during adolescence and throughout the life course.

Chapter 3: “I Did it My Way”: The Peer Context of Adolescent Romantic Relationship Inauthenticity

Adolescence is typically a time of increasing involvement in romantic relationships (Collins, Welsh, and Furman 2009; Giordano 2003). Recently, 36% of 13- year-olds and 70% of 17-year-olds reported having a “special romantic relationship” within the past 18 months (Carver, Joyner, and Udry 2003). Despite the increasing prevalence of romantic relationships throughout adolescence, little sociological research focuses on the dynamics of adolescent romantic relationships (Brown et al. 1999), especially when compared to the large number of studies on adolescent friendships and sexual relationships (Giordano, Manning, et al. 2006).

Emerging research on adolescent romantic relationships suggests they are developmentally significant. Early sexual behavior most often occurs with romantic partners (Manning et al. 2005). Romantic relationships are also linked to adolescent depressive symptoms (Joyner and Udry 2000). Under certain conditions romantic relationships have been shown to facilitate delinquent behavior through partner influence and strain (Aalsma et al. 2012; Kreager and Haynie 2011; McCarthy and Casey 2008). In addition, psychological aggression also occurs within roughly one-third of heterosexual adolescent romantic relationships, while physical aggression occurs within roughly 1 in

10 relationships (Halpern et al. 2001). Conversely, intimate and supportive romances can

enhance adolescent well-being (e.g., Simons and Barr forthcoming) and lead to more successful adult union formations (Collins 2003).

Relationship inauthenticity—the level of incongruence between what one thinks and feels and what one actually says and does within relational contexts—is one way in which romantic relationships influence adolescent development and well-being (Impett et al. 2008). Unfortunately, youth often forgo their emotional needs and desires in efforts to avoid conflict and maintain close relationships, which may lead to relationship inauthenticity (Impett et al. 2008). Understanding the conditions which lead to relationship inauthenticity is important since compromised authenticity within relationships has been associated with low self-esteem and depression among adolescents

(Impett et al. 2008; Theran 2011; Tolman et al. 2006) and diminished sexual self-efficacy among girls (Impett et al. 2006).

Up to this point, little research has explored the social factors that contribute to inauthentic romantic relationships. Building on insights from cultural sociology, social network perspectives, and social learning theory, the current study focuses on romantic relationship “scripts”—cultural templates for ordering behavior within romantic contexts—to conceptualize how friends influence relationship inauthenticity.

Specifically, I examine the link between “script discordance,” defined as the degree to which the sequencing of events within one’s ideal romantic relationship script diverges from the sequencing of events within close friends’ ideal scripts, and adolescent romantic relationship inauthenticity. I argue friends influence relationship inauthenticity in large part through reinforcement processes. Individuals’ scripts are more strongly reinforced

when there is greater agreement between their scripts and the scripts of their friends.

Conversely, one’s ideal script is less likely to be reinforced when it is discordant with the scripts of one’s friends. Lacking peer reinforcement for their ideal strategy of action, adolescents may be more likely to enact inauthentic romances when they experience high levels of script discordance.

This chapter explores the relationship between script discordance and romantic relationship inauthenticity with data from the National Longitudinal Study of Adolescent

Health (hereafter, Add Health). Using sequence analysis (Abbott and Tsay 2000), I measure romantic relationship inauthenticity by quantifying the extent to which the ordering of events within one’s ideal romantic relationships script (e.g., holding hands, going on dates, having sex, etc.) diverges from the sequencing of events within one’s subsequent romantic relationship (see Harding 2007). I measure script discordance by quantifying the extent to which the ordering of events in respondents’ ideal scripts deviates from the sequencing of close friends’ ideal scripts. I then test the association between script discordance and relationship inauthenticity. Finally, I test whether respondents’ level of involvement with friends and the overall level of popularity of their friendship group alter the association between script discordance and relationship inauthenticity.

ADOLESCENT RELATIONSHIP INAUTHENTICITY

Developmental psychologists are increasingly focusing on the causes and consequences of inauthentic relationships among adolescents. Relationship inauthenticity

refers to a condition in which individuals behave in manners that are inconsistent with how they desire to conduct themselves. Impett and colleagues (2008) note relationship inauthenticity differs from related concepts such as self-assertion (i.e., sharing needs and wants to realize agreement with others) and self-disclosure (i.e., communicating intimate details about oneself). Inauthenticity specifically entails inaccurate behavioral representations of how one thinks and feels within relationships.

According to developmental psychological perspectives, relationship inauthenticity largely stems from age- and gender-related pressures to maintain relationships at the expense of one’s true feelings and desires. Impett and colleagues

(2008) argue girls openly articulate positive and negative feelings to others in early childhood. But in early adolescence, girls encounter pressures to act in ways that are inconsistent with their true feelings and desires. Such pressures may result in a condition in which girls are reluctant to express their needs and emotions in order to avoid relationship conflict (Brown and Gilligan 1993; Tolman 2005). Boys also face pressures to behave in inauthentic ways in peer settings. For instance boys are pressured to express hypermasculinity in their behavior in order to achieve and maintain peer status (Chu,

Porche, and Tolman 2005; Pascoe 2007; Pollack 1998). Gender-specific pressures potentially lead to inauthentic relationships with friends and romantic partners among adolescents (see Pascoe 2007).

Psychologists have identified a number of individual characteristics that are associated with relationship inauthenticity. Among girls, low body image is positively associated with inauthenticity, while inauthenticity tends to decline with age (Impett et al.

2008). Research also indicates features of romantic dyads influence romantic relationship inauthenticity. Among college students, Neff and Suizzo (2006) found those occupying subordinate positions in romantic partnerships experience more relationship inauthenticity than those who occupy equal or socially-dominant positions.

While characteristics of individuals and romantic partners influence relationship dynamics, processes related to friendship groups may also play important roles in adolescent relationship inauthenticity. In the following sections, I build on insights from cultural sociology, social learning theory, and social network perspectives to formulate hypotheses regarding the association between script discordance, structural features of friendship groups, and relationship inauthenticity.

SCRIPT DISCORDANCE AND STRATEGIES OF ACTION

My analysis builds upon an emerging approach to culture that focuses on cognitive processes to explain how culture shapes action. Proponents of this perspective suggest culture informs behavior in large part by informing behavioral repertoires that individuals draw upon throughout interaction (Harding 2007). Culture in part consists of strategies that individuals incorporate into a continuously-evolving repertoire. Culture in turn shapes action by providing individuals with behavioral strategies that may be used to

“manage the social world” (Lamont and Small 2008:81). In this respect, culture informs which scripts or strategies of action are possible and more or less probable.

David Harding, a leading proponent of cognitive cultural theory, has conducted numerous studies focusing on cultural heterogeneity, which he defines as the “presence

of a diverse array of competing and conflicting cultural models” (Harding 2010:143). In one study, Harding (2007) used data from Add Health to demonstrate heterogeneity in relationship scripts within neighborhood contexts decreases the correspondence between the ordering of events in one’s ideal romantic script and sequencing of events within one’s actual subsequent relationship script. He argues cultural heterogeneity promotes deviation from ideal scripts in romantic relationships through exposure to a wider array of strategies of action that may be followed within relationships. Conversely, cultural homogeneity diminishes script variability, thus increasing the likelihood that one will

“stick to the ideal script” in romantic relationships.

While Harding emphasizes the consequences of heterogeneity in relationship scripts, I focus on the extent to which individuals’ ideal scripts deviate from their friends’ ideal scripts. My approach recognizes the importance of cultural heterogeneity within larger contexts. However, I argue experiencing script discordance increases relationship inauthenticity primarily through reinforcement processes, rather than exposure to heterogeneous arrays of scripts.

Distinguishing between heterogeneity and discordance is theoretically important.

Script heterogeneity, as measured by Harding (2007), captures the extent to which each member’s script within a particular context differs from one another. Heterogeneity increases with the extent to which every individual’s script entails a different ordering of events from everyone else’s script. However, one may be embedded in a context in which there is complete script homogeneity among others, but experience a high level of script discordance. This condition is illustrated in Figure D.1. In this figure, idealized romantic

relationship scripts are represented by the sequences of letters. From individual i’s perspective, there is complete script homogeneity among the other group members.

However there is a great deal of script discordance for individual i, given i’s script is a polar opposite of every other group members’ script. Focusing on the extent of heterogeneity at the group level in this case would grossly underestimate the individual’s exposure to strategies of action that diverge from his or her own script. As such, I focus on the extent to which the sequences of events within individuals’ own scripts diverge from the scripts of their friends, rather than the overall extent of script heterogeneity among group members. I examine the association between script discordance and adolescent relationship inauthenticity within the context of adolescent friendships, given the importance of peers in adolescent development and romantic relationship dynamics

(Connolly et al. 2004; Giordano 2003).

THE PEER CONTEXT OF RELATIONSHIP INAUTHENTICITY

Certain features of peer contexts may influence inauthenticity in adolescent relationships. In her ethnography of a Northern California high school, Pascoe found girls often engaged in “compulsive heterosexuality” (2007:86) in order to capture boys’ erotic attention and bolster peer social status. Conversely, boys achieved peer status in large part through compulsive heterosexuality in which they broadcasted exaggerated (and often fictitious) tales of sexual conquests to groups of friends. However, during one-on- one situations with Pascoe, boys were far less likely to behave in manners that expressed sexual dominance and actually articulated strong emotional connections with girls and

feelings of insecurity with the opposite sex. As these findings suggest, features of adolescent peer contexts may promote compulsive practices and public displays that counter one’s inner feelings and desires.

Research suggests cultural conceptions of romantic and sexual relationships among close friends and peer groups influence adolescents’ behavior in romantic and sexual relationships. For instance, Lyons and colleagues (2010) found girls to be keenly aware of sexual double standards—differing standards of sexual permissiveness across gender groups—within school contexts. However, the authors found liberal attitudes towards sexual permissiveness within some girls’ close-knit friendship groups protected some against the adverse consequences that may accompany sexual activity within schools with strong sexual double standards (e.g., social exclusion). Similarly, Soller and

Haynie (2012b) found a strong association between sexualized relationship scripts among adolescent peer group members and sexual intercourse but found no association between school-level sexualized scripts and intercourse. Such findings suggest relationship scripts among smaller, more intimate peer groups are particularly important in guiding adolescents’ behavior in sexual and romantic relationships.

But what are the processes through which close-knit peer groups shape romantic relationship inauthenticity? As Harding (2010:143–144) suggests, the presence of particular cultural models requires social support within larger contexts. While institutions such as religions, schools, and media sustain certain cultural frames and scripts, models that are not widely reinforced by the larger society are primarily sustained through social network processes within localized contexts (e.g., neighborhoods, peer

groups). Exactly how interpersonal relationships affect the development and performance of individual strategies of action is one of the more underdeveloped aspects of cultural sociology.

One way in which friends may influence relationship inauthenticity is through reinforcement processes. Ideal relationship scripts that are congruent with friends’ ideal scripts are reinforced by group members throughout interaction. Conversely, discordant scripts are less likely to be reinforced by group members. I argue high levels of script discordance—a condition in which the sequence of events within one’s ideal script contradicts the ordering of events in one’s friends’ scripts—promotes relationship inauthenticity because one’s script finds little support among group members. Failing to find support for one’s ideal strategy of action may lead to the abandonment of the ideal script and the adoption of an alternative script.

This brings me to my first hypothesis. Given that script discordance leads to a condition in which one’s own ideal script is less strongly reinforced among peer group members, I hypothesize:

H.1. discordance between an individual’s script and the scripts of his or her friends will be positively associated with relationship inauthenticity.

I argue friendship networks link strategies of action from larger contexts and relationship inauthenticity in two important ways. First, friends reinforce certain scripts among group members. Second, structural features of friendship networks determine the frequency of interaction and the intensity of peer associations. I argue peer network structure influences romantic relationship inauthenticity in large part by altering reinforcement processes. Below I incorporate insights from differential association/social 64

learning theory (Akers 2009) and social network perspectives to specify the conditions under which script discordance will be most strongly associated with relationship inauthenticity.

NETWORK PROCESSES AND RELATIONSHIP AUTHENTICITY

Differential association theory (1947) suggests numerous mechanisms through which network structure alters the association between script discordance and relationship inauthenticity. Although the theory is most often applied in studies of peer effects on delinquency, it suggests the process of learning criminal behavior entails the same mechanisms as learning conforming behavior. Therefore, this general theory of behavior may be used to understand peer influence on behavior within romantic relationships.

As Sutherland (1947) notes, associations vary in their priority, duration, frequency, and intensity. Priority relates to the chronology of associations throughout the life course. Frequency and duration capture the regularity and length of interaction with associations. Intensity refers to the “prestige” of the source of a criminal or anti-criminal pattern of behavior. Expanding upon Sutherland’s theory, Akers’ (2009) social learning theory suggests priority, duration, frequency, and intensity influence offending because they determine the extent to which groups reinforce behavior and definitions. Differential reinforcement, defined as “the balance of anticipated or actual rewards and punishments that follow or are consequences of behavior” (Akers 2009:67), represents an important mechanism through which group members selectively reinforce certain behavioral scripts

among group members. Behavior that aligns with definitions that are reinforced among friendship groups is more likely to be enacted than those that lack reinforcement among group members. Importantly, reinforcement is not constant, but is rather a function of priority, duration, frequency, and intensity. I test whether frequency—as measured by the extent of interactions with friends—and intensity—as measured by friends’ social status—alter the association between script discordance and relationship inauthenticity.

From a social network perspective, high levels of interaction with friends provide opportunities for individuals to articulate and reinforce their own ideal ordering of romantic relationships to others. Thus, the extent of one’s interactions with friends may impact relationship inauthenticity by providing opportunities for group members to reinforce certain scripts. As reinforcement supports congruent scripts, frequent interactions with friends who maintain concordant scripts will likely lead to more authentic relationships. Conversely, adolescents may be increasingly likely to disregard their ideal scripts when they frequently interact with friends who maintain and reinforce discordant scripts. Individuals may thus be especially likely to enact inauthentic relationships when they frequently interact with friends who have discordant relationship scripts. This brings me to my second hypothesis:

H.2. interaction with friends accentuates the positive association between script discordance and relationship inauthenticity.

The intensity of associations is consequential because it determines the probability that certain scripts are reinforced within friendship groups. Intensity was never precisely defined by Sutherland. As a result, it remains one of the more nebulous components of differential association theory and has been operationalized in different 66

ways in empirical tests of the theory (Haynie 2001; Matsueda 1982). However, Akers suggests intensity has to do with the “significance, salience, or importance of the association to the individual” (Akers 2009:65). I argue friends who are of high social status within larger peer settings represent particularly significant and salient associations as they may entail more social rewards (e.g., increased popularity) and bolster one’s own peer status (Dijkstra et al. 2010). Accordingly, behavior and scripts are likely more strongly reinforced within high status peer groups than within low status peer groups.

Research suggests high social status alters peer influence. For instance, Haynie

(2001) found beta centrality—a network measure of status that weighs status by the status of one’s friends—accentuates the association between peer and respondent delinquency. Similarly, friends’ social status has been shown to accentuate peer influence on both pro- and antisocial behavior and attitudes (Allen et al. 2012; Cohen and Prinstein

2006; Ellis and Zarbatany 2007; Shi and Xie 2012). Cohen and Prinstein (2006) argue friends’ social status promotes peer influence because of a desire to emulate behavior that is performed by high status others.

From a social learning/differential association theoretical perspective, reinforcement involves a wide range of social rewards that are valued within the larger society and its constituent subgroups. Within adolescent school contexts, having friends with high social status is socially-rewarding because such friends signify, bolster, and help maintain one’s own social standing. Accordingly, the status of one’s peers is directly linked to the intensity (i.e., salience or importance) of associations. The intensity of

associations in turn shapes the probability of script reinforcement from the larger peer group.

If the intensity of one’s associations promotes reinforcement processes, high status peer groups more persuasively reinforce particular scripts than groups that are lower in the overall social hierarchy. This would lead to a situation in which scripts that are congruent with popular peers’ scripts are more likely to be reinforced and enacted in subsequent relationships. Conversely, adolescents who maintain discordant ideal scripts and who are embedded in high status peer groups are especially likely to experience relationship inauthenticity because particularly salient associations do not reinforce their scripts. If these processes occur, then friends’ popularity likely accentuates the positive association between script discordance and the extent to which individuals deviate from their ideal romantic relationships in subsequent relationships. This brings me to my final hypothesis:

H.3. the average status of one’s peers accentuates the positive association between script discordance and relationship inauthenticity.

METHODS

Data

Add Health is a nationally representative longitudinal school-based study that explores the etiology of health outcomes and behaviors among young people in the

United States. All U.S. high schools that included an 11th grade and had at least 30 enrollees were eligible for participation. A random sample of 80 high schools was compiled that was stratified by region, urbanicity, school type, ethnic makeup, and size. 68

Each high school’s largest feeder school was recruited when available, which resulted in a sample of more than 130 schools. More than 90,000 adolescents completed an in-school survey between 1994 and 1995. Roughly 20,000 adolescents completed the first in-home interview, which gathered detailed information on respondents’ sexual behavior and relationship scripts. Nearly 15,000 respondents completed the wave 2 interview, which took place approximately 1 year after the first in-home interview.

Sample

The Add Health research team attempted to interview every respondent from 16 schools (14 small and 2 large) as part of the first two in-home interviews. As population samples were collected from these schools, this “saturated” sample contains information on a wide variety of attributes (including romantic relationship scripts) for every friend.

Accordingly, I restrict my analysis to this saturated sample (N = 3,702).8 I exclude one of the saturated schools (individual N = 45) because its low response rate precludes the measurement of social network characteristics. I exclude an additional 959 respondents who were lost to follow up or were not interviewed at wave 2.9 Finally, I exclude 1,637 respondents who did not form a new romantic relationship between interview waves, did not provide the ordering of events within their ideal scripts or first subsequent romantic

8Prior research that has utilized the saturated sample suggests it is generally comparable to the full Add Health sample with regards to respondents’ personal characteristics (Haynie 2002). One notable difference between the samples is that small schools are over-represented in the saturated sample.

9Those who were in the 12th grade at the time of the wave 1 interview were not interviewed at wave 2. 69

relationship, or were missing survey weights. My sample includes 1,013 respondents nested in 15 schools.

Dependent Variable: Relationship Inauthenticity

During the wave 1 interview, respondents indicated whether they (and a hypothetical romantic partner) would not experience a number of events within an “ideal romantic relationship” within the next year. These items include the following 17 events:

1) We would go out together in a group; 2) I would meet my partner’s parents; 3) I would tell other people that we were a couple; 4) I would see less of my other friends so I could spend more time with my partner; 5) We would go out together alone; 6) We would hold hands; 7) I would give my partner a present; 8) My partner would give me a present; 9) I would tell my partner that I loved him/her; 10) My partner would tell me that he/she loved me; 11) We would think of ourselves as a couple; 12) We would talk about contraception or sexually transmitted diseases; 13) We would kiss; 14) We would touch each other under our clothing; 15) We would have sex; 16) My partner or I would get pregnant; and 17) We would get married. Respondents were then asked to provide the ideal sequencing of the events that would take place within the hypothetical relationship.

I use a subsample of these items (see below) to construct each respondent’s ideal romantic relationship script.

At wave 2, respondents reported on details of specific romantic relationships that took place within the 18 months prior to the date of the wave 2 interview. Respondents identified relationship partners in two ways. First, respondents could name up to three

individuals with whom they had “special romantic relationship.” Respondents who did not have a “special romantic relationship” were asked whether they 1) held hands, 2) kissed on the mouth, or 3) told another person that they liked or loved another person.

Respondents who engaged in all these three activities with at least one person were asked to identify up to 3 “liked” relationship partners. I use information on both special and liked romantic relationships in this analysis.10

Respondents indicated the ordering of a similar set of events that actually occurred within up to 3 romantic/liked relationships. Minor differences are in wave 2, items 7, 8, 9, 10, and 17 from the ideal relationship script were dropped from the actual relationship script, and respondents indicated whether the respondent and his/her partner

“gave each other presents,” and “told each other you loved each other,” and “touched each other’s genitals (private parts).” Following Harding (2007), I make the ideal and actual relationship sequences comparable by dropping genital touching from the actual script and combining items 7 (gave present) and 8 (received present), and items 9 (said I love you) and 10 (told I love you) in the ideal script into single items representing gift exchange and expressing love, respectively. The ordering of the combined items is based on the first of the two items that occurred (e.g., if giving a gift was the 4th item and

10I focus on both special and liked relationships in this analysis for two primary reasons. First, every completed “liked” relationship once had the potential to develop into a “special” romantic relationship. Second, a relationship that was once thought of as special romantic relationship may later be interpreted as a “liked” relationship, especially if the relationship was rocky, ended poorly, or was especially inauthentic. As such, excluding liked romantic relationships may bias my results towards the null by excluding relationships that are retrospectively not identified as special romantic relationships.

receiving a gift was the 8th item, then gift exchange was coded as the 4th item). In total, each ideal and actual romantic sequence may consist of up to 14 events.

I quantify the difference in the content and sequencing of events in respondents’

“ideal” relationship script and the first liked/special romantic relationship that was initiated after the date of the wave 1 interview to measure relationship inauthenticity. I argue this difference captures an important dimension of romantic relationship inauthenticity, namely incongruence between ideal ordering of events in a relationship and one’s actions within a romantic relationship. I use optimal matching (OM), a particular form of sequence analysis (Abbott and Tsay 2000), to quantify the minimal

“cost” of transforming (through substitutions, insertions, and deletions of events) the actual script into the ideal romantic relationship script for each respondent. More information on sequence analysis procedures is provided in the Analytic Strategy section.

Key Independent Variables

Script Discordance. As part of the first in-home interview, respondents from the

“saturated” schools were asked to nominate up to five female and five male friends from the school roster.11 I use these nominations to identify school-based friendships and measure script discordance. In this study, I use both sent and received nominations to identify friends. After identifying each respondent’s set of friends, I match each friend

11Due to a survey implementation error, 28 of the respondents in the final sample were only asked to nominate one best male and one best female friend during the wave 1 in home interview. In order to keep these respondents in the sample I replaced missing non- best friend nominations with nominations from the in-school interview. As a robustness check I ran models with these respondents excluded. Results were nearly identical to those presented in this chapter. 72

with his or her ideal relationship script, and use OM to quantify the minimum cost of transforming the respondent’s ideal script and into each friend’s ideal script. Script discordance represents the average minimum cost of transforming one’s script into the script of every other friend. This value is undefined for isolates, or individuals with no sent or received ties. In order to avoid dropping isolates, I recode script discordance to the mean value of script discordance for isolates and include a binary variable indicting isolate status (0=no, 1=yes) in linear regression models of relationship inauthenticity.12

Friend Involvement. Respondents also indicated whether they engaged in a number of activities with each nominated friend outside of school during the week leading up to the wave 1 interview. I focus on three activities, including visiting a friend’s house, hanging out with a friend after school, and hanging out with a friend during the weekend to measure friend involvement. I first made each network tie non- directed such that a tie exists between i and j if i nominated j, or vice versa. I then created a weighted tie between the respondent and his or her network partner that ranged from 0, which indicates neither i nor j engaged in any activity in the past week, to 3, which indicates that i and j engaged in 3 activities with one another in the past week. Following

Haynie and Osgood (2005), I summed the values of the weighted ties across each respondent’s send and receive network and divided the resulting values by the square root of each respondent’s number of friends. To aid in the interpretation of the interaction

12I ran additional models that were similar to those presented in this chapter, but for which isolates were dropped. Results from those analyses were nearly identical to those presented in this chapter.

between friend involvement and script discordance, I center involvement around its grand mean. Isolates’ friend involvement was recoded to the mean value of involvement.

Friends’ Popularity. I measure friends’ popularity by first determining whether each respondent is highly popular. In this study, respondents are highly popular if they are in the 90th percentile of in terms of the number of received friendship nominations

(0=no, 1=yes).13 I match this measure to each send and receive network partner and calculate the proportion of friends who are highly popular. This measure ranges from 0, indicating none of the respondent’s friends are highly popular, to 1, indicating all of the respondent’s friends are highly popular. Finally, I take the square root of this measure to reduce skewness and center the measure around its grand mean. Values of isolates’ friends’ popularity were recoded to the mean value of friends’ popularity.

Partner characteristics

Respondent/partner social overlap. At wave 2, respondents answered a series of questions regarding their relationships with romantic partners prior to the romantic relationship onset. I use responses to a number of questions to measure relationship characteristics that may impact relationship inauthenticity. Research suggests friends’ approval of romantic partners increases the longevity of romantic relationships (Felmlee

13I avoid using the raw number of friend nominations because of the high level of variation in the sizes of the student populations across schools. Students in smaller schools have fewer possible alters to nominate than students from large schools. Accordingly, students from small schools are more likely than students in large schools to be attached to frequently nominated peers simply because of the restricted friend pool. Indeed, the school mean popularity score is negatively correlated with school size (r = - .368). Creating within-school measures of popularity takes school size into account when measuring friends’ popularity. 74

2001). I include a measure of shared mutual friends with a binary variable indicating whether the respondent and the relationship partner shared friends prior to relationship onset (0=no, 1=yes) to account for this potentially confounding factor.14

Overlap in the friendship networks of romantic partners may increase intimacy and interdependency among couples (Sprecher and Felmlee 2000). Accordingly, I include a measure of friend/partner network overlap, which indicates the proportion of the respondent’s friends who knew the partner prior to the onset of the relationship (as indicated by the respondent). I recoded initial responses, which ranged from 1 (“All of them”) to 5 (“None of them”) such that higher values indicate more network overlap.

Being friends prior to romantic relationship onset may also promote intimacy, reciprocity, and companionate love in romantic relationships (Furman and Simon 1999). I thus include a measure of friends prior to relationship, which is binary and indicates whether the respondent and the partner were friends prior to the onset of the relationship

(0=no, 1=yes).

Relationship characteristics. I control for a number of relationship characteristics that may influence inauthenticity. I include a binary measure that indicates whether the relationship is ongoing (0=no, 1=yes). I also include a binary variable that indicates whether the relationship in question is a “special romantic” relationship (0=no, 1=yes,

“liked” relationship is the reference category). Long intervals between the first interview and relationship onset may affect the measurement of relationship inauthenticity, as

14Unfortunately, Add Health did not directly measure friends’ approval of the respondents’ partner. However I argue that friends would most likely approve of the partner as a person if they are initially friends with the partner. 75

longer intervals potentially allow individuals to change their ideal romantic relationship script. The measure of relationship inauthenticity may thus be confounded with changes in ideal scripts, especially among respondents whose relationships started long after the first interview. Accordingly, I control for time to relationship which represents the number of days that have elapsed between the first in-home interview and relationship onset, divided by 10. All variables capturing relationship characteristics were measured at wave 2.

Partner demographics. I control for partners’ demographic characteristics

(measured at wave 2) that may influence romantic relationship inauthenticity. First, age dissimilarity has been associated with adverse outcomes and may be linked to poor relationship quality (Loftus, Kelly, and Mustillo 2011). Following Loftus and colleagues

(2011), I include binary measures of younger partner, which equals 1 if the relationship partner was at least three years younger than the respondent (0 if otherwise), and older partner, which equals 1 if the partner was at least three year older than the respondent (0 if otherwise). Similar age serves as the reference category. Finally, I include a binary variable that indicates whether the respondent and partner are the same race/ethnicity

(0=no, 1=yes).

Control Variables

Peer network characteristics. I construct three variables that capture respondents’ structural positions within the school networks (measured at the wave 1) that may be associated with relationship inauthenticity. First, popularity may decrease relationship

inauthenticity by signaling social desirability (thus increasing power within romantic relationships) by increasing social proximity to compatible romantic partners. I measure popularity with a count of the number of nominations received from other schoolmates.

Friendliness may also increase access to compatible partners through high involvement with peers. I measure friendliness with the number of nominations that the respondent sent to other students in the school (McCarthy and Casey 2008). Finally isolate is a binary variable indicating whether the respondent has no sent or received friendship nominations (0=no, 1=yes).

Psychological characteristics. I control for psychological characteristics

(measured at wave 1) that are associated with romantic relationship dynamics and inauthenticity. Relationship inauthenticity is negatively associated with self-esteem among girls (Impett et al. 2008). Following Longmore and colleagues (2004), I measure self-esteem with a 5-item scale that taps respondents’ agreement with statements such as

“you have a lot to be proud of” and “you like yourself just the way you are.”15 Responses ranged from 1 (“strongly disagree”) to 5 (“strongly disagree”). Self-esteem represents the mean of reverse-coded items (α=.838).

Depression has been linked to relationship inauthenticity (Tolman and Porche

2000). Following Perreira and colleagues (2005), I measure depression with five items adopted from the CES-D scale (Radloff 1977) that indicate prevalence of emotional and

15I drop one item from Longmore and colleagues (2004) self-esteem measure that indicates whether the respondent feels he/she is doing everything about right, as I argue this item is more indicative of self-centeredness and is used in my measure of low self- control.

mental health problems (e.g., felt depressed, sad) throughout the past week (α=.793).

Responses ranged from 0 (“never or rarely”) to 3 (“all or most of the time”). Values presented are the means of the items.

Gottfredson and Hirschi (1990) suggest low self-control leads to unstable and strained relationships. Following Perrone and colleagues (2004) I measure low self- control with a scale that includes 5 items that tap various dimensions of self-control. The first item measures respondents’ agreement with the statement: “you feel like you are doing everything just about right.” Responses ranged from 1 (“strongly disagree”) to 5

(“strongly agree”). The next three items assess the frequency in which respondents had trouble paying attention, had difficulties finishing their homework, and had problems with their teachers during last school year (1=“never/rarely” to 5=“every day”). A final item indicates how often respondents had trouble keeping their mind focused during the past week. Responses ranged from 0 (“never”) to 3 (“everyday”). Low self-control consists of the mean of the standardized items (α=.755).

Low body image is positively associated with relationship inauthenticity among girls (Impett et al. 2008). I use responses to the question, “How do you think of yourself in terms of weight?” to measure body image. Initial responses ranged from 1 (“very underweight”) to 5 (“very overweight”). Following Vogt Yuan (2010) I construct two dummy variables that indicate overweight body image (very to slightly overweight; 0=no,

1=yes) and underweight body image (slightly to very underweight; 0=no, 1=yes). About the right weight serves as the reference category.

Religiosity. Religiosity is strongly associated with sexual behavior and romantic relationship outcomes (Bearman and Brückner 2001). I control for religiosity with 4-item scale (measured at wave 1) that captures the importance of religion for the respondent (1 being “very important” to 4 being “not important at all”) and the frequency of prayer, religious service attendance, and religious youth group participation (1 being “once a week or more” to 4 being “never”). I recoded items such that higher values indicate greater religiosity and took the mean of the standardized items (α=.837).

Individual demographics. Because physical development is associated with adolescent sexual behavior (Halpern, Kaestle, and Hallfors 2007), I measure physical maturity with the question “how advanced is your physical development compared to other boys/girls your age?” (measured at wave 1). Responses ranged from 1 (“I look younger than most”) to 5 (“I look older than most”). I also control for age, gender, race

(binary indicators for black, Latino and other race/ethnicity; white as reference), single- parent household (0=no, 1=yes), parent socioeconomic status (consisting of the standardized mean of parents’ highest occupational status and education level). All variables capturing individual demographic characteristics were measured at wave 1.

Finally, especially short or lengthy scripts may be discordant with friends’ scripts through large numbers of insertions or deletions. Accordingly, I control for ideal script length, which represents the number of events within the respondent’s ideal romantic relationship script.

Analytic Strategy

Optimal matching procedures. I use optimal matching (OM) to measure relationship inauthenticity and script discordance. OM quantifies differences between two data sequences according to the insertions, deletions, and substitutions that are required to transform one script into the other. The OM algorithm determines the minimal total

“cost” of transforming one sequence into another. Following Harding (2007), I empirically define substitution costs by estimating a substitution cost matrix (which has

14 rows and columns, see Table C.1.), the elements of which represent the logged inverse probability of transitioning from one event to every other script event. The cost matrix is based on the ideal relationship scripts from all Add Health respondents who participated in the wave 1 interview. Elements of the cost matrix are smaller (and less costly) if items frequently follow/precede one another (say, going out together alone then holding hands) and larger (and more costly) if the items rarely follow/precede one another (say holding hands then having sex). Following Stovel and colleagues (1996), I set the insertion/deletion cost to the largest value of the substitution cost matrix. I also normalize the difference between two scripts by dividing by the length of the longer script, which makes the final difference measure represent the average cost per event in the longer script (Harding 2007). Finally, in order to take into account differences in the ability to progress through ideal scripts in ongoing versus completed relationships, I truncate ideal relationships to be equal in length to the actual relationship in terms of the number of events for respondents in ongoing relationships. Sequence analysis procedures were performed with the TraMineR package (Gabadinho et al. 2011) for the R statistical analysis software program.

I use optimal matching to measure both relationship inauthenticity and script discordance. To measure relationship inauthenticity, I ran an optimal matching procedure that quantifies the difference between the ordering of events in a particular respondent’s ideal script and actual romantic relationship. This process was repeated for each respondent in the sample. The resulting values comprise my measure of relationship inauthenticity. I then ran an optimal matching procedure that measures the minimum costs of changing a respondent’s ideal relationship script to the script of each of his or her friend(s). Upon running the OM procedure, symmetric N x N matrices with 0’s on the diagonal resulted, where N equals the number of friends in the send and receive network plus the respondent. The elements of these matrices represent the costs of changing respondent i’s script into every other friend’s script. In addition, the first row/column of the matrix represents the difference between the focal respondent’s ideal script and his or her friends’ scripts. To measure script discordance, I summed of the first row of the matrix and divided that number by N - 1. The resulting measure represents the average minimum cost of transforming respondent i’s script into the scripts of his or her friend(s).

Finally, because I interact script discordance with peer network measures, I center the variable at its grand mean.

Correcting for selection into new romantic relationships. I model relationship inauthenticity with a two-equation estimation procedure based on Heckman (1976) which accounts for self-selection into a new romantic relationship after wave 1. Following

McCarthy and Casey (2008) I first estimated a selection model which predicted the probability of entering a new romantic relationship with probit regression. The selection

model included a number of covariates that are associated with romantic relationships among adolescent relationships in past research but are only associated with the substantive outcome through the selection process. Next, I calculated an inverse Mills ratio by dividing the probability density function by the cumulative distribution function from the probit model. The resulting measure, which captures the hazard of non-selection into romantic relationships, was used as a predictor in my substantive models consisting of linear regressions of relationship inauthenticity.

My selection model incorporated a number of the aforementioned control variables, including age, race, gender, religiosity, socioeconomic status, depression, low self-control, self-esteem, physical maturity, popularity, and friendliness. I also included a number of other variables that may impact the likelihood of forming a new relationship after the wave 1 interview, including parental attachment, parental approval of sex, grade point average, body mass index (BMI), ongoing relationship, prior romantic relationship, desire for romantic relationship, abstinence pledge status, and prior sexual intercourse. Information on the construction of these additional selection variables is provided in Table C.2. Finally, I control for unmeasured school characteristics that may influence selection into romantic relationships with school fixed effects.

Survey Weights and Missing Data. For my substantive models, missing values on independent variables were imputed with the ICE command in the Stata12 statistical software program (Royston 2004). I use the mi svy command suite in Stata12 to estimate linear regression models that take into account unequal probability of selection and school clustering when estimating models with imputed data.

Modeling strategy. My analysis proceeds as follows. I first display the results from the probit model used to predict the formation of a new romantic relationship. Next,

I present regression models of relationship inauthenticity, controlling for selection and other individual and partner variables. Finally, I present sensitivity analyses that help rule out competing explanations regarding the association between script discordance and relationship authenticity.

RESULTS

Selection model

Table C.4. displays the coefficients and robust standard errors from a probit model used to measure the hazard of forming a new romantic relationship between interview waves. Results indicate boys and those in ongoing relationships at wave 1 are less likely to form a new relationship between waves. Conversely, respondents who have had a recent romantic relationship and would like to have a romantic relationship have higher likelihoods of forming a new relationship. Grade point average and parental attachment have negative and marginally-significant associations with the likelihood of forming a romantic relationship, while low self-control and physical development have positive and marginally significant associations with the outcome. Finally, popularity is positively associated with the outcome and body mass index is negatively associated with the likelihood of starting a new relationship.

Script discordance, friend involvement, and relationship inauthenticity

I now turn to the primary models of interest that test the association between script discordance and romantic relationship inauthenticity. I first estimate a baseline model (Model 1) that controls for age, gender, race, single parent household, socioeconomic status, religiosity, self-esteem, depression, self-control, body image, physical maturity, popularity, friendliness, isolate status, hazard of relationship, ideal script length, and time to relationship. In Model 2, I introduce partner/partnership characteristics including romantic relationship, ongoing relationship, older partner, younger partner, same race, mutual friends, friend/partner network overlap, and prior friends. I test hypothesis 1 in Model 3 by introducing script discordance. Results from

Models 1 through 3 are displayed in Table C.5.

Turing to results from Model 1 (Table C.5.), age has a negative and marginally- significant association with relationship inauthenticity (b=-.044, p<.10). This finding reflects past research that demonstrates inauthenticity tends to decline as adolescents progress into emerging adulthood. Results also indicate compared to whites, members of other race/ethnic groups have higher levels of relationship inauthenticity (b=.289, p<.05).

Time to relationship is also positively associated with relationship inauthenticity (b=.011, p<.05). Finally, relationships that are ongoing at wave 2 are associated with lower levels of relationship inauthenticity than relationships that have ended (b=-.366, p<.001). The negative association between ongoing relationships and the outcome is perhaps expected, given adolescents may be more likely to terminate inauthentic relationships than more authentic relationships. No other associations from Model 1 are statistically significant.

In Model 2 (Table C.5.) I introduce variables capturing partner/partnership characteristics. None of the measures introduced in Model 2 are significantly associated with the outcome. In addition, introducing partner/partnership characteristics has little impact on the significant associations observed in Model 2. I test hypothesis 1 in Model 3

(Table C.5.), which predicts a positive association between script discordance and relationship inauthenticity. In support of the hypothesis, script discordance has a strong positive association with relationship inauthenticity after controlling for characteristics of individuals, partners, and relationships (b=.287, p<.01). Results indicate a one standard deviation increase in script discordance is associated with a .159 standard deviation increase in relationship inauthenticity. With the inclusion of script discordance the magnitude of the positive coefficient for ideal script length becomes statistically significant (b=.063, p<.05) and the negative association between self-esteem and the outcome becomes marginally-significant (b=-.090, p<.10).

Table C.6. displays results from models in which I test hypotheses 2 and 3. In

Model 4, I introduce friend involvement and friends’ popularity and an interaction term between script discordance and friend involvement. In Model 5, I omit the friend involvement*script discordance interaction and introduce friends’ popularity its interaction with script discordance. Finally in Model 6, I reintroduce the friend involvement*script discordance interaction term. All models in Table C.6. include the control variables from Model 3 in Table C.5. For the sake of presentation, I omit the coefficients for control variables from the table.16

16 All omitted coefficients are available from the author upon request.

Turning to the results for Models 4 through 6 (Table C.6.), both the partial effects of script discordance (b=.306, p<.01) and friend involvement (b=.052, p<.01) are positive and significant. Because script discordance and friend involvement are centered at their grand means, the coefficient for script discordance is interpreted as its association with relationship inauthenticity when peer involvement is at its mean and vice versa. While the coefficients for discordance and involvement are positive and significant, the non- significant interaction coefficient in Model 4 suggests friend involvement does not accentuate the association between script discordance and relationship inauthenticity.

Model 4 fails to support hypothesis 2.

In Model 5 (Table C.6.) I omit the script discordance*friend involvement interaction and introduce the interaction between friends’ popularity and script discordance. Again the coefficient for friend involvement is interpreted as its association with relationship inauthenticity when script discordance is at its mean. The coefficient for friends’ popularity fails to reach statistical significance. However in support of hypothesis 3, the interaction coefficient is positive and significant (b=.543, p<.05), indicating friends’ popularity accentuates the association between script discordance and romantic relationship inauthenticity. Finally, in Model 6, I reintroduce the interaction between script discordance and friend involvement. Doing so slightly increases the magnitude of the coefficient for the interaction between script discordance and friends’ popularity (b=.660, p<.05), however the interaction coefficient for script discordance and friend involvement remains non-significant.

Figure D.2. graphically displays the predicted values of relationship inauthenticity across different levels of script discordance for adolescents with low (1.5 standard deviations below the mean), versus high (1.5 standard deviations above the mean) levels of friends’ popularity. This figure is based on the results from Model 6 in Table C.6., with all other measures held at their grand means. As the figure illustrates, there is a negligible association between script discordance and relationship inauthenticity among adolescents who are attached to less popular friends. Conversely, we observe a much sharper increase in relationship inauthenticity as script discordance increases among adolescents whose friendship groups consist of higher proportions of highly-popular friends.

Sensitivity Analysis

I ran two additional models in order to help rule out competing explanations for the results from Model 3 in Table C.5. One especially plausible explanation is adolescents may experience more relationship inauthenticity if their partners’ ideal scripts are discordant with their own ideal script. This is because adolescents may experience more difficulty in achieving their ideal relationships if their scripts are incongruent with the scripts of romantic partners than if the scripts were similar. If this process occurs, script discordance would be most strongly associated with relationship inauthenticity among adolescents who become romantically involved with friends and individuals within whom they have high levels of social overlap.

In order to help ensure the results from previous models reflect a process rooted in script discordance with friends (versus romantic partners), I interact script discordance with three measures that capture friendship and network overlap prior to relationship onset, namely respondent/partner network overlap, mutual friends, and prior friends.

Positive coefficients on these interaction terms would provide evidence that the results may be due to this alternative mechanism. This model includes all of the variables and interaction terms from Model 6 in Table C.6.

Results for this analysis are displayed in Model 7 in Table C.7. For the sake of presentation coefficients for control variables are omitted from the table. Results indicate that none of the interaction effects introduced in Model 7 reach statistical significance.

Results from Model 7 support the notion that script discordance influences relationship inauthenticity thorough peer reinforcement processes.

Another alternative explanation is discordant scripts lead adolescents to enact romantic relationships that more closely reflect the scripts of their friends. In such a case, the association between script discordance and relationship authenticity would be explained by adolescents relying on ideal relationship scripts of friends to guide their behavior within ideal romantic relationships. If this process occurs, then one would expect a negative association between script discordance and a measure that captures the extent to which individuals’ actual romantic relationship diverge from the ideal romantic relationship scripts of their friends. To help rule out this competing explanation, I created a measure of behavioral discordance, which quantifies the average minimum cost of transforming one’s actual romantic relationship script to the ideal romantic relationship

scripts of one’s friends using OM. I then regressed behavioral discordance on script discordance and the control variables from Model 3 in Table C.6. Results from this final model are presented in Model 8 in Table C.7.

Results from the model indicate script discordance is positively but not significantly associated with behavioral discordance (b = .141, n.s.), suggesting script discordance among friends does not lead adolescents to form romances that are more congruent with the ideal ordering of events among their friends. Together Models 7 and 8 provide further support that congruence in scripts with one’s friends promotes authentic relationships through a social reinforcement process.

DISCUSSION AND CONCLUSION

Psychologists are increasingly focusing on the personal consequences of relationship inauthenticity. Research focusing on the causes of relationship authenticity has also largely focused on its individual precursors. However, as sociological research suggests, cultural features of adolescent contexts may lead to relationship inauthenticity within romantic relationships (Harding 2007; Pascoe 2007). Building on insights from cultural sociology and differential association/social learning perspectives, this study focused on the association between script discordance within one’s friendship groups and social network processes to understand why certain adolescents are more able to enact their ideal romantic relationship than others.

I introduced the concept of script discordance, which captures the extent to which the content and sequencing of events within one’s ideal romantic relationship script

diverges from the ideal scripts of one’s friends. Using optimal matching and social network data, I operationalized script discordance and measured its association with romantic relationship inauthenticity among a subsample of respondents from Add Health.

I found script discordance has a positive and significant association with inauthenticity within subsequent romantic relationships. I also found friends’ overall popularity accentuates the association between script discordance and the outcome.

This study opens up a number of potential avenues of research regarding of the association between culture and romantic relationship dynamics. First, I demonstrate the role script discordance within friendship groups plays in shaping adolescent romantic relationships. Individuals may be less likely to enact their ideal romantic relationships when their scripts are not reinforced by others within their friendship groups. Research focusing on other forms of cultural discordance (e.g., cultural frames and worldviews) may shed more insight into why individuals fail to achieve their ideal relationships or more generally behave in a manner that is inconsistent with their own attitudes and ideal strategies of action.

In addition, script discordance in adolescence may shape romantic relationships that are formed in emerging adulthood and beyond. Life course research focusing on the consequences of exposure to script discordance in adolescence may advance the understanding of how childhood friendships shape subsequent relationships. More broadly, individuals may behave in a manner that is inconsistent with their ideal strategies of action when they are exposed to discordant scripts within their friendship groups. Future research that focuses on discordance with regards to other scripts and

strategies of action (e.g., ideal education trajectories, how one finds a job) may shed more light onto the processes through which individuals achieve (or fail to achieve) important life objectives.

This study also contributes to cognitive approaches to culture by identifying how the structure of relations among friends alters the association between script discordance and romantic relationship inauthenticity. Results from this study reflect findings from recent studies focusing on the interactive effects of network structure and cultural processes on adolescent sexual behavior (Soller and Haynie 2012a, 2012b). In particular, script discordance is most strongly associated with relationship inauthenticity among individuals who are embedded in high status peer groups. This has important implications for the longitudinal exposure to discordance scripts. Individuals may be less inclined to end friendships with culturally-incompatible friends if the friends in questions are of high status. Beyond friendship dyads, triadic relations (e.g., sharing a third friend) or dense friendship groups may also discourage adolescents from ending discordant relationships.

Faced with such structural constraints, individuals may adjust their ideal relationship scripts to more closely match the scripts of their close associates. Applying Stochastic

Actor-Based models (e.g., SIENA) that take into account these social network processes may help understand the coevolution of scripts and friendship networks.

While my study advances the understanding of the role friendships play in shaping adolescent romantic relationship inauthenticity, it is not without limitations.

First, the initial wave of Add Health was conducted in the mid-1990s. Subsequent large- scale studies that build on Add Health’s scope and design are necessary to determine

whether friends continue to have as significant of an impact on romantic relationships in the 21st century. With that said, Add Health remains the sole data source that includes detailed data on both romantic relationship scripts and adolescent friendship networks among adolescents nested in a number of schools. Second, ideal romantic relationship scripts were measured at the time of the wave 1 interview. Ideally, the script would have been measured at the time of relationship onset. Exposure to discordant scripts may have led respondents to subsequently change their ideal script to be more compatible with group members. If this is the case, then some of the results may have been attributable to scripts changing, rather than relationship inauthenticity. However, a supplemental model based on Model 3 from Table C.5 (results not displayed but available upon request) that included an interaction between script discordance and time to relationship onset failed to reach statistical significance. This suggests that the time span between script measurement and relationship onset (which provides greater opportunity to change one’s script), does not alter the association between script discordance and relationship inauthenticity. This provides further support for my claim that script discordance among friends is a crucial factor in shaping romantic relationship inauthenticity.

Despite these limitations, this study contributes to the understanding of the interactive association between cultural and network processes and adolescent romantic relationship dynamics. I hope that future research builds on my findings to better understand the causes of inauthentic relationships in adolescence and beyond.

Chapter 4: Caught in a Bad Romance: Adolescent Romantic Relationships and Mental Health

Romantic involvement is a hallmark of adolescence (Giordano, Manning, et al.

2006). Unfortunately, few studies examine the health and developmental consequences of early romantic relationships. Instead, research on youth romance is in large part overshadowed by work focusing on early sexual behavior. Nevertheless, findings from a limited number of studies attest to the significance of early romantic relationships for adolescent health and well-being (Collins et al. 2009; Giordano 2003).

Relationship inauthenticity—the extent of incongruence between thoughts/feelings and actions within relational contexts—is a key mechanism through which personal relationships influence adolescent well-being (Impett et al. 2008).

Relationship demands, coupled with the desire to gain approval, may suppress authenticity and promote behavior that reflects what adolescents perceive relationship partners wish to observe (Harter 1999). Compromised authenticity is in turn linked to depression among adolescents (Impett et al. 2008). Despite evidence pointing to effects of inauthentic peer relationships on adolescent well-being, little sociological research focuses on the mental health consequences of inauthenticity within youths’ romantic relationships.

This study explores the mental health consequences of adolescent romantic relationship inauthenticity. Integrating insights from sociological theories of culture

(Swidler 1986, 2001) and identity (Burke and Stets 2009; Cast and Burke 2002), I suggest romantic relationship inauthenticity compromises mental health by disrupting the process of self-verification. Self-verification occurs when meanings within a situation match or confirm meanings of a particular identity (Cast and Burke 2002). As Swidler

(2001:87) notes, individuals develop strategies of action based on actual and ideal self- concepts. Relationship scripts—ideal progressions of actions and emotional states within romantic relationships (e.g., holding hands, kissing, etc.)—are meaningful components of identity that reflect idealized romantic selves. Enacting meaningful ideal scripts within relationships verifies components of the self and enhances mental health. Conversely, deviating from idealized romantic relationship scripts blocks self-verification processes and thus compromises mental well-being (Cast and Burke 2002). In this study I test whether relationship inauthenticity—conceptualized as behavior that diverges from the sequencing of events within one’s idealized relationship script (Harding 2007)—is associated with multiple dimensions of adolescent mental health.

I also test whether the association between relationship inauthenticity and adolescent mental health varies by gender. While some suggest romantic involvement equally enhances boys’ and girls’ well-being (La Greca and Harrison 2005), past research suggests romantic dynamics have stronger effects on girls’ mental health (Joyner and

Udry 2000). Gender variation in the association between romantic relationship outcomes and mental health may be attributable to the increased salience of interpersonal

This study uses data from the first two waves of the National Longitudinal Study of Adolescent Health (hereafter, Add Health). Using sequence analysis (Abbott and Tsay

2000), I measure romantic relationship inauthenticity by quantifying the extent to which the ordering of events within respondents’ ideal romantic relationships script diverges from the sequencing of events within their first subsequent romantic relationship

(Harding 2007). I then test associations between relationship inauthenticity and multiple dimensions of adolescent mental health (e.g., severe depression, suicide ideation etc.).

This study underscores the importance of gender and inauthenticity in determining how romantic involvement influences adolescent mental health.

BACKGROUND

Adolescent Romance and Mental Health

The transition into adolescence brings increased risk of mood disorders and emotional problems (Costello et al. 2006). Social and physiological changes contribute to the increased risk of poor mental health throughout this life-stage. “Storm and stress” perspectives (Arnett 1999) note adolescents encounter new situations that compromise their well-being. In particular, relational stressors—i.e., strains rooted in personal relationships—often trigger adolescent mental health problems (Exner-Cortens,

Eckenrode, and Rothman 2013; Joyner and Udry 2000; Meier 2007).

At the same time, adolescence brings increased interested in romantic relationships for most youth. Early romances may contribute to mood disorders among adolescents for a number of reasons. For instance, romantic involvement often disrupts friend and parent-child relationships (Joyner and Udry 2000). Dating violence—which occurs within significant portions of adolescent romantic relationships—increase youths’ risk of depression and suicide ideation (Exner-Cortens et al. 2013). Rejection and breakups are also often emotionally-taxing experiences (Joyner and Udry 2000). Romantic involvement also increases the likelihood of sexual activity. While the association between sexual behavior and mental health depends on several factors (e.g., age, gender, romantic attachment) (Meier 2007; Spriggs and Halpern 2008), sexual intercourse is on average positively associated with adolescent depression (Hallfors et al. 2005). Given the stressors rooted in early romances, it is perhaps not surprising that romantic involvement increases the risk of depression among adolescents (Joyner and Udry 2000).

Most research on adolescent romance and mental health focuses on more notable events, such as breakups, sexual activity, and partner aggression. While these momentous events are important to consider, more commonplace features of early romantic relationships are also consequential for adolescent well-being. For instance, relationship quality most likely determines how romantic relationships affect delinquency (Giordano et al. 2010; McCarthy and Casey 2008). Recently, Simons and Barr (forthcoming) found no average association between romantic involvement and offending. Conversely, love and commitment within romantic relationships protected against youthful offending.

Importantly, cognitive processes (e.g., non-hostile views of relationships) mediate the

association between relationship quality and delinquency. Although these studies focus on externalizing symptoms and behavior, these results suggest the effect of romantic involvement on adolescent mental and behavioral health is contingent upon more general relationship processes and characteristics.

A more holistic account of adolescent romances may help identify additional mechanisms through which early relationships influence adolescent mental health. For instance, Collins (2003) argues behavior that adheres to personal relationship ideals enhances positive emotions. Conversely, relationships that depart from idealized notions represent a source of negative emotions (Stets and Burke 2005). As I argue below, romantic relationship inauthenticity—conceptualized as the extent to which the ordering of events within relationships deviate from ideal relationship scripts—may also adversely affect mental health by impeding self-verification.

Ideal Scripts, Ideal Selves

In her seminal article on culture and action, Ann Swidler (1986) argues culture is best understood as a “tool-kit” of worldviews, habits, skills, and styles that individuals draw upon when formulating strategies of action. Strategies of action are persistent ways of ordering of behavior that allow groups and individuals to achieve desired ends or solve problems. Individuals formulate cultural repertoires—which contain varieties of symbols and strategies—through interactions with others and within institutional contexts (e.g., religion, media, schools etc.). Culture thus provides individuals with strategies for

managing the social world, thereby influencing behavior by regulating the strategies of action that are available to pursue.

In subsequent work, Swidler (2001) elaborates the relationship between strategies of action and identity. She proposes culture operates in large part by attaching meanings to the self—a process that informs which strategies of action are conceptualized as more- or-less possible, promising, or worthy of enacting, given one’s self-conception (Swidler

2001:81). In essence, strategies of action allow individuals to pursue ends that reflect and enhance key components of salient identities. Individuals in turn formulate and enact strategies of action to become a certain kind of idealized self. Culture relates to identity processes in part because individuals construct and pursue strategies that are in concert with meanings attached to the self. In this sense, romantic strategies of action help individuals constitute ideal selves within romantic relationships.

Cultural “scripts” represent templates for sequencing behavior over time. Scripts are akin to Swidler’s concept of strategies of action in that they both specify how to achieve desired outcomes or solve problems (Harding 2007). Beyond guiding behavior within intimate settings, romantic relationship scripts are crucial in developing ideal selves related to romantic and intimate partnerships. For instance, in his study of inner- city black adolescents, Anderson (1989) found boys develop and enact scripts that reflect the esteemed “player” identity. Boys cultivate “game”—a component of one’s cultural repertoire that comprises various habits, styles, and scripts that constitute the player identity. Enacting sexualized relationship scripts ultimately verifies the player identity.

Anderson’s respondents describe one particular script as taking a girl on a “walk through the woods” (Anderson 1989:62). This script entails a series of actions intended to demonstrate one’s status as an upstanding young man to a girl (e.g., attending church, visiting with the girl’s family, etc.). The ultimate objective of the script is to have sexual relations, which affirms the boy’s status as a “player,” thus verifying the identity. At the same time, relationship scripts provide the cultural material that helps constitute one’s game. As Anderson illustrates, relationship scripts represent important components of boys’ masculine identities. At the same time, an important part of the player script is geared towards enhancing a girl’s positive emotions by appealing to her desires to enact conventional romantic relationships.

Swidler and Anderson highlight the significance of culture in identity formation.

However, they do little to addresses the mental health consequences of enacting relationships that diverge from ideal scripts. Sociological theories of identity (Burke and

Stets 2009; Cast and Burke 2002) explicitly attend to the association between identity and mental health. By integrating insights from identity perspectives and Swidler’s work, I elaborate how inauthentic relationships disrupt the self-verification process and contribute to poor mental health.

Identity, Inauthenticity, and Mental Health

Burke (1991:837) defines an identity as a “set of ‘meanings’ applied to the self in a social role or situation defining what it means to be who one is” (emphasis in original).

According to this perspective, the self is composed of a hierarchical set of identities that

reflect various roles (e.g., parent, romantic partner) and social structural positions (Burke and Stets 2009). Identities are hierarchically-organized in that they vary in salience or importance within individuals. Adolescents’ relative inexperience with romantic relationships means the romantic partner role is novel. Because novel roles are more salient for individuals (McGuire and McGuire 1982), romantic roles are likely particularly important components of adolescent identities (Meier 2007).

Identity standards—defined as “meanings that are assigned to the self in the performance of a given role” (Andersson 2012:292)—provide reference points with which to guide action throughout the self-verification process. Individuals continuously regulate behavior in order to adhere to identity standards (Andersson 2012; Burke and

Stets 2009). Verification of romantic-related identities requires that behavior be consistent with established meanings and definitions associated with partner roles—e.g., emotionally supporting one’s partner, doing house/yard work, behaving in traditionally masculine or feminine manner, etc.

I argue cultural scripts are important components of identity standards that allows for self-verification over the course of a relationship. While verification is continuous, mostly unconscious, and routinized, cultural scripts—regardless of length—enable romantic partners to affirm key components of salient identities. For instance, enacting sexual scripts allows one to express components of identity standards regarding romantic, sexual, or gender roles (Cornwell and Laumann 2011; Kornrich, Brines, and Leupp

2013). In this way, romantic relationship scripts encode sequences of actions that are

100

consistent with the kind of persons they wish to be, thus allowing individuals with a means to verify ideal selves in relationships.

There are emotional consequences for mismatches between meaningful actions and identity standards. Self-verification enhances feelings of individual self-worth and personal mastery, thus promoting mental well-being (Cast and Burke 2002). Conversely, failing to affirm identity standards leads to emotional distress and contributes to mood disorders such as depression (Andersson 2012; Cast and Burke 2002).

Based on the above discussion, relationship authenticity likely enhances mental health because it entails reflecting salient identity standards, thereby verifying one’s identity as a romantic partner. Conversely, failure to match behavior to identity standards inhibits self-verification processes, thereby damaging one’s sense of personal mastery

(Thoits 2012). Thus deviating from ideal sequences—or engaging in romantic relationship inauthenticity—harms mental health to the extent that is entails failure to match actions to romantic identity standards (Cast and Burke 2002). Accordingly I hypothesize:

H.1. Romantic relationship inauthenticity is positively associated with poor mental health.

As identities are hierarchically nested within individuals, verifying salient identities may have especially important consequences for mental health. As research on adolescent relationships suggests romantic relationships are particularly important for girls’ identities, relationship inauthenticity may have especially strong effects on their mental health.

101

Gender and the Cost of Inauthentic Romances

Sociological perspectives suggest romantic relationships are particularly important for girls’ identity formation and self-concepts (Rosenfield and Mouzon 2013;

Simon and Barrett 2010). According to these perspectives, gender socialization leads girls’ to develop senses of self that are more intimately bound to their interpersonal relationships. For instance, Cross and Madson (1997) propose higher rates of assigning childcare and other domestic tasks to girls increase their sense of nurturance and relatedness from an early age. Conversely, boys are more inclined to engage in activities and tasks outside of the home that allow more freedom and independence. In the end girls are socialized to more fully consider their relationships and pursue harmony with others, while boys are socialized to be more independent and distinguish themselves from others

(Markus and Kitayama 1991).

The importance of interpersonal relationships for girls’ developing sense of self is revealed in adolescent research. As Impett et al. (2006) note, psychological research focusing on girls’ desire to maintain relationships suggests the importance of relational process in girls’ identities. Qualitative sociological studies further suggest the importance of romantic relationships in girls’ self-concepts. For instance, Eder et al.’s (1995) middle school ethnography revealed the importance of personal appearance and being in love for girls’ developing sense of self. Similarly, Pascoe’s (2007) high school ethnography suggests a girl’s status within the school peer hierarchy is in large part determined by the status of the boy with whom she is romantically involved. While these studies do not explicitly focus on identity formation, they do suggest romantic relationships are more

102

salient components of girls identities when compared to boys. In turn, the extent of behavioral divergence from romantic identity standards—as captured by romantic relationship inauthenticity—may have especially important consequences for girls’ mental health. This is because verifying romantic roles through authentic behavior becomes increasingly tied to mental health as such roles become more salient for an individual’s self-concept. This leads to the hypothesis:

H.2. The association between romantic relationship inauthenticity and poor mental health is particularly strong among girls.

DATA AND METHODS

This study uses data from Add Health, which includes a longitudinal nationally- representative survey focusing the social context of adolescents’ health and development in the United States. Respondents were nested within randomly-selected schools drawn from a clustered random sample of 80 high schools stratified by ethnic composition, population size, public/private status, geographic region, and urbanicity. High schools including an eleventh grade and at least 30 enrollees were eligible for participation. Each school’s largest feeder school was recruited when available.

Respondents in the present study completed a wave 1 in-home interview in 1995 and a wave 2 interview that took place roughly 12 months after wave 1. Nearly 15,000 respondents participated in both waves. Of these respondents, 6,173 formed a new romantic relationship between waves. I exclude 550 of those respondents who did not provide information on ideal romantic scripts or actual romantic relationships. I also

103

exclude respondents with missing data on dependent variables (N=12) or survey weights

(N=295). My final sample includes 5,316 respondents (2,905 girls and 2,411 boys).

Measures

Dependent variables. I analyze four dependent variables, all measured at wave 2:

(1) severe depression, (2) high somatic symptoms, (3) suicide ideation, and (4) suicide attempt.

Severe depression is measured with the CES-D (Radloff 1977). Following

Perreria et al. (2005), the measure is based on a subset of 5 items from the original 20- items.17 Respondents indicated the frequency with which they experienced symptoms including having the blues, feeling life was not worth living, feeling depressed, feeling sad, and feeling happy throughout the week leading up to wave 2. Ordinal responses ranged from 0 (“never or rarely”) to 3 (“most of the time/all of the time”). To measure severe depression, I summed the items, with responses for “felt happy” reverse-coded to indicate higher depression (α=.779). Scores ranged from 0 to 15. I then identified those with severe depression by setting cut-points for the summed scale. Research using the full

CES-D (which has a possible total score of 60) determined cut-points for identifying major depressive disorders among adolescents to be 22 (a ratio of .367/1) for boys and 24

(a ratio of .4/1) for girls (Roberts et al. 1991). Following research using reduced CES-D scales (Warren et al. 2010), I lowered the cut-point for severe depression to 6 for girls

17Perreria et al.’s (2005) abbreviated CES-D scale demonstrates improved psychometric properties across race/ethnic groups compared to the full scale. As an empirical check, I ran analysis based on another depression measure that includes 19 of the original items. Results from those models were nearly identical to those presented in this chapter. 104

and boys (a ratio of .4/1). My binary measure of severe depression, indicates whether the respondent’s depressive symptomology score was greater than or equal to the cut-point

(0=no, 1=yes).

Somatic symptoms are measured using 12 items tapping the frequency of physical conditions during the past 12 months (Walsemann, Bell, and Maitra 2011). Conditions include: 1) headache; 2) feeling hot all over suddenly, for no reason; 3) stomachache or upset stomach; 4) cold sweats; 5) feeling physically weak, for no reason; 6) feeling really sick; 7) waking up feeling tired; 8) dizziness; 9) chest pains; 10) aches, pains, or soreness in muscles or joints; 11) trouble falling asleep or staying asleep; and 12) trouble relaxing.

Ordinal responses ranged from 0 (“never) to 4 (“everyday”). To measure high somatic symptoms I first summed the items (α=.799).18 Following Walsemann et al.’s (2011) approach to measuring high somatic symptoms with Add Health data, youth are classified as experiencing high somatic symptoms if their score was in the 75th percentile across all

Add Health respondents, which in this study was 12 (1=yes).

The third outcome is suicide ideation. At wave 2, respondents were asked

“During the past 12 months, did you ever seriously think about committing suicide?”

Responses were binary, with 1 indicating “yes.” My measure of suicide ideation indicates

18Unfortunately, somatic symptoms were only measured in reference to the past year. Accordingly, relationship inauthenticity might have followed somatic symptoms. To assess the potential for reverse causality, I regressed a continuous measure of relationship inauthenticity on prior somatic symptoms (measured at wave 1; not displayed). A positive coefficient for somatic symptoms in these models would suggest it predicts subsequent relationship inauthenticity. Prior somatic symptoms was not associated with relationship inauthenticity for boys or girls. While this does not rule out the possibility of reverse causality, it provides support this study’s theoretical model.

105

whether the respondent contemplated suicide in the past year (1=yes).19 The final mental health outcome is suicide attempt. Those experiencing suicide ideation were asked

“During the past 12 months, how many times did you actually attempt suicide?”

Responses ranged from 0 (“0 times”) to 4 (“6 or more times”). I measure suicide attempts by collapsing responses to this question into binary categories, with 1 indicating “one or more suicide attempt” and 0 indicating “no attempts.” Respondents who did not experience suicide ideation were recoded to 0.20

Romantic relationship inauthenticity. I measure romantic relationship inauthenticity with information on respondents’ ideal romantic relationship sequences and events that took place within respondents’ first subsequent romantic relationship. At wave 1, respondents indicated whether they would experience 17 events in an “ideal romantic relationship” at this stage in their lives. Events include: 1) We would go out together in a group; 2) I would meet my partner’s parents; 3) I would tell others we were a couple; 4) I would see less of my friends to spend more time with my partner; 5) We would go out alone; 6) We would hold hands; 7) I would give my partner a present; 8)

19Generally, respondents in my study had worse mental health compared to the entire sample. My sample only includes respondents who formed a new romantic relationship between waves. Thus differences in mental health between the reduced sample and the entire Add Health sample are perhaps expected as romantic involvement is associated with higher depression among some adolescents (Joyner and Udry 2000).

20Respondents were not asked the dates of suicide ideation or suicide attempts. I am thus unable to determine whether suicide ideation/attempts followed experiences with relationship inauthenticity. As with somatic symptoms, I regressed a continuous measure of relationship inauthenticity on suicide ideation and suicide attempt prior to wave 1 (not displayed). These variables were not associated with relationship inauthenticity and the associations did not vary by gender, suggesting neither suicide attempt nor ideation predict later relationship inauthenticity.

106

My partner would give me a present; 9) I would tell my partner I loved him/her; 10) My partner would tell me he/she loved me; 11) We would think of ourselves as a couple; 12)

We would discuss contraception or sexually transmitted diseases; 13) We would kiss; 14)

We would touch each other under our clothing; 15) We would have sex; 16) I/My partner would get pregnant; and 17) We would get married. Respondents then provided the sequence in which chosen events would ideally unfold. I use a subset of these items (see below) to measure respondents’ ideal romantic relationship scripts.

At wave 2, respondents reported on actual events within specific romantic relationships that took place during the last 18 months. Relationship partners were identified in two ways. First, respondents identified up to three individuals with whom they had “special romantic relationship.” Those who did not report having a special romantic relationship were asked whether they 1) held hands, 2) kissed, or 3) told another person that they liked/ loved him/her after wave 1. Those engaging in all three activities identified up to 3 “liked” relationship partners. For this analysis, I use both special and liked romantic relationships to measure romantic relationship inauthenticity.21

Respondents provided the actual ordering of events within each relationship.

Possible events were identical to those from the ideal script with a few exceptions. For actual relationship scripts, ideal script items 7, 8, 9, 10, and 17 were excluded.

21I focus on both relationship types for two reasons. First, every “liked” relationship has/once had the potential to become a “special” relationship. Second, relationships once thought of as “special” romantic relationships may be reinterpreted as “liked” relationships, especially for rocky and inauthentic relationships. Excluding “liked” romantic relationships may thus bias my results towards the null. That said, the vast majority of relationships in this study are special relationships.

107

Additionally, respondents indicated whether they and their partner told each other you loved each other, exchanged presents, and touched each other’s genitals/private parts. I drop genital touching from the actual script and combine items 7 (gave present) and 8

(received present), and items 9 (said I love you) and 10 (told I love you) from the ideal script into single items representing gift exchange and expressing love (respectively) in order to make the ideal and actual relationship sequences fully comparable.22 In total, ideal and actual romantic scripts consist of up to 14 events.

I quantify differences in the sequencing of ideal relationships and subsequent romantic relationships to measure relationship inauthenticity. This difference captures an important dimension of relationship inauthenticity, namely discrepancies between how relationships ideally and actually unfold. Using sequence analysis (Abbott and Tsay

2000), I quantify the minimal “cost” of transforming respondents’ actual scripts into the ideal romantic relationship scripts through inserting, deleting, and substituting events.

The resulting values capture the extent of relationship inauthenticity, with higher values indicating greater inauthenticity. In order to test for non-linearity in the association between relationship inauthenticity and mental health, I break the continuous measure of relationship inauthenticity into four quartiles. The first quartile of romantic relationship inauthenticity serves as the reference category for all models in this study. I describe sequence analysis procedures in more detail in the analytic strategy section below.

22The ordering of combined items is based on the order of the first of the two combined events. For instance, if “saying I love you” was event 3 and being told “I love you” event 6, then “professed love to each other” was coded event 3. 108

Control variables. I control for a number of demographic characteristics and individual/relationship confounders. All models include lagged dependent variables

(measured at wave 1). Descriptions of control measures, including items on which they are based, construction procedures, and scale reliabilities (when appropriate), are provided in Table E.1. Descriptive statistics for variables used to predict mental health outcomes, as well as dependent variables and relationship inauthenticity, are displayed in

Table E.2.

Analytic Strategy

Optimal matching. I measure relationship inauthenticity using Optimal Matching

(OM), a particular form of sequence analysis. OM quantifies differences between two data sequences according to the substitutions, deletions, and insertions required to transform the sequences to be equivalent. The OM algorithm alters sequences based on the minimal total “cost” of transforming scripts into another. Substitution costs are empirically defined by estimating a 14x14 substitution cost matrix (see Table C.1.).

Matrix elements represent the logged inverse probability of transitioning from event i to event j. The cost matrix is estimated according to transition probabilities derived from ideal romantic relationship scripts among all Add Health respondents participating in wave 1. Substitutions are larger and more costly if items rarely follow/precede one another (e.g., holding hands then having sexual intercourse). Conversely, substitutions are smaller and less costly if items frequently follow/precede one another (e.g., going out

109

together alone then holding hands). Insertions/deletions are set to the largest substitution cost (Harding 2007).

The OM procedure used to measure relationship inauthenticity quantifies the difference in the ordering of events for a respondent’s actual and ideal relationship scripts. I normalize total costs by dividing them by the length of the longer script, which makes the difference represent the average cost per event in the longer script (Harding

2007). To take into account differences in the opportunity to progress through ideal scripts for completed versus ongoing relationships, I truncate ideal relationships to be equal in length to the actual relationship (in terms of the number of events) for those in ongoing relationships. This process is repeated for each respondent. Sequence analyses are performed with the TraMineR package for the R statistical analysis software program

(Gabadinho et al. 2011).

Selection into new romantic relationships. I model the association between relationship inauthenticity and mental health using a two-equation estimation procedure based on Heckman (1976). This procedure accounts for self-selection into new romantic relationships between waves. Following McCarthy and Casey (2008), I first run a probit regression model that estimates the probability of entering a new relationship between waves. This model includes covariates that are associated with the likelihood of forming romantic relationships among adolescents in past research but are only theoretically associated with mental health through selection processes.23 After estimating the selection

23Descriptions of variables included in the selection model are provided in Table E.1. Descriptive statistics for variables included in the selection model but not included in models of mental health are not displayed but are available upon request. 110

model, I divided the probability density function by the cumulative distribution function to calculate an inverse Mills ratio. The resulting values capture the hazard of non- selection into my sample and are used as a predictor in regressions of adolescent mental health. Selection model results are displayed in Table E.2.

Survey weights and missing data. For models of adolescent mental health, missing values on independent variables are imputed with the ICE command in the Stata12 statistical software program (Royston 2004). I estimate weighted regression models of imputed data that adjust for the unequal probability of selection and school clustering with Stata’s mi svy command suite (Chantala and Tabor 2010).

Modeling strategy. I estimate a series of logistic regressions to test the association between relationship inauthenticity and the four binary measures of mental health. I estimate models separately by gender because I hypothesize the association between romantic relationship inauthenticity and mental health varies by gender. Wald tests confirm this is the appropriate modeling strategy (Frisco, Houle, and Martin 2010).

For a particular measure of mental health, I first test the association between romantic relationship inauthenticity and the outcome, adjusting for individual characteristics. I then introduce potential relationship confounders (e.g., physical/psychological aggression) in a subsequent model. Results for girls’ severe depression and high somatic symptoms are displayed in Table E.3., while models of girls’ suicide ideation and suicide attempt are presented in Table E.4. I present results for boys’ severe depression and high somatic symptoms in Table E.5. and results for boys’ suicide ideation and suicide attempts in Table E.6. Tables E.3. through E.6. display

111

unstandardized coefficients and standard errors. For the sake of presentation, I omit coefficients and standard errors for individual control variables from Tables E.3. through

E.6. (omitted coefficients are available upon request).

RESULTS

Relationship Inauthenticity and Girls’ Mental Health

Table E.4. examines the association between relationship inauthenticity and depressive symptoms (Models 1 and 2) and somatic symptoms (Models 3 and 4) among girls. Model 1 indicates that compared to those in the lowest quartile of relationship inauthenticity, girls in the third and fourth quartiles have higher log-odds of severe depression after controlling individual characteristics. Upon introducing relationship characteristics in Model 2, the difference in log-odds of severe depression among those in the third quartile slightly decreases in magnitude, while the coefficient for those in the fourth quartile slightly increases in magnitude. According to results from Model 2, a girl in the third quartile of relationship inauthenticity has 1.62 times the odds of severe depression and a girl in the fourth quartile has 2.08 times the odds of severe depression compared to a girl in the first quartile of relationship inauthenticity,

Models 3 and 4 (Table E.4.) test the association between relationship inauthenticity and high somatic symptoms among girls. Turing to Model 3, I find a positive and significant difference in the log-odds of high somatic symptoms among girls in the fourth quartile of relationship inauthenticity compared to girls in the first quartile.

Upon introducing relationship characteristics in Model 4, the magnitude of the coefficient

112

for the fourth quartile of relationship inauthenticity slightly increases. Based on estimates from Model 4, girls in the highest quartile of relationship inauthenticity have 1.49 times the odds of experiencing high somatic symptoms compared to girls in the lowest quartile.

Table E.5. examines the association between relationship inauthenticity and girls’ suicide ideation (Models 1 and 2) and suicide attempt (Models 3 and 4). In addition to individual characteristics, Models 1 through 4 control for prior severe depression and binary variables indicating whether 1) a friend/family member attempted suicide and 2) a friend/family member committed suicide (both prior to wave 1). Focusing on girls’ suicide ideation in Model 1, girls in both the third and fourth quartiles of relationship inauthenticity have higher log-odds of suicide ideation compared to girls in the lowest quartile. Upon introducing relationship characteristics in Model 2, the coefficient representing the difference between the third and first quartiles of relationship inauthenticity decreases to a non-significant level. The positive coefficient for the fourth quartile of relationship inauthenticity also decreases in magnitude, though it remains significant. Based on results from Model 2, girls in the highest quartile of relationship inauthenticity have 1.64 times the odds of experiencing suicide ideation compared to girls in the lowest quartile.

Models 3 and 4 (Table E.5.) present results for girls’ suicide attempt. First, I find girls in the third quartile of relationship inauthenticity have higher odds of suicide attempt compared to girls in the first quartile. The coefficient for the fourth quartile of relationship inauthenticity is positive although not significant, suggesting relationship inauthenticity has a non-linear association with the log-odds of suicide attempt. Upon

113

introducing relationship characteristics in Model 4, I find the difference in the log-odds of suicide attempt among girls in the first and third quartiles of relationship inauthenticity decreases but remains significant. Based on results from Model 4, girls in the third quartile of relationship inauthenticity have 2.23 times the odds of suicide attempt compared to those in the first quartile.

Relationship Inauthenticity and Boys’ Mental Health

I now examine the association between romantic relationship inauthenticity and boys’ mental health. I present results for boys’ severe depression and high somatic symptoms in Table E.6. Models of boys’ suicide ideation and suicide attempt are displayed in Table E.7.

Turing to results for severe depression (Table E.6.), Model 1 indicates there are no significant differences in the log-odds of severe depression across romantic relationship inauthenticity quartiles for boys. The null associations between romantic relationship inauthenticity quartiles and boys’ severe depression remain non-significant upon introducing romantic relationship characteristics in Model 2.

Models 3 and 4 (Table E.6.) test the association between relationship inauthenticity and boys’ high somatic symptoms. As with severe depression, there are no significant differences in the log-odds of high somatic symptoms across relationship inauthenticity quartiles among male respondents in Model 3. These null associations remain non-significant in Model 4 after controlling relationship characteristics.

114

Models in Table E.7. test associations between relationship inauthenticity and boys’ suicide ideation and suicide attempt. All models in Table E.7. also include all relevant individual controls as well as variables indicating whether a friend/family member committed suicide or attempted suicide. Turing to results for boys’ suicide ideation, I find no significant differences in the log-odds of suicide ideation among boys in the second, third, or fourth quartiles of romantic relationship inauthenticity when compared to boys in the first quartile in Model 1. After introducing relationship characteristics in Model 2, all coefficients for the romantic relationship inauthenticity quartiles remain non-significant. Models 3 and 4 (Table E.7.) examine the association between romantic relationship inauthenticity and boys’ suicide attempt. Again, I find no significant differences in the log-odds of suicide ideation across quartiles of romantic relationship inauthenticity among boys in Model 3. All coefficients representing the differences in boys’ log-odds of suicide attempt across relationship inauthenticity quartiles remain non-significant in Model 4 after introducing relationship characteristics.

DISCUSSION AND CONCLUSION

This study contributes to the understanding of gender, romantic relationships, and adolescent mental health. Previous research highlights the mental health consequences of early romantic involvement. However, most prior work centers on more noteworthy events, such as sexual activity, partner aggression, and break-ups. Integrating theories of culture and identity, I examine the mental health consequences of adolescent romantic relationship inauthenticity—conceptualized as the extent to which the ordering

115

relationship events diverge from the idealized relationship scripts. This study demonstrates romantic relationship inauthenticity is tied to important dimensions of adolescent psychological well-being, however its association with mental health varies by gender.

Empirically, I find relationship inauthenticity is only associated with girls’ mental health. Specifically, girls who experience relationships that strongly diverge from their ideal scripts are at increased risk of severe depression, high somatic symptoms, and suicide ideation. In addition, girls in the third quartile of relationship inauthenticity are at increased risk of suicide attempt compared to girls in the first quartile. Conversely, I find no evidence that romantic relationship inauthenticity contributes to poor mental health among boys. These findings reflect past work that suggests girls’ are at greater risk for the adverse effects of romantic involvement than their male counterparts.

More generally, results point to the importance of more general features of relationships in determining how romantic involvement impacts mental health. While I do not intend to downplay the significance of partner aggression and early and/or unwanted sexual activity for adolescent health, my study suggests a more holistic approach to adolescent romantic relationships—one attuned to culture and relationship progressions—provides novel insights into how romantic involvement potentially harms adolescent mental health.

I also integrate sociological theories of culture and identity to specify how romantic relationship inauthenticity influences mental health. Drawing from Swidler

(1986) I suggest romantic relationship inauthenticity is usefully conceptualized as the

116

extent to which behavior within romantic relationships diverges from how relationships ideally would unfold. Noting Swidler’s (2001) work on culture and identity, I further argue romantic scripts are important components of adolescents’ emerging identities.

Increasing romantic involvement throughout adolescence, coupled with the novelty of romantic roles for youth, suggests romantic relationships are central to several youths’ lives. Thus verifying romantic roles becomes increasingly linked to mental health during adolescence. The amplified importance of interpersonal relationships for girls may mean verification of romantic roles is intimately tied to their psychological well-being.

I also contribute to the sociology of identity by explicating the role of cultural scripts in identity verification processes. Cultural scripts pertaining to romantic and sexual behavior detail meaningful strategies of action that, when enacted, facilitate self- verification process (Cornwell and Laumann 2011; Kornrich et al. 2013). For better or worse, cultural scripts present individuals with means to constitute and verify important components of self. For instance, some “player” scripts entail behavior that verifies esteemed identities but increases the risk of poor health (e.g., concurrent and casual sexual partnering). In addition, such scripts involve “playing” female partners and most likely involve some level of romantic relationship inauthenticity on their part. Self- verification through certain relationship scripts may lead to adverse outcomes for both partners. Future research focusing on the emergence of different scripts across cultural contexts may provide more insight into the link between culture and identity. At the same time, research focusing on the social precursors of relationship inauthenticity will

117

increase the understanding of why individuals fail to “stick to the script” within romantic relationships.

As with all studies I must address a number of study limitations. For boys, externalizing behaviors (e.g., aggression, substance use, etc.) may be more common reactions to disrupted self-verification processes (Rosenfield and Mouzon 2013). Thus null associations between romantic relationship inauthenticity and boys’ mental health found in this study may thus reflect gendered reactions to relationship inauthenticity. In order to rule out this explanation, I ran parallel models of delinquency (violent and non- violent forms) and numerous substance use behaviors (results not displayed).

Relationship inauthenticity was not associated with these outcomes for boys. Conversely, relationship inauthenticity was positively associated with cigarette smoking among girls.

These supplementary models further attest to heightened salience of relationship inauthenticity for girls’ mental and behavioral health.

I also should note adolescence is a period of identity formation and exploration

(Erikson 1968). Accordingly, idealized romantic relationships evolve and romantic role salience fluctuates within individuals throughout adolescence. Unfortunately, I was unable to measure respondents’ ideal relationship scripts immediately prior to, or during, the first subsequent romantic relationship. As a result I could not assess the extent to which respondents’ scripts changed between wave 1 and relationship onset. My measure of relationship inauthenticity may thus fail to fully capture discordance between ideas and relationship events for respondents who changed scripts prior to relationship onset.

However I ran supplementary models that interacted relationship inauthenticity with the

118

length of time between wave 1 and relationship onset. Negative coefficients on interaction parameters from these models would indicate relationship inauthenticity matters more for individuals whose relationships started sooner after wave 1 than later.

This in turn would suggest respondents changed their scripts between the time of the first interview and relationship onset. Across these supplementary models, the association between relationship inauthenticity and mental health variables did not vary by this time span. Thus, it appears individual differences in the time lapse between wave 1 and relationship onset did not impact my results.

Another limitation is that I cannot control for important factors that may confound the association between romantic relationship inauthenticity and mental health. For instance, while partner aggression and sexual activity were controlled, I could not assess partner infidelity. Infidelity and inauthenticity could share common causes and confound one another’s association with mental health. Future research capturing more detailed relationship information may help assess the relative effects of romantic relationship inauthenticity on adolescent mental health. Additionally, prospective accounts capturing adolescents’ mental well-being as romances unfold may advance the understanding of why girls are more adversely affected by relationship authenticity than boys.

Despite these limitations, my study contributes to the understanding of gender, culture, identity, romantic involvement, and adolescent mental health. Romantic relationships are particularly relevant to adolescent mental health for a number of reasons. One important, yet often overlooked aspect of youth romances is their role in self-verification. Adolescents use scripts to guide actions within various contexts.

119

Perhaps as importantly, youth utilize scripts throughout the self-verification process. Self- verification is in turn central to psychological well-being. My study suggests enacting cultural scripts within romantic relationships is vital to identity verification and youths’ subsequent mental health. In the end, enacting romantic scripts may be more closely related to mental health than may be commonly assumed.

120

Chapter 5: Conclusion

The transition from late childhood to adolescence brings increased interest and involvement in sexual and romantic relationships for most youth. However it is only in recent years that social scientists have begun to focus much attention on the developmental significance of early romances. In addition, research on adolescent sexual activity tends to be narrow in focus, with most studies conceptualizing sexual activity primarily in terms of its entailed risks (e.g., early pregnancy, STIs, etc.). The limited amount of social scientific research on romantic relationships, coupled with the restricted focus on research on adolescent sexual activity, has led to an incomplete understanding of how romantic relationships and sexual activity influences youths’ health and development. Additionally, few have attempted to identify how certain features of adolescent social contexts influence dynamics of early romantic relationships and determine the impact of early romantic and sexual involvement on adolescent well-being.

The three studies that comprise this dissertation aimed to advance the understanding of the social dynamics of early relationships and sexual activity among youth in the United States. Grounding my analysis in sociological theories of culture

(Harding 2007; Kirk and Papachristos 2011; Swidler 1986, 2001), I demonstrate how cultural features of adolescent environments shape how adolescents’ behavioral approach to romantic relationships, and how conceptions of romantic relationships and sexual

121

behavior potentially contribute to adolescent mental health. I also aimed to enrich the theoretical understanding of culture and action by integrating insights from identity theory (Burke and Stets 2009; Cast and Burke 2002), social network perspectives

(Haynie 2001), social learning theory (Akers 2009), and gender perspectives (Kornrich et al. 2013; Rosenfield and Mouzon 2013) into my analyses. Incorporating these complementary perspectives helped me identify important contingencies in the link between culture and action, and highlight the conditions under which romantic and sexual involvement most often lead to poor mental health. Below I summarize key findings from the three analytical chapters and note key contributions of my dissertation with regards to the understanding of adolescent romantic relationships and sexual activity, the link between culture and action, and policies aimed at promoting adolescent health and well- being. I conclude by addressing limitations of my dissertation and noting how future research that builds on this collection of studies may advance the understanding of adolescent romantic relationships and sexual behavior.

SUMMARY OF RESULTS

In Chapter 2 I examine the mental health consequences of sexual intercourse among adolescents. Motivated in part by recent trends in “slut shaming”—publicly degrading girls on the basis of actual or perceived sexual behavior—within high schools,

I focus on school sexual double standards as key modifiers in the relationship between sexual intercourse and mental health. This study is rooted in a framing approach to culture, which proposes individuals rely on collective, interpretive schema that simplify

122

and condense the social world by punctuating and encoding objects, situations, events, experiences, within social environments (Snow and Benford 1992). Further, this perspective suggests frames help individuals share understandings of the social and personal consequences of actions (Kirk and Papachristos 2011).

I suggest the sexual double standard in part functions as a school-based cultural frame that helps adolescents identify the personal consequences of sexual activity.

Encoded in this frame is the notion that sexual activity is socially-rewarding for boys, and potentially leads to stigma among girls. Integrating insights from identity theory, I suggest sexual double standards pertaining to the social benefits of sexual intercourse may lead actors to codify sexual intercourse as integral to esteemed male and degraded female gender roles. This lead to the hypothesis that sexual intercourse leads to poor mental health among girls and enriches boys’ mental health as the sexual double standard becomes more salient within a school.

Results largely support this hypothesis. I find compared to girls who abstained from sexual intercourse between interview waves, girls who engaged in sexual intercourse are more likely to experience severe depression. While this was true for both intercourse categories, the slope of non-romantic sexual intercourse was slightly steeper compared to girls who only had sexual intercourse in a romantic relationship. I found no variation in the associations between the sexual intercourse categories boys’ severe depression across different levels of the sexual double standard. However I did find compared to boys who abstained from sexual intercourse between study waves, boys who had sexual intercourse in a non-romantic relationship have higher likelihoods of high

123

self-esteem as the sexual double standard in a school becomes more salient. There were no differences in the odds of high self-esteem across different levels of the sexual double standard for girls in different sexual intercourse categories.

Chapter 2 examines the peer context of romantic relationship inauthenticity.

Drawing from work by Swidler (1986) and Harding (2007), I suggest romantic relationship inauthenticity in part refers to the extent to which the sequencing of actions and events within a romantic relationship diverges from the sequencing of events within one’s ideal relationship script. This chapter focuses on the peer context of relationship inauthenticity, and examines whether discordant scripts within one’s friendship group leads to increased inauthenticity in subsequent relationships. In this study, script discordance refers to the extent to which the ordering of events within one’s ideal romantic relationship script diverges from the sequencing of events within one’s close friends’ romantic scripts. One key finding is that adolescents experience more relationship inauthenticity as one’s ideal script becomes more discordant with one’s peers’ scripts.

I also draw from differential association and social learning theory to formulate hypotheses that predict peer-group social network processes alter the association between script discordance and relationship inauthenticity. I find friends’ popularity (but not involvement with friends) accentuates the association between script discordance and relationship inauthenticity.

The study presented in Chapter 3 highlights the utility of incorporating both cultural and social network processes into the study of peer influence to better understand

124

the cultural context of romantic relationships. I also demonstrate how script discordance influences adolescent romantic relationships through reinforcement processes, but also note an important contingency in the link between cultural features of one’s peer group and one’s actions. Findings from this study help explain why certain adolescents behave in manners that are inconsistent with their own ideal strategies of action.

The final analytical chapter (Chapter 4) examines the mental health consequences of relationship inauthenticity. Integrating insights from theories of culture (Swidler 1986,

2001) and identity (Burke and Stets 2009; Cast and Burke 2002), I propose relationship inauthenticity compromises mental health because it disrupts self-verification processes.

This is because individuals formulate ideal relationship scripts that are based on ideal self-concepts (Swidler 2001). Enacting ideal relationship scripts verifies important components of the self and enhances mental health. On the other hand, deviating from ideal relationship scripts hinders self-verification, thus compromising mental well-being.

Drawing from gender perspectives (Cross and Madson 1997; Rosenfield and

Mouzon 2013), I hypothesized the association between romantic relationship inauthenticity and mental health varies by gender. This is because gender socialization processes potentially leads to an increased salience of interpersonal relationships for girls’ self-concepts (Rosenfield and Mouzon 2013). Accordingly, the inner workings of romantic relationships may have especially strong effects on girls’ mental health when compared to boys.

Results indicate relationship inauthenticity is only associated with multiple dimensions of mental health for girls but not boys. Specifically, I find girls who

125

experience relationships that strongly diverge from their ideal relationship scripts have higher risks of severe depression, high somatic symptoms, and suicide ideation. Girls in the third quartile of relationship inauthenticity were also at increased risk of suicide attempt (compared to girls in the first quartile). Findings from Chapter 4 reflect past work that suggests girls’ have a greater risk for the adverse effects of romantic involvement than boys.

CONTRIBUTIONS OF DISSERTATION

This dissertation contributes to the sociological understanding of adolescent romantic relationships and sexual behavior. More generally, this dissertation advances theories of culture by identifying important contingencies in the link between culture and action and demonstrating the importance of identity with regards to the operation of culture. Finally, as this research illustrates the importance of sexual behavior and romantic relationships for adolescent health and well-being, this dissertation has a number of implications for programs and policies aimed at promoting adolescent well- being.

Understanding Culture, Sexual Behavior, and Romantic Relationships

One of the more noteworthy findings from Chapter 2 is that sexual intercourse (in either romantic or non-romantic relationships) was not on average associated with severe depression or high self-esteem for girls. However, sexual intercourse appears to have more severe effects on girls’ depression as the sexual double standard in a school

126

increases. Together these results suggest sexual activity has a negligible impact on the psychological well-being of girls who are embedded schools in which the sexual double standard is less severe. This finding is important because little research considers how cultural features of adolescent contexts shape the consequences of adolescent sexual behavior. The findings presented in Chapter 2 suggest sexual activity does not exert universal effects on adolescent well-being. Rather the effect of sex on mental health depends cultural structures within the local environment.

Little research also focuses on the social context of adolescent romantic relationship inauthenticity. Findings from Chapter 3 highlight both cultural and network structural features are important in determining the extent to which individuals relationships reflect their ideal romantic relationships. This finding mirrors recent work that suggests both network structure and cultural features of peer environments matter for adolescent sexual behavior (Soller and Haynie 2012b). More generally, Chapter 3 contributes to cultural sociology by illustrating the importance of cultural reinforcement in shaping the extent to which one’s ideal scripts inform one’s subsequent actions.

Chapter 4 advances the understanding of the mental health consequences of early romantic involvement. A number of studies suggest romantic involvement is developmentally significant but the effects of early romances on the well-being of youth depend on a variety of factors (Collins 2003; Giordano 2003; Joyner and Udry 2000). By integrating insights from complementary theories of gender, culture and identity, Chapter

4 elaborates the importance of enacting relationship scripts for girls’ mental health. More

127

generally, this study suggests enacting ideal scripts that are integral to salient identities may be more central to psychological health than may be commonly assumed.

Policy Implications

This dissertation has a number of policy implications. As I did at the end of

Chapter 2, I caution against interpreting the study on the sexual double standard as identifying conditions under which early sexual activity entails neutral or positive effects on adolescent mental health. Rather, I aimed to elaborate the mechanisms through which gendered cultural structures determine how sexual behavior affects adolescent mental health. As suggested by the study, the effects of adolescent sexual activity on mental health are contingent upon the extent of the sexual double standard in a school. That said one of the major justifications for abstinence only programs is that early sexual activity harms mental health. However, as illustrated by Chapter 2, early sexual activity may only harm mental health among girls who are embedded in contexts with gender-biased cultural structures. Accordingly, one method for combatting the adverse effects of sexual activity and promoting health sexuality development is to address sexual double standards in school contexts. In the end, such programs may promote healthy sexuality development and yield benefits throughout the life course.

Chapter 3 highlights the social precursors of romantic relationship inauthenticity, while Chapter 4 suggests relationship inauthenticity adversely affects girls’ mental health. In the later chapter, I suggest relationship inauthenticity is more strongly associated with girls’ mental health because of the salience of romantic relationships for

128

their identity formation. Apart from efforts aimed at reducing romantic relationship inauthenticity, girls’ mental health may be enhanced by programs that help them develop identities that are independent of their romantic involvement, such as those related to sports or other extracurricular activities. Such an effort may lead girls to develop more independent self-concepts that can be enhanced and verified through processes that do not depend on romantic involvement.

LIMITATIONS AND FUTURE DIRECTIONS

This dissertation has a number of limitations that must be considered in future research on adolescent romantic relationships and sexual activity. First, the first wave of the Add Health study was conducted in 1994. As a result it remains uncertain whether the associations observed in this dissertation would be found nearly 20 years later. Although

I have no reason to suspect the findings presented here are period specific, future research that gathers detailed information on romantic relationships, peer networks, and the cultural climate of school contexts may help verify the key findings of this dissertation.

With that said, Add Health remains the sole data source that includes both complete network data as well as detailed information on romantic and sexual relationships from a nationally-representative sample of youth. As a result, Add Health remains the best data source to test the theoretical model outlined in this dissertation. Future studies aiming to build on the scope and breadth of Add Health would do well to gather information on romantic and sexual behavior and cultural climates as they are most likely remain key factors in adolescent well-being and development.

129

Another key limitation of this dissertation is that I am only able to focus on heterosexual romantic relationships and sexual activity. Because Add Health lacks comprehensive measurements of same-sex romantic relationships and sexual activity I was forced to restrict my analysis to opposite sex relationships. As a result I am uncertain whether the associations observed in the present study apply to gay, lesbian, bisexual, or transgendered youth. The dynamics of non-heterosexual relationships are sociologically compelling given political debates regarding same-sex marriage and parenting in recent years. Future research examining the relationship dynamics of early non-heterosexual relationships may provide further insight into the formation and inner-workings of such relationships in adulthood.

Another limitation of this dissertation relates to the clustered sample of Add

Health. While Add Health’s school-based sample enables an unprecedented opportunity to study school effects on youth development, adolescents are embedded within a number of important contexts (e.g., family, neighborhoods, non-school youth groups) that expose adolescents to cultural frames and relationship scripts (Harding 2007, 2010).

Unfortunately Add Health’s design does not allow the measurement of cultural scripts and frames from within most non-school environments. Furthermore, the study samples from this dissertation are restricted to adolescents who were in school. Accordingly, I am unable to make inferences regarding the consequences of romantic relationships and sexual activity among adolescents not attending school. It may be that the processes detailed in this dissertation operate differently for such youth. Future research that gathers more detailed information on the cultural climates of non-school contexts for

130

adolescents not attending schools may promote the understanding of the cultural processes being examined in this dissertation.

Despite these limitations, this dissertation advances the understanding of the link between culture and action as it pertains to adolescent sexual behavior and romantic relationships. This dissertation also lays the groundwork for future research that may promote adolescent health and well-being. For instance, while delaying sexual onset is likely best for most adolescents, this study suggests the adverse effects of sexual intercourse on mental health depend on gender and cultural frames within the larger school context. Research that takes a contextual approach to understanding the sexual double standard may shed more light onto the processes through which gender-biased cultural frames develop within and across key contexts. This may help alleviate the adverse mental health consequences of sexual activity, particularly among girls. In addition, identifying additional social precursors to adolescent romantic relationship inauthenticity may shed more light onto how romantic involvement influences adolescent mental health. Finally from a life course perspective, it remains unclear whether inauthenticity in early romances influences romantic relationships during adulthood. It may be the case that adolescents who engage in inauthentic relationships form inauthentic or otherwise unsatisfying relationships during adulthood. Future research examining the association between romantic relationship inauthenticity in early adolescence and features of adult relationships may shed more light onto the dynamics of adult unions.

The findings demonstrate key contingencies in the association between ideal strategies of action and relationship outcomes. I also demonstrate the mental health

131

consequences that potentially accompany mismatches between one’s ideal scripts and actual romantic relationships. Finally, I illustrate the importance of cultural frames in determining how one’s actions influences subsequent mental health. As these studies suggest, culture influences youth development by both informing and reinforcing ideal strategies of action within interpersonal relationships but also by informing interpretations of one’s self and others.

132

References

Aalsma, Matthew C., Melissa Y. Carpentier, Faouzi Azzouz, and J. Dennis Fortenberry. 2012. “Longitudinal Effects of Health-harming and Health-protective Behaviors Within Adolescent Romantic Dyads.” Social Science & Medicine 74(9):1444–51.

Abbott, Andrew, and Angela Tsay. 2000. “Sequence Analysis and Optimal Matching Methods in Sociology Review and Prospect.” Sociological Methods & Research 29(1):3–33.

Akers, Ronald L. 2009. Social Learning and Social Structure: A General Theory of Crime and Deviance. New Brunswick, NJ: Transaction Publishers.

Allen, Joseph P., Joanna Chango, David Szwedo, Megan Schad, and Emily Marston. 2012. “Predictors of Susceptibility to Peer Influence Regarding Substance Use in Adolescence.” Child Development 83(1):337–50.

Anderson, Elijah. 1989. “Sex Codes and Family Life Among Poor Inner-city Youths.” The Annals of the American Academy of Political and Social Science 501(1):59– 78.

Andersson, Matthew A. 2012. “Identity Crises in Love and at Work Dispositional Optimism as a Durable Personal Resource.” Social Psychology Quarterly 75(4):290–309.

Armour, Stacy, and Dana L. Haynie. 2007. “Adolescent Sexual Debut and Later Delinquency.” Journal of Youth and Adolescence 36(2):141–52.

Armstrong, Elizabeth A., Paula England, and Alison C. K. Fogarty. 2012. “Accounting for Women’s Orgasm and Sexual Enjoyment in College Hookups and Relationships.” American Sociological Review 77(3):435–62.

Armstrong, Elizabeth A., Laura Hamilton, and Paula England. 2010. “Is Hooking Up Bad for Young Women?” Contexts 9(3):22–27.

Arnett, Jeffrey Jensen. 1999. “Adolescent Storm and Stress, Reconsidered.” American Psychologist 54(5):317–26.

133

Bearman, Peter S., and Hannah Brückner. 2001. “Promising the Future: Virginity Pledges and First Intercourse.” The American Journal of Sociology 106:859–912.

Brown, B. Bradford, Candice Feiring, and Wyndol Furman. 1999. “Missing the Love Boat: Why Researchers Have Shied Away from Adolescent Romance.” Pp. 1–16 in The development of romantic relationships in adolescence, Cambridge studies in social and emotional development., edited by Wyndol Furman, Brown B. Bradford, and Candice Feiring. New York: Cambridge University Press.

Brown, Lyn Mikel, and Carol Gilligan. 1993. “Meeting at the Crossroads: Women’s Psychology and Girls’ Development.” Feminism & Psychology 3(1):11–35.

Brumariu, Laura E., and Kathryn A. Kerns. 2010. “Parent–child Attachment and Internalizing Symptoms in Childhood and Adolescence: A Review of Empirical Findings and Future Directions.” Development and Psychopathology 22(01):177– 203.

Burke, Peter J. 1991. “Identity Processes and Social Stress.” American Sociological Review 56(6):836–49.

Burke, Peter J., and Jan E. Stets. 2009. Identity Theory. New York: Oxford University Press.

Carver, Karen, Kara Joyner, and J. Richard Udry. 2003. “National Estimates of Adolescent Romantic Relationships.” Pp. 23–56 in Adolescent Romantic Relations and Sexual Behavior: Theory, Research, and Practical Implications., edited by Paul Florsheim. Mahwah, NJ,: Lawrence Erlbaum Associates Publishers.

Cast, Alicia D., and Peter J. Burke. 2002. “A Theory of Self-Esteem.” Social Forces 80(3):1041–68.

Chantala, Kim, and Joyce Tabor. 2010. “Strategies to Perform a Design-Based Analysis Using the Add Health Data.” National Longitudinal Study of Adolescent Health. Retrieved May 25, 2011 (http://www.cpc.unc.edu/projects/addhealth/faqs/aboutdata/weight1.pdf).

Chu, Judy Y., Michelle V. Porche, and Deborah L. Tolman. 2005. “The Adolescent Masculinity Ideology in Relationships Scale Development and Validation of a New Measure for Boys.” Men and Masculinities 8(1):93–115.

Cloward, Richard A., and Lloyd E. Ohlin. 1960. Delinquency and Opportunity: A Theory of Delinquent Gangs. Glencoe: Free Press.

134

Cohen, Geoffrey L., and Mitchell J. Prinstein. 2006. “Peer Contagion of Aggression and Health Risk Behavior Among Adolescent Males: An Experimental Investigation of Effects on Public Conduct and Private Attitudes.” Child Development 77(4):967–83.

Collins, W. Andrew. 2003. “More Than Myth: The Developmental Significance of Romantic Relationships During Adolescence.” Journal of Research on Adolescence 13(1):1–24.

Collins, W. Andrew, Deborah P. Welsh, and Wyndol Furman. 2009. “Adolescent Romantic Relationships.” Annual Review of Psychology 60(1):631–52.

Connolly, Jennifer, Wendy Craig, Adele Goldberg, and Debra Pepler. 2004. “Mixed- Gender Groups, Dating, and Romantic Relationships in Early Adolescence.” Journal of Research on Adolescence 14(2):185–207.

Cornwell, Benjamin, and Edward O. Laumann. 2011. “Network Position and Sexual Dysfunction: Implications of Partner Betweenness for Men.” American Journal of Sociology 117(1):172–208.

Costello, E. Jane, Alaattin Erkanli, and Adrian Angold. 2006. “Is There an Epidemic of Child or Adolescent Depression?” Journal of Child Psychology and Psychiatry 47(12):1263–71.

Crawford, Mary, and Danielle Popp. 2003. “Sexual Double Standards: A Review and Methodological Critique of Two Decades of Research.” The Journal of Sex Research 40(1):13–26.

Cross, Susan E., and Laura Madson. 1997. “Models of the Self: Self-Construals and Gender.” Psychological Bulletin 122(1):5–37.

Dijkstra, Jan Kornelis, Antonius H. N. Cillessen, Siegwart Lindenberg, and René Veenstra. 2010. “Basking in Reflected Glory and Its Limits: Why Adolescents Hang Out With Popular Peers.” Journal of Research on Adolescence 20(4):942– 58.

Eder, Donna, Catherine C. Evans, and Stephen Parker. 1995. School Talk: Gender and Adolescent Culture. New Brunswick, NJ: Rutgers University Press.

Ellis, Wendy E., and Lynne Zarbatany. 2007. “Peer Group Status as a Moderator of Group Influence on Children’s Deviant, Aggressive, and Prosocial Behavior.” Child Development 78(4):1240–54.

Erikson, Erik H. 1968. Identity: Youth and Crisis. New York: W.W. Norton.

135

Exner-Cortens, Deinera, John Eckenrode, and Emily Rothman. 2013. “Longitudinal Associations Between Teen Dating Violence Victimization and Adverse Health Outcomes.” Pediatrics 131(1):71–78.

Felmlee, Diane. 2001. “No Couple Is an Island: A Social Network Perspective on Dyadic Stability.” Social Forces 79(4):1259–87.

Frisco, Michelle L., Jason N. Houle, and Molly A. Martin. 2010. “The Image in the Mirror and the Number on the Scale Weight, Weight Perceptions, and Adolescent Depressive Symptoms.” Journal of Health and Social Behavior 51(2):215–28.

Furman, Wyndol, and Valerie A. Simon. 1999. “Cognitive Representations of Adolescent Romantic Relationships.” Pp. 75–98 in The development of romantic relationships in adolescence, Cambridge studies in social and emotional development., edited by Wyndol Furman, Brown B. Bradford, and Candice Feiring. New York: Cambridge University Press.

Gabadinho, Alexis, Gilbert Ritschard, Nicolas S. Müller, and Matthias Studer. 2011. “Analyzing and Visualizing State Sequences in R with TraMineR.” Journal of Statistical Software 40:1–37.

Giordano, Peggy C. 2003. “Relationships in Adolescence.” Annual Review of Sociology 29:257–81.

Giordano, Peggy C., Robert A. Lonardo, Wendy D. Manning, and Monica A. Longmore. 2010. “Adolescent Romance and Delinquency: A Further Exploration of Hirschi’s ‘Cold and Brittle’ Relationships Hypothesis.” Criminology 48(4):919–46.

Giordano, Peggy C., Monica A. Longmore, and Wendy D. Manning. 2006. “Gender and the Meanings of Adolescent Romantic Relationships: A Focus on Boys.” American Sociological Review 71(2):260–87.

Giordano, Peggy C., Monica A. Longmore, Wendy D. Manning, and Miriam J. Northcutt. 2009. “Adolescent Identities and Sexual Behavior: An Examination of Anderson’s Player Hypothesis.” Social Forces 87(4):1813–43.

Giordano, Peggy C., Wendy D. Manning, and Monica A. Longmore. 2006. “Adolescent Romantic Relationships: An Emerging Portrait of Their Nature and Developmental Significance.” Pp. 127–50 in Romance and sex in adolescence and emerging adulthood: Risks and opportunities, edited by A. C. Crouter and A. Booth. Mahwah, NJ: Erlbaum.

Goodson, Patricia, Eric R. Buhi, and Sarah C. Dunsmore. 2006. “Self-esteem and Adolescent Sexual Behaviors, Attitudes, and Intentions: a Systematic Review.” Journal of Adolescent Health 38(3):310–19. 136

Gottfredson, Michael R., and Travis Hirschi. 1990. A General Theory of Crime. Stanford University Press.

La Greca, Annette M., and Hannah Moore Harrison. 2005. “Adolescent Peer Relations, Friendships, and Romantic Relationships: Do They Predict Social Anxiety and Depression?” Journal of Clinical Child & Adolescent Psychology 34(1):49–61.

Hall, Granville Stanley. 1904. Adolescence: Its Psychology and Its Relations to Physiology, Anthropology, Sociology, Sex, Crime, Religion and Education. Englewood Cliffs, NJ: Prentice-Hall.

Hallfors, Denise D., Martha W. Waller, Daniel Bauer, Carol A. Ford, and Carolyn T. Halpern. 2005. “Which Comes First in Adolescence—Sex and Drugs or Depression?” American Journal of Preventive Medicine 29(3):163–70.

Halpern, Carolyn T., Selene G. Oslak, Mary L. Young, Sandra L. Martin, and Lawrence L. Kupper. 2001. “Partner Violence Among Adolescents in Opposite-Sex Romantic Relationships: Findings From the National Longitudinal Study of Adolescent Health.” Am J Public Health 91(10):1679–85.

Halpern, Carolyn Tucker, Christine E. Kaestle, and Denise Dion Hallfors. 2007. “Perceived Physical Maturity, Age of Romantic Partner, and Adolescent Risk Behavior.” Prevention Science 8(1):1–10.

Hamilton, Laura, and Elizabeth A. Armstrong. 2009. “Gendered Sexuality in Young Adulthood Double Binds and Flawed Options.” Gender & Society 23(5):589–616.

Harding, David J. 2007. “Cultural Context, Sexual Behavior, and Romantic Relationships in Disadvantaged Neighborhoods.” American Sociological Review 72(3):341–64.

Harding, David J. 2010. Living the Drama: Community, Conflict and Culture Among Inner-City Boys. Chicago: University of Chicago Press.

Harter, Susan. 1999. The Construction of the Self: A Developmental Perspective. The Guilford Press.

Haynie, D. L. 2001. “Delinquent Peers Revisited: Does Network Structure Matter?” American Journal of Sociology 106(4):1013–57.

Haynie, Dana L. 2002. “Friendship Networks and Adolescent Delinquency: The Relative Nature of Peer Delinquency.” Journal of Quantitative Criminology 18(2):99–134.

Haynie, Dana L., and D. Wayne Osgood. 2005. “Reconsidering Peers and Delinquency: How Do Peers Matter?” Social Forces 84(2):1109–30.

137

Heckman, James J. 1976. “The Common Structure of Statistical Models of Truncation, Sample Selection and Limited Dependent Variables and a Simple Estimator for Such Models.” Annals of Economic and Social Measurement 5(4):475–92.

Huet, Ellen. 2012. “High school says students had ‘Fantasy Slut League’.” The San Francisco Chronicle, October 23, Metro; p. C2.

Impett, Emily A., Deborah Schooler, and Deborah L. Tolman. 2006. “To Be Seen and Not Heard: Femininity Ideology and Adolescent Girls’ Sexual Health.” Archives of Sexual Behavior 35(2):129–42.

Impett, Emily A., Lynn Sorsoli, Deborah Schooler, James M. Henson, and Deborah L. Tolman. 2008. “Girls’ Relationship Authenticity and Self-esteem Across Adolescence.” Developmental Psychology 44(3):722–33.

Joyner, Kara, and J. Richard Udry. 2000. “You Don’t Bring Me Anything but Down: Adolescent Romance and Depression.” Journal of Health and Social Behavior 41(4):369–91.

Kirk, David S., and Andrew V. Papachristos. 2011. “Cultural Mechanisms and the Persistence of Neighborhood Violence.” American Journal of Sociology 116(4):1190–1233.

Kornrich, Sabino, Julie Brines, and Katrina Leupp. 2013. “Egalitarianism, Housework, and Sexual Frequency in Marriage.” American Sociological Review 78(1):26–50.

Kreager, Derek A., and Dana L. Haynie. 2011. “Dangerous Liaisons? Dating and Drinking Diffusion in Adolescent Peer Networks.” American Sociological Review 76(5):737–63.

Lamont, Michele, and Mario Louis Small. 2008. “How Culture Matters: Enriching Our Understanding of Poverty.” Pp. 76–102 in The Colors of Poverty: Why Racial and Ethnic Disparities Persist, edited by Ann Lin and David Harris. New York: Russell Sage Foundation.

Larson, Matthew, and Gary Sweeten. 2012. “Breaking up Is Hard to Do: Romantic Dissolution, Offending, and Substance Use During the Transition to Adulthood*.” Criminology 50(3):605–36.

Loftus, Jeni, Brian Kelly, and Sarah Mustillo. 2011. “Depressive Symptoms Among Adolescent Girls in Relationships with Older Partners: Causes and Lasting Effects?” Journal of Youth and Adolescence 40(7):800–813.

138

Longmore, Monica A., Wendy D. Manning, Peggy C. Giordano, and Jennifer L. Rudolph. 2004. “Self-Esteem, Depressive Symptoms, and Adolescents’ Sexual Onset.” Social Psychology Quarterly 67(3):279–95.

Lyons, Heidi, Peggy C. Giordano, Wendy D. Manning, and Monica A. Longmore. 2010. “Identity, Peer Relationships, and Adolescent Girls’ Sexual Behavior: An Exploration of the Contemporary Double Standard.” Journal of Sex Research 48(5):437–49.

Manning, Wendy D., Monica A. Longmore, and Peggy C. Giordano. 2005. “Adolescents’ Involvement in Non-romantic Sexual Activity.” Social Science Research 34(2):384–407.

Markus, Hazel R., and Shinobu Kitayama. 1991. “Culture and the Self: Implications for Cognition, Emotion, and Motivation.” Psychological Review 98(2):224–53.

Matsueda, Ross L. 1982. “Testing Control Theory and Differential Association: A Causal Modeling Approach.” American Sociological Review 47(4):489–504.

Matsueda, Ross L. 1992. “Reflected Appraisals, Parental Labeling, and Delinquency: Specifying a Symbolic Interactionist Theory.” The American Journal of Sociology 97(6):1577–1611.

McCarthy, Bill, and Teresa Casey. 2008. “Love, Sex, and Crime: Adolescent Romantic Relationships and Offending.” American Sociological Review 73(6):944–69.

McGuire, William J., and Claire V. McGuire. 1982. “Significant Others in Self-space: Sex Differences and Developmental Trends in the Social Self.” Psychological Perspectives on the Self 1:71–96.

Mead, George Herbert. 1934. Mind, Self, and Society. Chicago: University of Chicago Press.

Meier, Ann M. 2007. “Adolescent First Sex and Subsequent Mental Health.” American Journal of Sociology 112(6):1811–47.

Moran, Jeffrey. 2000. Teaching Sex: The Shaping of Adolescence in the 20th Century. Cambridge, MA: Harvard University Press.

Neff, Kristin D., and Marie-Anne Suizzo. 2006. “Culture, Power, Authenticity and Psychological Well-being Within Romantic Relationships: A Comparison of European American and Mexican Americans.” Cognitive Development 21(4):441–57.

Pascoe, C. J. 2007. Dude, You’re a Fag: Masculinity and Sexuality in High School. Berkeley: University of California Press. 139

Perreira, Krista M., Natalia Deeb-Sossa, Kathleen Mullan Harris, and Kenneth Bollen. 2005. “What Are We Measuring? An Evaluation of the CES-D Across Race/Ethnicity and Immigrant Generation.” Social Forces 83(4):1567–1601.

Perrone, D., C. J. Sullivan, T. C. Pratt, and S. Margaryan. 2004. “Parental Efficacy, Self- Control, and Delinquency: A Test of a General Theory of Crime on a Nationally Representative Sample of Youth.” International Journal of Offender Therapy and Comparative Criminology 48(3):298–312.

Pollack, William. 1998. Real Boys: Rescuing Our Sons from the Myths of Boyhood. New York: Heny Holt.

Radloff, Lenore Sawyer. 1977. “The CES-D Scale.” Applied Psychological Measurement 1(3):385–401.

Reiss, Ira L. 1964. “The Scaling of Premarital Sexual Permissiveness.” Journal of Marriage and Family 26(2):188–98.

Roberts, Robert E., Peter M. Lewinsohn, and John R. Seeley. 1991. “Screening for Adolescent Depression: A Comparison of Depression Scales.” Journal of the American Academy of Child & Adolescent Psychiatry 30(1):58–66.

Rosenberg, Morris. 1965. Society and the Adolescent Self-image. Prince: Princeton University Press.

Rosenfield, Sarah, and Dawne Mouzon. 2013. “Gender and Mental Health.” Pp. 277–96 in Handbook of the Sociology of Mental Health, edited by Carol S. Aneshensel, Jo C. Phaelan, and Alex Bierman. New York: Springer.

Rostosky, Sharon S., Brian L. Wilcox, Margaret Laurie Comer Wright, and Brandy A. Randall. 2004. “The Impact of Religiosity on Adolescent Sexual Behavior: A Review of the Evidence.” Journal of Adolescent Research 19(6):677–97.

Royston, Patrick. 2004. “Multiple Imputation of Missing Values.” Stata Journal 4(3):227–41.

Shaw, Clifford, and Henry D. McKay. 1942. Juvenile Delinquency and Urban Areas. Chicago: University of Chicago Press.

Shi, Bing, and Hongling Xie. 2012. “Socialization of Physical and Social Aggression in Early Adolescents’ Peer Groups: High-status Peers, Individual Status, and Gender.” Social Development 21(1):170–94.

Shields, Ryan T., and Kevin M. Beaver. 2011. “The Effects of Nonshared Environments on Adolescent Depression: Findings From a Sample of Monozygotic Twins.” Journal of Adolescent Health 48(6):572–78. 140

Simon, Robin W., and Anne E. Barrett. 2010. “Nonmarital Romantic Relationships and Mental Health in Early Adulthood Does the Association Differ for Women and Men?” Journal of Health and Social Behavior 51(2):168–82.

Simons, Ronald L., and Ashley B. Barr. forthcoming. “Shifting Perspectives: Cognitive Changes Mediate the Impact of Romantic Relationships on Desistance from Crime.” Justice Quarterly

Snow, David A., and Robert D. Benford. 1992. “Master Frames and Cycles of Protest.” in Frontiers in Social Movement Theory, edited by Aldon D. Morris and Carol Mueller. New Haven, CT: Yale University Press.

Snow, Kate, and Kelly Hagan. 2009. “‘Slut List’: Teen Girls Hazed at N.J. High School.” ABC News, Retrieved January 26, 2013 (http://abcnews.go.com/GMA/teen-girls- hazed-slut-list/story?id=8649050).

Social Security Act. 1996. 42 U.S.C. Title 5, § 510 (b2E). Separate Program for Abstinence Education.

Soller, Brian, Christopher R. Browning, and Aubrey L. Jackson. 2012. “Legal cynicism and parental appraisals of adolescent violence.” Presented at the annual meeting of the American Sociological Association, August 20, Denver, CO.

Soller, Brian, and Dana L. Haynie. 2012a. “Putting the sexual double standard in its place: frames, peer status, and sexual behavior.” Presented at the annual meeting of the American Sociological Association, August 19, Denver, CO.

Soller, Brian, and Dana L. Haynie. 2012b. “Scripts, sects, and sex: A network approach of reference groups.” Presented at the Annual Meeting of the Population Association of America, May 4, San Francisco, CA.

Sprecher, Susan, and Diane Felmlee. 2000. “Romantic Partners’ perceptions of Social Network Attributes with the Passage of Time and Relationship Transitions.” Personal Relationships 7(4):325–40.

Spriggs, Aubrey L., and Carolyn T. Halpern. 2008. “Sexual Debut Timing and Depressive Symptoms in Emerging Adulthood.” Journal of Youth and Adolescence 37(9):1085–96.

Staiger, Annegret. 2005. “‘Hoes Can Be Hoed Out, Players Can Be Played Out, but Pimp Is for life’—The Pimp Phenomenon as Strategy of Identity Formation.” Symbolic Interaction 28(3):407–28.

Stets, Jan E., and Peter J. Burke. 2005. “Identity Verification, Control, and Aggression in Marriage.” Social Psychology Quarterly 68(2):160–78.

141

Stovel, Katherine, Michael Savage, and Peter Bearman. 1996. “Ascription into Achievement: Models of Career Systems at Lloyds Bank, 1890-1970.” American Journal of Sociology 102(2):358–99.

Sutherland, Edwin. 1947. Principles of Criminology. Philadelphia: J.B. Lippincott.

Swidler, Ann. 1986. “Culture in Action: Symbols and Strategies.” American Sociological Review 51(2):273–86.

Swidler, Ann. 2001. Talk of Love: How Culture Matters. Chicago: University of Chicago Press.

Theran, Sally A. 2011. “Authenticity in Relationships and Depressive Symptoms: A Gender Analysis.” Personality and Individual Differences 51(4):423–28.

Thoits, Peggy A. 2012. “Role-Identity Salience, Purpose and Meaning in Life, and Well- Being Among Volunteers.” Social Psychology Quarterly 75(4):360–84.

Tolman, Deborah L. 2005. Dilemmas of Desire: Teenage Girls Talk About Sexuality. Boston: Harvard University Press.

Tolman, Deborah L., Emily A. Impett, Allison J. Tracy, and Alice Michael. 2006. “Looking Good, Sounding Good: Femininity Ideology and Adolescent Girls’ Mental Health.” Psychology of Women Quarterly 30(1):85–95.

Tolman, Deborah L., and Sara I. McClelland. 2011. “Normative Sexuality Development in Adolescence: A Decade in Review, 2000–2009.” Journal of Research on Adolescence 21(1):242–55.

Tolman, Deborah L., and Michelle V. Porche. 2000. “The Adolescent Femininity Ideology Scale: Development and Validation of a New Measure for Girls.” Psychology of Women Quarterly 24(4):365–76.

Turner, C. F. et al. 1998. “Adolescent Sexual Behavior, Drug Use, and Violence: Increased Reporting with Computer Survey Technology.” Science 280(5365):867–73.

Vogt Yuan, Anastasia. 2010. “Body Perceptions, Weight Control Behavior, and Changes in Adolescents’ Psychological Well-Being Over Time: A Longitudinal Examination of Gender.” Journal of Youth and Adolescence 39(8):927–39.

Walsemann, Katrina M., Bethany A. Bell, and Debeshi Maitra. 2011. “The Intersection of School Racial Composition and Student Race/ethnicity on Adolescent Depressive and Somatic Symptoms.” Social Science & Medicine 72(11):1873–83.

142

Warner, Tara D., Peggy C. Giordano, Wendy D. Manning, and Monica A. Longmore. 2011. “Everybody’s Doin’ It (right?): Neighborhood Norms and Sexual Activity in Adolescence.” Social Science Research 40(6):1676–90.

Warren, Jocelyn T., S. Marie Harvey, and Jillian T. Henderson. 2010. “Do Depression and Low Self-Esteem Follow Abortion Among Adolescents? Evidence from a National Study.” Perspectives on Sexual and Reproductive Health 42(4):230–35.

Welsh, Deborah P., Sharon S. Rostosky, and Myra Christen Kawaguchi. 2000. “A Normative Perspective of Adolescent Girls’ Developing Sexuality.” Pp. 111–40 in Sexuality, society, and feminism, Psychology of women; 4., edited by C. B. Travis and J. W. White. Washington, DC, US: American Psychological Association.

143

Appendix A: Tables from Chapter 2

144

Table A.1. Descriptive Statistics Total Sample Girls Boys (N=8,513) (N=4,296) (N=4,217) Variables Mean (SD) Mean (SD) Mean (SD) Individual Variables Severe Depression .13 .16 .09 High Self-Esteem .20 .18 .23 No Sexual Intercourse Between Waves .46 .46 .47 Romantic-Only Sexual Intercourse .33 .37 .28 Non-Romantic Sexual Intercourse .21 .17 .25 Age 16.73 (1.00) 16.67 (.98) 16.80 (1.02) Male .50 White .55 .54 .55 Latino .17 .17 .18 Black .20 .21 .18 Other .08 .08 .09 Socioeconomic Status -.05 (.89) -.08 (.89) -.01 (.89) Single Parent Household .31 .32 .31 Parental Attachment 4.47 (.58) 4.42 (.65) 4.52 (.51) Religiosity .00 (.83) .08 (.82) -.07 (.83) Abstinence Pledge .12 .16 .09 Problem with Peers .83 (.97) .81 (.95) .85 (.98) Peer Attachment 3.71 (.87) 3.66 (.90) 3.75 (.84) Perceived Benefits of Sexual Intercourse 2.42 (.59) 2.11 (.48) 2.73 (.53) Prior Sexual Intercourse .49 .46 .52 Change in Problem with Peers -.10 (1.08) -.12 (1.03) -.07 (1.14) Change in Peer Attachment -.05 (.84) -.03 (.87) -.07 (.81) Prior Severe Depression .14 .18 .10 Prior High Self-Esteem .19 .19 .19 School Variables (N=75) Sexual Double Standard .61 (.12) Proportion Had Sexual Intercourse .49 (.15) Region: South .39 Region: West .20 Region: Midwest .24 Region: Northeast .17 Suburban .52 Urban .29 Rural .19 Private School .11 School Size 1179.53 (810.39)

145

Table A.2. Multilevel Logistic Regression Models of Severe Depression Regressed on School Sexual Double Standards and Sexual Behavior Relationship Context Variable Model 1 Model 2 Model 3 Girls Intercept -2.01 ** (.08) -2.06 ** (.16) -2.07 ** (.15) Female*Romantic Sexual Intercourse .11 (.20) .10 (.20) Female*Non-Romantic Sexual .06 (.25) -.06 (.25) Intercourse Sexual Double Standard (SDS)*Female 1.11 * (.55) 1.11 (.56) -.50 (.68) SDS*Female*Romantic Sexual 2.08 * (1.04) Intercourse SDS*Female*Non-Romantic Sexual 4.42 ** (1.28) Intercourse Boys (.00) Intercept -2.77 ** (.09) -2.88 ** (.17) -2.89 ** (.17) Male*Romantic Sexual Intercourse .15 (.30) .17 (.31) Male*Non-Romantic Sexual Intercourse .24 (.26) .28 (.27) SDS*Male 1.24 (.67) 1.17 (.72) 1.56 (1.57) SDS*Male*Romantic Sexual -.27 (2.38) Intercourse SDS*Male*Non-Romantic Sexual -.62 (1.85) Intercourse Note: Robust standard errors in parentheses. Selected coefficients are omitted from the table (see appendix). Missing values on individual-level variables were imputed using multiple imputation with 10 replications. Individual N = 8,513; School N = 75. **p < .01, *p < .05 (two-tailed tests).

146

Table A.3. Multilevel Logistic Regression Models of High Self-Esteem Regressed on School Sexual Double Standards and Sexual Behavior Relationship Context Variable Model 1 Model 2 Model 3 Girls Intercept -2.03 ** (.09) -2.06 ** (.16) -2.06 ** (.16) Female*Romantic Sexual Intercourse .10 (.22) .07 (.22) Female*Non-Romantic Sexual .01 .01 (.22) (.22) Intercourse Sexual Double Standard (SDS)*Female -.21 (.42) -.23 (.41) -.54 (.50) SDS*Female*Romantic Sexual 1.46 (1.33) Intercourse SDS*Female*Non-Romantic Sexual .34 (1.24) Intercourse Boys Intercept -1.54 ** (.07) -1.42 ** (.09) -1.42 ** (.09) Male*Romantic Sexual Intercourse -.20 (.16) -.16 (.16) Male*Non-Romantic Sexual Intercourse -.25 (.14) -.40 ** (.13) SDS*Male 1.12 * (.44) 1.18 * (.44) 1.00 * (.46) SDS*Male*Romantic Sexual -.63 (.87) Intercourse SDS*Male*Non-Romantic Sexual 3.11 * (1.24) Intercourse Note: Robust standard errors in parentheses. Selected coefficients are omitted from the table (see appendix). Missing values on individual-level variables were imputed using multiple imputation with 10 replications. Individual N = 8,513; School N = 75. **p < .01, *p < .05 (two-tailed tests).

147

Table A.4. Coefficients for Control Variables Omitted from Table A.2. Variable Model 1 Model 2 Model 3 Individual Control Variables Age .15 ** (.04) .14 ** (.04) .14 ** (.04) Latino .26 (.22) .25 (.22) .26 (.23) Black -.12 (.15) -.13 (.15) -.14 (.16) Other .66 ** (.19) .68 ** (.19) .67 ** (.19) Socioeconomic Status -.10 (.07) -.09 (.07) -.09 (.07) Single Parent Household .07 (.12) .06 (.12) .06 (.12) Parental Attachment -.13 (.09) -.13 (.09) -.13 (.09) Religiosity -.05 (.08) -.05 (.08) -.05 (.08) Abstinence Pledge .37 ** (.19) .37 ** (.19) .34 ** (.18) Problem with Peers .32 ** (.07) .32 ** (.07) .32 ** (.07) Change in Problem with Peers .30 ** (.06) .30 ** (.06) .31 ** (.06) Peer Attachment -.48 ** (.08) -.48 ** (.08) -.48 ** (.08) Change in Peer Attachment -.36 ** (.06) -.36 ** (.06) -.36 ** (.06) Friend Attachment -.19 ** (.11) -.20 ** (.11) -.20 ** (.11) Change in Friend Attachment -.14 ** (.07) -.13 ** (.07) -.13 ** (.07) Perceived Benefits of Sexual Intercourse .31 ** (.09) .31 ** (.09) .31 ** (.10) Prior Sexual Intercourse .25 ** (.14) .18 (.14) .16 (.13) Prior Severe Depression 1.64 ** (.15) 1.64 ** (.15) 1.66 ** (.15) School Control Variables Proportion Had Sexual Intercourse*Female .82 (.52) .84 (.53) 1.01 ** (.58) Proportion Had Sexual Intercourse*Male .87 (.72) .82 (.70) .84 (.71) Region: West -.16 (.17) -.16 (.17) -.15 (.18) Region: Midwest -.30 ** (.15) -.31 ** (.16) -.30 ** (.17) Region: Northeast -.05 (.18) -.05 (.18) -.03 (.19) Urban .08 (.17) .08 (.17) .07 (.19) Rural -.09 (.14) -.10 (.14) -.11 (.15) School Size .00 (.00) .00 (.00) .00 (.00) Note: Robust standard errors in parentheses. Missing val ues on individual -level variables were imputed using multiple imputation with 10 replications. Individual N = 8,513; School N = 75. **p < .01, *p < .05 (two-tailed tests).

148

Table A.5. Coefficients for Control Variables Omitted from Table A.3. Variable Model 1 Model 2 Model 3 Individual Control Variables Age -.06 (.04) -.05 (.04) -.06 (.04) Latino -.35 (.28) -.34 (.28) -.35 (.27) Black .24 ** (.12) .26 ** (.12) .25 ** (.12) Other -.53 ** (.31) -.54 ** (.31) -.54 ** (.32) Socioeconomic Status .02 (.05) .02 (.05) .02 (.05) Single Parent Household -.04 (.09) -.04 (.09) -.06 (.09) Parental Attachment .45 ** (.10) .45 ** (.10) .45 ** (.10) Religiosity .17 ** (.07) .17 ** (.07) .16 ** (.07) Abstinence Pledge .19 (.15) .19 (.15) .18 (.16) Problem with Peers -.12 (.08) -.12 (.08) -.12 (.08) Change in Problem with Peers -.11 (.07) -.11 (.07) -.11 (.07) Peer Attachment .49 ** (.09) .49 ** (.09) .48 ** (.09) Change in Peer Attachment .47 ** (.08) .47 ** (.08) .47 ** (.08) Friend Attachment .43 ** (.09) .43 ** (.09) .45 ** (.08) Change in Friend Attachment .30 ** (.07) .30 ** (.07) .30 ** (.07) Perceived Benefits of Sexual Intercourse -.14 (.09) -.13 (.09) -.13 (.09) Prior Sexual Intercourse .25 ** (.09) .29 ** (.11) .29 ** (.11) Prior High Self-Esteem 1.56 ** (.11) 1.56 ** (.11) 1.56 ** (.11) School Control Variables Proportion Had Sexual Intercourse*Female -.08 (.48) .02 (.48) .05 (.47) Proportion Had Sexual Intercourse*Male .67 (.59) .56 (.60) .63 (.58) Region: West -.07 (.14) -.07 (.14) -.05 (.14) Region: Midwest .16 (.13) .16 (.13) .18 (.13) Region: Northeast -.27 ** (.14) -.27 ** (.14) -.26 ** (.14) Urban -.16 (.11) -.15 (.11) -.16 (.11) Rural -.08 (.12) -.08 (.12) -.07 (.12) School Size .00 (.00) .00 (.00) .00 (.00) Note: Robust standard errors in parentheses. Missing values on individual -level variables were imputed using multiple imputation with 10 replications. Individual N = 8,513; School N = 75. **p < .01, *p < .05 (two-tailed tests).

149

Appendix B: Figures from Chapter 2

150

Figure B.1. Predicted Probability of Severe Depression Among Girls by the Sexual Double Standard and Sexual Relationship Categories

0.25

0.20

0.15

0.10

0.05 Abstained Between Waves Romantic Sexual Behavior

Probability of Severe Depression Severe of Probability Non-Romantic Sexual Partner 0.00 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Sexual Double Standard (SD)

151

Figure B.2. Predicted Probability of High Self-Esteem Among Boys by the Sexual Double Standard and Sexual Relationship Categories

0.35

0.30

Esteem - 0.25

0.20

0.15

0.10 Abstained Between Waves

Probability of High Self High of Probability 0.05 Romantic Sexual Behavior Non-Romantic Sexual Partner 0.00 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Sexual Double Standard (SD)

152

Appendix C: Tables from Chapter 3

153

Table C.1. Substitution Cost Matrix for Ideal Relationship Script Items. A B C D E F G H I J K L M N A 0.00 1.70 2.31 3.15 1.46 1.84 2.65 3.23 2.41 3.62 3.28 4.31 5.45 6.13 B 2.14 0.00 1.81 3.08 1.85 2.42 2.08 2.56 2.30 2.92 3.23 3.79 4.84 6.19 C 2.45 2.00 0.00 2.36 2.05 1.93 2.21 2.60 2.17 3.43 2.87 4.04 5.43 6.58 D 2.95 2.55 2.64 0.00 1.64 2.55 1.94 2.38 2.66 2.48 3.03 2.81 4.15 4.41 E 2.79 2.71 2.84 3.20 0.00 1.16 2.12 2.73 2.34 3.37 2.13 3.91 5.49 6.35 F 3.07 2.78 2.70 3.92 2.26 0.00 1.84 2.60 2.21 3.99 1.12 4.71 6.05 6.81 G 3.22 2.58 3.34 3.15 2.83 2.76 0.00 1.08 2.45 2.28 2.44 2.94 4.34 5.24 H 3.66 2.70 3.29 3.39 3.08 2.91 1.93 0.00 1.78 1.74 2.07 2.42 3.67 5.43 I 3.14 2.75 0.96 3.76 2.95 2.47 2.71 2.65 0.00 2.50 2.22 4.16 5.46 6.23 J 3.75 3.30 3.75 3.48 3.38 3.58 2.85 2.45 3.14 0.00 1.93 1.13 1.86 4.49 K 3.00 2.31 2.70 3.37 2.66 2.82 2.08 2.13 2.04 2.37 0.00 2.09 3.70 5.93 L 4.27 3.79 4.42 3.99 4.75 5.00 3.27 2.89 4.23 1.90 3.85 0.00 0.50 3.43 M 3.19 2.83 3.56 2.68 4.05 4.30 2.45 2.46 3.39 3.17 4.07 2.38 0.00 0.88 N 2.50 2.76 3.13 2.04 3.33 3.63 2.87 3.16 3.23 2.47 3.45 2.20 1.37 0.00 Note: Bolded letters correspond to the following relationship events: A) Go out together in a group, B) I would meet my partner’s parents, C) I would tell other people that we were a couple, D) I would see less of my other friends, E) We would go out together alone, F) We would hold hands, G) Exchange presents, H) We would say “I love you” to each other, I) We would think of ourselves as a couple, J) We would talk about contraception or sexually transmitted diseases, K) We would kiss, L) We would touch each other under our clothing; M) We would have sex, and N) My partner/I would get pregnant.

154

Table C.2. Descriptive Information on Additional Variables from Table C.4. Variable Mean SD Min Max Parental Mean responses from the following 4 4.57 (.57) 1 5 Attachment questions: “How close do you feel to your resident mother/father?” and “How much do you think she/he cares about you?” (1=“Not at all,” 5 = “Very much”). For respondents in single parent households, the mean of the 2 appropriate items are used (α=.724). Body Mass Index BMI is calculated as follows: (Weight / 22.98 (4.53) 11.75 51.37 (BMI) Height2)*703. Weight is in pounds and height is in inches. Height and weight are self-reported. Grade Point Mean of the standardized values of most -.07 (.77) -1.88 1.31 Average recent grades in math, English, history, and science (self-reported) on a 4.0 scale (α=.755). Abstinence Pledge Binary variable indicating whether .15 (.36) 0 1 Status respondent has taken a public pledge to remain a virgin until married (0 = no, 1 = yes). Parental Approval Mean responses from the following 2 1.76 (.85) 1 5 of Sex questions: “How would she [your mother] feel about your having sex at this time in your life?” and “How would he [your father] feel about your having sex at this time in your life?” (1 = “Strongly disapprove to 5 = “Strongly approve”). For respondents in single parent households, the mean of the 2 appropriate items are used (r = .705). Ongoing Binary variable indicating whether .40 (.49) 0 1 Relationship respondent was in a romantic relationship at wave 1 (0=no, 1= yes). Recent Romantic In the last 18 months—since [MONTH, .57 (.45) 0 1 Relationship YEAR]—have you had a special romantic relationship with any one? (0=no, 1=yes). Desire for Responses to the question: “How much .84 (.37) 0 1 Romantic would you like to have a romantic Relationship relationship in the next year?” (1 = “Not at all” to 5 = “Very much”). Responses were collapsed to (0 = “Not at all/Very Little” and 1 = “Somewhat to Very much.” Prior Sexual Binary variable indicating whether .42 (.49) 0 1 Intercourse respondent had sexual intercourse prior to the Wave 1 interview (0=no, 1=yes).

155

Table C.3. Descriptive Statistics. Variable Mean SD Min Max Relationship Inauthenticity 2.723 (.843) .000 5.823 Script Discordance .000 (.467) -1.194 2.844 Friend Involvement .000 (1.774) -2.907 5.811 Friends’ Popularity .000 (.278) -.300 .700 Age 16.267 (1.416) 12.693 20.553 Male .460 0 1 White .570 0 1 Black .138 0 1 Latino .150 0 1 Other .142 0 1 Single Parent Household .274 0 1 Socioeconomic Status -.023 (.811 ) -1.733 1.762 Religiosity .103 (.781) -1.541 1.128 Self-Esteem 4.043 (.610) 1.667 5.000 Depression .694 (.580) 0 3 Low Self-Control .070 (.626) -1.119 2.389 Underweight Body Image .153 0 1 Overweight Body Image .326 0 1 Physical Development 3.314 (1.081 ) 1 5 Popularity 3.742 (3.196) 0 19 Friendliness 3.560 (2.494) 0 10 Isolate .045 0 1 Time to Relationship 19.484 (11.729 ) .100 45.400 Romantic Relationship .922 0 1 Relationship is Ongoing .484 0 1 Ideal Script Length 10.549 (1.904 ) 3 14 Relationship Hazard .851 (.221) .266 1.549 Older Partner .088 0 1 Younger Partner .029 0 1 Same Race Partner .789 0 1 Network Overlap -.001 (1.248 ) -2.676 1.324 Mutual Friends .399 0 1 Prior Friends .429 0 1 N =1,013

156

Table C.4. Probit Regression of the Hazard of New Romantic Relationship Between Waves 1 and 2. Variable b SE Age .011 (.025) Male -.147 * (.056) Black .200 (.128) Latino .085 (.124) Other .191 (.127) Single Parent Household .040 (.064) Socioeconomic Statusa .007 (.035) Parental Attachmenta -.085 + (.049) Religiositya .037 (.039) Abstinence Pledgea -.027 (.075) Self-Esteem .073 (.054) Depression .043 (.055) Low Self-Control .084 + (.051) Physical Developmenta .049 + (.025) Body Mass Indexa -.018 ** (.006) Grade Point Averagea -.077 + (.040) Popularity .029 ** (.010) Friendliness .013 (.012) Parental Approval of Sexa .027 (.037) Ongoing Relationship -.150 * (.065) Recent Romantic Relationshipa .416 *** (.062) Desire for Romantic Relationshipa .259 (.075) Prior Sexual Intercoursea -.046 (.067) Intercept -.619 (.513) Note: aDummy variable is included for missing cases. Standard errors are in parentheses. Coefficients for school fixed effects and dummy variables for missing cases are omitted from table. N = 2,617. ***p < .001, **p < .01, *p < .05, +p < .10 (two-tailed tests).

157

Table C.5. Linear Regression Models of Romantic Relationship Inauthenticity Regressed on Script Discordance. Variable Model 1 Model 2 Model 3 Age -.044 + (.020) -.050 + (.023) -.051 + (.024) Male .008 (.112) .007 (.105) -.019 (.099) Black .269 (.183) .249 (.174) .226 (.173) Latino .087 (.204) -.063 (.240) -.067 (.256) Other .289 * (.108) .249 * (.103) .251 * (.103) Single Parent Household .064 (.106) .065 (.110) .067 (.107) Socioeconomic Status .034 (.045) .042 (.045) .043 (.044) Religiosity -.033 (.047) -.029 (.058) -.032 (.055) Self-Esteem -.059 (.046) -.056 (.049) -.090 + (.045) Depression .024 (.066) .030 (.047) .012 (.049) Low Self-Control .021 (.102) .014 (.087) .003 (.078) Underweight Body Image .051 (.073) .054 (.067) .039 (.065) Overweight Body Image -.014 (.068) -.010 (.071) -.022 (.070) Physical Development -.014 (.032) -.014 (.030) -.020 (.033) Popularity .015 (.011) .014 (.010) .014 (.011) Friendliness .000 (.014) .002 (.015) .004 (.015) Isolate .326 (.203) .318 (.206) .320 (.205) Time to Relationship .011 * (.004) .010 * (.003) .010 * (.003) Romantic Relationship .083 (.093) .090 (.092) .144 (.090) Relationship is Ongoing -.366 *** (.067) -.372 *** (.065) -.370 *** (.071) Ideal Script Length .035 (.023) .035 (.024) .063 * (.025) Relationship Hazard .188 (.264) .173 (.259) .138 (.241) Older Partner .015 (.117) .043 (.113) Younger Partner .149 (.297) .186 (.286) Same Race Partner -.145 (.129) -.120 (.150) Network Overlap .006 (.073) .004 (.070) Mutual Friends .058 (.081) .067 (.082) Prior Friends -.113 (.101) -.122 (.102) Script Discordance .287 ** (.064) Intercept 2.872 ** (.684) 3.143 ** (.722) 3.009 ** (.733) Note: Standard errors are in parentheses. Missing values on independent variables were imputed using multiple imputation with 10 replications. N = 1,013. ***p < .001, **p < .01, *p < .05, +p < .10 (two-tailed tests).

158

Table C.6. Linear Regression Models of Romantic Relationship Inauthenticity Regressed on Script Discordance, Friend Involvement, and Friends’ Popularity. Variable Model 4 Model 5 Model 6 Script Discordance .306 ** (.089) .338 *** (.063) .316 ** (.092) Friend Involvement .052 ** (.012) .054 ** (.011) .048 ** (.011) Friends’ Popularity .105 (.089) .109 (.097) .118 (.099) Script Discordance*Friend Involvement -.025 (.048) -.054 (.051) Script Discordance*Friends’ Popularity .543 * (.197) .660 * (.206) Intercept 3.278 ** (.739) 3.219 ** (.764) 3.218 ** (.746) Note: Standard errors are in parentheses. Selected coefficients are omitted from the table (available upon request). Missing values on individual-level variables were imputed using multiple imputation with 10 replications. N = 1,013. ***p < .001, **p < .01, *p < .05.

159

Table C.7. Sensitivity Analyses. Model 7 Model 8 Relationship Behavioral Inauthenticity Discordancea Network Overlap -.001 (.065) .010 (.040) Mutual Friends .079 (.087) -.008 (.078) Prior Friends -.111 (.096) -.093 (.118) Script Discordance .408 * (.140) .141 (.082) Friend Involvement .048 ** (.010) Friends’ Popularity .110 (.106) Script Discordance*Friend Involvement -.056 (.055) Script Discordance*Friends’ Popularity .700 ** (.205) Script Discordance*Network Overlap -.021 (.048) Script Discordance*Mutual Friends -.192 (.147) Script Discordance*Prior Friends -.041 (.159) Intercept 3.176 ** (.733) 2.673 ** (.682) N 1,013 967 Note: aIsolates dropped from analysis. Standard errors are in parentheses. Selected coefficients are omitted from the table (available upon request). Missing values on individual-level variables were imputed using multiple imputation with 10 replications. **p < .01, *p < .05, +p < .10 (two-tailed tests).

160

Appendix D: Figures from Chapter 3

161

Figure D.1. Illustration of High Script Discordance and Low Script Heterogeneity.

Individual i FEDCBA

Friend 1 Friend 2 Friend 3 Friend 4 Friend 5 ABCDEF ABCDEF ABCDEF ABCDEF ABCDEF

162

Figure D.2. Predicted Values of Relationship Inauthenticity by Script Discordance and High and Low Friend Popularity.

3.25

3.00

2.75

2.50

Relationship Inauthenticity Relationship 2.25 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Script Discordance (SD) High Friends' Popularity (+1.5 SD) Low Friends' Popularity (-1.5 SD)

163

Appendix E: Tables from Chapter 4

164

Table E.1. Descriptions of Control Variables Variable Definition Individual Measures Age Age (in years) at wave 1 Malea Binary variable indicating male gender (1=yes) Black Binary variable indicating respondent is black (1=black) Latino/a Binary variable indicating respondent is Latino/a (1=Latino/a) Other Binary variable indicating respondent’s race/ethnicity is other than white, black, or Latino/a (1=other) Socioeconomic Status Interval measure consisting of the standardized values of parents’ highest occupational status and education level (r=.458). One-Parent Household Binary variable indicating respondent lives in a one-parent household (1=yes) Parental Attachment 5-item scale tapping parent/child bonding as captured by responses to questions such as “how close do you feel to your mother?” and “how much do you think your father cares about you?” Each question is asked in reference to the mother and the father, for a potential total of 10 questions. The maximum value from each paired item is taken and the mean the five-items is measured (α=.844). School Connectedness 8-item scale capturing attachment to teachers and schoolmates (e.g., “You feel close to people at your school” and “You feel like you are part of your school”) (see Resnick et al. 1997). School connectedness represents the mean of the standardized items with higher indicating greater school connectedness (α=.771). Religiosity 4-item scale capturing the frequency of prayer, service attendance, and religious youth-group participation, and a variable indicating how important religion is to the respondent. Religiosity consists of the mean of the reverse-coded and standardized items (α=.840). Ideal Script Length Count variable indicating the number of activities respondents would engage in within ideal romantic relationships. Friend/Family Member Binary variable indicating a family or family member attempted suicide in Attempted Suicide the 12 months prior to wave 1 (1=yes) Friend/Family Member Binary variable indicating a family or family committed suicide in the 12 Committed Suicide months prior to wave 1 (1=yes) Grade Point Averagea Self-reported average of grades in 4 subjects, based on a 4-point scale. Verbal Abilitya Scores from Add Health’s abridged computerized version of the Peabody Picture Vocabulary Test (AH-PVT). Scores were standardized scores to have a mean of 100 and a standard deviation of 15. Self-Assessed Self-rated intelligence relative to people their age. Responses ranged from Intelligencea 1 (moderately below average) to 6 (extremely above average) Parental Approval of Perception that parents would strongly disapprove (1) or strongly approve Sexual Intercoursea (5) of them having sex. Responses were averaged across parents. a 2 Body Mass Index Calculated from height and weight as follows: weightkg/heightm Note: aMeasure is only used as a predictor in the selection model. bVariable was measured at wave 2. All other variables were measured at wave 1. (continued on next page)

165

Table E.1. Descriptions of Control Variables (continued) Variable Definition Individual Measures Physical Developmenta Ordinal responses to the question “How advanced is your physical development compared to other boys/girls your age?” Responses ranged from 1 (“I look younger than most”) to 5 (“I look older than most”). Desire for Romantic Desire for a romantic relationship is based on responses to the question: Relationshipa “How much would you like to have a romantic relationship in the next year?” Ordinal responses ranged from 1 (not at all) to 5 (very much). Current Romantic Binary variable indicating romantic involvement at wave 1 (1=yes) Involvementa Prior Romantic Binary variable indicating romantic involvement prior to wave 1 (1=yes) Relationshipa Prior Sexual Binary variable indicating sexual intercourse prior to wave 1 (1=yes) Intercoursea Driven Cara Binary variable indicating whether respondent has driven an automobile (1=yes) New Romantic Binary variable indicating the respondent formed a new romantic Relationshipa,b relationship after the date of the wave 1 survey (1=yes) Relationship Measures Relationship Script Count variable indicating the number of activities respondents engaged in Length within first subsequent romantic relationship. Time to Relationshipb Continuous measure indicating the number of days that elapsed between the first in-home interview and relationship onset, divided by 10 Special Romantic Binary variable indicating whether the respondent the first subsequent Relationshipb romantic relationship is a “special romantic” versus “liked” relationship (1=special romantic relationship, 0=liked relationship) Older Partnerb Binary variable indicating the partner was at least 3 years older than the respondent (1=yes) Younger Partnerb Binary variable indicating the partner was 3 (or more) years younger than the respondent (1=yes) Ongoing Relationshipb Binary variable indicating relationship was ongoing at wave 2 (1=yes). Sexual Intercourse in Binary variable indicating the respondent and partner had sexual Relationshipb intercourse (1=yes) Psychological Abuseb Binary variable indicating the respondent was ever insulted in public, cursed or swore at, or threatened with violence by the partner (1=yes) Physical Abuseb Binary variable indicating the romantic partner ever threw something at or pushed/shoved the respondent (1=yes) Note: aMeasure is only used as a predictor in the selection model. bVariable was measured at wave 2. All other variables were measured at wave 1.

166

Table E.2. Probit Regression of the Hazard of New Romantic Relationship Between Waves 1 and 2. b (se) Age .00 (.01) Male -.14 *** (.02) Black -.09 ** (.03) Latino -.09 ** (.03) Other -.16 *** (.04) Socioeconomic Statusa .03 * (.01) Grade Point Averagea -.08 *** (.02) Verbal Abilitya .00 *** (.00) Self-Assessed Intelligence a .04 ** (.01) Body Mass Indexa -.02 *** (.00) Physical Developmenta .08 *** (.01) Parental Attachmenta -.04 * (.02) Parental Approval of Sexual Intercoursea .02 (.01) Prior Sexual Intercoursea .10 ** (.03) Desire for Romantic Relationshipa .06 *** (.01) Current Romantic Involvementa -.14 *** (.03) Prior Romantic Relationshipa .46 *** (.03) Driven Cara .16 *** (.03) Note: Standard errors are in parentheses. aDummy variable is included for missing cases. Missing values on independent variables were replaced with the variable’s grand mean. N = 14,316. *p<.05, **p<.01, ***p<.001 (two-tailed tests)

167

Table E.3. Descriptive Statistics by Gender. Girls (N=2,905) Boys (N=2,411) Mean/ Mean/ Proportion (SD) Proportion (SD) Dependent Variables Severe Depression .17 .08 High Somatic Symptoms .38 .25 Suicide Ideation .16 .09 Suicide Attempt .07 .02 Relationship Variables Romantic Relationship Inauthenticity 1st Quartile (reference) .26 .24 2nd Quartile .26 .24 3rd Quartile .25 .25 4th Quartile .24 .27 Relationship Script Length 9.76 (3.05) 9.59 (3.08) Time to Relationship 18.72 (11.62) 19.97 (12.00) “Special Romantic” Relationship .93 .90 Age Discordance Same Age (reference) .71 .78 Older Partner .26 .04 Younger Partner .03 .19 Ongoing Relationship .49 .41 Sexual Intercourse .51 .56 Psychological Abuse .19 .21 Physical Abuse .06 .08 Individual Variables Age 15.76 (1.48) 16.06 (1.51) Race-Ethnicity White (reference) .54 .52 Black .21 .20 Latino/a .15 .17 Other .10 .11 One-Parent Household .31 .31 Socioeconomic Status .01 (.86) .07 (.86) Parental Attachment 4.44 (.64) 4.55 (.50) School Connectedness -.01 (.64) -.04 (.63) Religiosity .04 (.81) -.06 (.83) Ideal Script Length 9.54 (2.35) 10.02 (2.39) Relationship Hazard .83 (.21) .89 (.20) Friend/Family Member Attempted Suicide .23 .14 Friend/Family Member Committed Suicide .05 .04 Lagged Dependent Variables Prior Severe Depression .18 .09 Prior High Somatic Symptoms .37 .27 Prior Suicide Ideation .20 .11 Prior Suicide Attempt .07 .02

168

Table E.4. Logistic Regressions of Girls’ Severe Depression and High Somatic Symptoms. Severe Depression High Somatic Symptoms Model 1 Model 2 Model 3 Model 4 Relationship Inauthenticity 2nd Quartile .19 .17 -.09 -.11 (.23) (.23) (.15) (.15) 3rd Quartile .55 ** .48 * .11 .06 (.20) (.21) (.16) (.16) 4th Quartile .69 ** .73 ** .34 * .40 * (.21) (.23) (.16) (.19) Relationship Script Length .02 .02 (.03) (.03) Time to Relationship .00 .00 (.01) (.01) “Special Romantic” Relationship .15 -.25 (.27) (.26) Older Partner .17 .18 (.16) (.12) Younger Partner .38 -.23 (.33) (.34) Ongoing Relationship -.10 -.08 (.20) (.14) Sexual Intercourse .33 .21 (.19) (.15) Psychological Abuse .63 ** .61 *** (.18) (.15) Physical Abuse .33 .15 (.25) (.24) Intercept -.84 -.88 1.39 1.69 (1.03) (1.11) (.94) (1.04) Notes: All models control for age, race/ethnicity, one-parent household, socioeconomic status, parental attachment, school connectedness, religiosity, ideal script length, relationship hazard, and lagged dependent variable. Standard errors are in parentheses. N = 2,905. *p<.05, **p<.01, ***p<.001 (two-tailed tests)

169

Table E.5. Logistic Regressions of Girls’ Suicide Ideation and Suicide Attempt. Suicide Ideation Suicide Attempt Model 1 Model 2 Model 3 Model 4 Relationship Inauthenticity 2nd Quartile .37 .33 .10 .08 (.23) (.23) (.32) (.32) 3rd Quartile .44 * .36 .87 ** .80 * (.19) (.20) (.29) (.31) 4th Quartile .57 * .50 * .32 .23 (.22) (.25) (.34) (.36) Relationship Script Length .03 .02 (.03) (.05) Time to Relationship .00 .01 (.01) (.01) “Special Romantic” Relationship -.09 -.09 (.28) (.36) Older Partner .14 .05 (.17) (.22) Younger Partner .01 .23 (.33) (.37) Ongoing Relationship -.39 * -.35 (.17) (.27) Sexual Intercourse -.13 .10 (.21) (.26) Psychological Abuse .40 * .12 (.19) (.26) Physical Abuse .40 .84 ** (.26) (.32) Intercept 1.39 1.43 1.19 1.53 (1.00) (1.10) (1.52) (1.70) All models control for age, race/ethnicity, one-parent parent household, socioeconomic status, parental attachment, school connectedness, religiosity, ideal script length, relationship hazard, prior depression, binary variables indicating whether a friend/family member attempted suicide and friend/family member committed suicide in past 12 months, and lagged dependent variable. Standard errors are in parentheses. N = 2,905. *p<.05, **p<.01, ***p<.001 (two-tailed tests)

170

Table E.6. Logistic Regressions of Boys’ Severe Depression and High Somatic Symptoms. Severe Depression High Somatic Symptoms Model 1 Model 2 Model 3 Model 4 Relationship Inauthenticity 2nd Quartile .06 .05 -.11 -.10 (.33) (.33) (.21) (.22) 3rd Quartile .09 .15 .21 .23 (.32) (.29) (.19) (.19) 4th Quartile .19 .43 .11 .20 (.30) (.32) (.20) (.26) Relationship Script Length .01 .01 (.04) (.03) Time to Relationship .01 .00 (.01) (.01) “Special Romantic” Relationship .26 .03 (.34) (.26) Older Partner .30 .25 (.39) (.37) Younger Partner .50 .16 (.29) (.18) Ongoing Relationship .26 .04 (.28) (.20) Sexual Intercourse .50 .14 (.28) (.18) Psychological Abuse .11 -.01 (.23) (.22) Physical Abuse .78 ** .12 (.28) (.24) Intercept -4.01 ** -3.95 * .33 .50 (1.42) (1.62) (1.10) (1.18) Notes: All models control for age, race/ethnicity, one-parent household, socioeconomic status, parental attachment, school connectedness, religiosity, ideal script length, relationship hazard, and lagged dependent variable. Standard errors are in parentheses. N = 2,905. *p<.05, **p<.01, ***p<.001 (two-tailed tests)

171

Table E.7. Logistic Regressions of Boys’ Suicide Ideation and Suicide Attempt. Suicide Ideation Suicide Attempt Model 1 Model 2 Model 3 Model 4 Relationship Inauthenticity 2nd Quartile .41 .38 .13 .08 (.25) (.25) (.63) (.67)

3rd Quartile .13 .04 .56 .33 (.31) (.30) (.59) (.61)

4th Quartile .17 -.02 .83 .30 (.26) (.32) (.63) (.71)

Relationship Script Length .00 -.04 (.04) (.09)

Time to Relationship .00 .01 (.01) (.02)

“Special Romantic” Relationship -.16 -.04 (.40) (.80)

Older Partner -.04 -1.66 (.45) (1.09)

Younger Partner .70 ** .67 (.23) (.58)

Ongoing Relationship -.39 -.76 (.26) (.54)

Sexual Intercourse -.09 .73 (.25) (.49)

Psychological Abuse -.25 .03 (.31) (.50)

Physical Abuse .21 -.33 (.40) (.64)

Intercept -2.03 -.35 -1.08 1.84 (1.61) (1.66) (3.17) (3.71)

Notes: All models control for age, race/ethnicity, one-parent parent household, socioeconomic status, parental attachment, school connectedness, religiosity, ideal script length, relationship hazard, prior depression, binary variables indicating whether a friend/family member attempted suicide and friend/family member committed suicide in past 12 months, and lagged dependent variable. Standard errors are in parentheses. N = 2,905. *p<.05, **p<.01, ***p<.001 (two-tailed tests)

172

Appendix F: Figure from Chapter 4

173

Figure F.1. Predicted Probability of Girls’ Poor Mental Health by Romantic Relationship Inauthenticity Quartiles.

.45 .40 .35 .30 .25 .20 .15

Predicted Probability Predicted .10 .05

.00 Severe High Somatic Suicide Suicide Attemptd Depressiona Symptomsb Ideationc

Romantic Relationship Authenticity Quartiles 1st Quartile 2nd Quartile 3rd Quartile 4th Quartile

Notes: All control variables are held at their grand means. aBased on Model 2 in Table E.3. bBased on Model 4 in Table E.3. cBased on Model 2 in Table E.4. dBased on Model 4 in Table E.4.

174