Family Structure, Social Capital, and Educational Outcomes in Two-Parent Families

Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2012 Family Structure, Social Capital, and Educational Outcomes in Two Parent Families Chelsea Lynn Garneau

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] THE FLORIDA STATE UNIVERSITY

COLLEGE OF HUMAN SCIENCES

FAMILY STRUCTURE, SOCIAL CAPITAL, AND EDUCATIONAL OUTCOMES IN TWO-

PARENT FAMILIES

CHELSEA LYNN GARNEAU

A Dissertation submitted to the Department of Family and Child Sciences in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Degree Awarded: Summer Semester, 2012 Chelsea Lynn Garneau defended this dissertation on June 13, 2012.

The members of the supervisory committee were:

B. Kay Pasley Professor Directing Dissertation

Kathryn Tillman University Representative

Frank D. Fincham Committee Member

Lenore McWey Committee Member

The Graduate School has verified and approved the above-named committee members, and certifies that the dissertation has been approved in accordance with university requirements.

For my loving friends and family, whose support and patience keep me going through the most trying times.

iii

ACKNOWLEDGEMENTS

Many people made it possible for me to get to this point through providing instrumental and emotional support when it was most needed. First, I must thank my major professor, Dr. Kay Pasley, who has provided guidance, nurtured my development as a researcher and professional, and gone out of her way to do so over the past three years. As her mentee, I have been encouraged to aim high and work hard to achieve my goals, and she has always been there to support me through the accompanying bouts of anxiety. I would also like to thank the rest of my committee, Dr. Frank Fincham, Dr. Lenore McWey, and Dr. Kathryn Tillman, for their time, patience, and flexibility throughout the past several years. They each brought unique strengths and experiences which improved my research, and I learned much from them along the way. Dr. Isaac Washburn served as my statistics consultant for my dissertation, and it is difficult to adequately express my gratitude for his time, effort, and support throughout this process. He made himself available almost any time I was in need of assistance, and I learned more from him in a few short months than in any coursework I could have taken. Also, thank you to Steve McClaskie, research associate for NLS, for your unsurpassed response time to my emails and your help in understanding and accessing NLSY97 data. I must also acknowledge my cohort and other classmates for their support and encouragement for both my career aspirations and need for self-care. My parents, siblings, and other family members also provided support and encouragement. Even when they did not always understand what it was I was going through, they were always willing to listen to me talk about my work, and I am grateful for that and many additional sacrifices they have made for me. Finally, thank you to my partner in everything, Seth Rosner, who was truly with me in every experience, for better or for worse. He has taught me to be more patient and believed in me when I doubted myself.

TABLE OF CONTENTS

List of Tables ...... vi Abstract ...... vii 1. INTRODUCTION ...... 1 Study Purpose ...... 3

2. REVIEW OF LITERATURE ...... 4 Theory ...... 4 Social Capital Theory ...... 4 Symbolic Interactionism and Role Identity ...... 6 Family Stress Theory ...... 7 Educational Outcomes ...... 8 Family Structure and Educational Outcomes ...... 10 Indicators of Social Capital and Educational Outcomes ...... 11 Social Capital and Family Structure ...... 13 Additional Factors Affecting Educational Outcomes ...... 14 Hypotheses ...... 15

3. METHODS ...... 16 Data Source ...... 16 Sample ...... 16 Measures ...... 18 Independent Variables ...... 18 Dependent Variables ...... 19 Control Variables ...... 20 Analytic Strategy ...... 22

4. STUDY RESULTS ...... 31 5. DISCUSSION ...... 36

APPENDICES A Tables and Figures ...... 44 B Social Capital Items ...... 62 C Human Subjects Approval Letter ...... 64

REFERENCES ...... 67

BIOGRAPHICAL SKETCH ...... 75

LIST OF TABLES

Table 1. Youth Demographic Characteristics ...... 45

Table 2. Bivariate Correlations of Study Variables ...... 46

Table 3. Results of Confirmatory Factor Analysis: Social Capital Items ...... 48

Table 4. Summary of Fit Indices for One-, Two-, Three-, and Second-order Factor Models of Social Capital ...... 49

Table 5. Crosstabulations by Gender, race, and Educational Outcomes for Full Sample and Family Structure Type ...... 51

Table 6. Analysis of Variance by Family Structure Type ...... 52

Table 7. Random-Intercept Logistic Regressions for Educational Outcomes – Full Sample ...... 52

Table 8. Random-Intercept Linear Regressions for Educational Outcomes – Full Sample ...... 55

Table 9. Random-Intercept Logistic Regressions for Educational Outcomes – Biological Children Only...... 56

Table 10. Random-Intercept Linear Regressions for Educational Outcomes – Biological Children Only...... 58

Table 11. Random-Intercept Logistic Regressions for Educational Outcomes – Within Blended Families ...... 59

Table 12. Random-Intercept Linear Regressions for Educational Outcomes – Within Blended Families ...... 61

ABSTRACT

This study examined how family structure moderates the relationship between social capital and educational outcomes for those in two-parent families. Social capital and family structure were measured when participants were 12-14 years old, and a variety of educational outcomes were examined from high school dropout and completion through postsecondary degree completion. Social capital was measured as parental monitoring, parent-child relationship quality, and parent- school involvement. Specifically, this study compared those in stepfamilies to those in intact two-parent families. It also examined differences between mutual biological children in blended stepfamilies and biological children in intact two-parent families, as well as mutual children in blended stepfamilies and stepchildren in blended stepfamilies. Results indicated that those in simple stepfather and blended stepfamilies had poorer educational outcomes overall compared to those in intact two-parent families. Having a higher quality parent-child relationship was associated with greater likelihood of completing a postsecondary degree, greater total years of school completed, and greater highest degree completed. Findings supporting the moderation hypothesis were few. Living in a blended stepfamily moderated the relationships between parent-child relationship quality and high school completion and parent-school involvement and completing a bachelor’s degree or higher. In both instances, the relationship was stronger for those in intact two-parent families than those in blended stepfamilies. Among children living with both biological parents, mutual children in blended stepfamilies were less likely to complete a postsecondary degree and had a lower highest degree completed than those in intact two-parent families. No significant differences were found for the educational outcomes of mutual children and stepchildren living in blended stepfamilies. Implications for future research and policy are discussed.

vii

CHAPTER ONE

INTRODUCTION

High school dropout and educational attainment are widely recognized issues in American education. Estimates from 2010 show that 13% of adults 25 years and older had not obtained a high school diploma/GED, 31% had only a high school diploma/GED, 16.8% had attended some college, and only 30% completed a bachelor’s degree or higher (U.S. Census Bureau, 2010). Importantly, educational attainment has lasting effects on the future success of both adolescents and young adults (e.g., Day & Newburger, 2002). Americans with a high school diploma earned roughly $7,000 more per year than those with less than a high school education, and having a bachelors’ degree increased one’s earnings by $22,000 per year compared to having only a high school diploma. Educational outcomes are heavily influenced by the nature of parenting for both adolescents (Teachman, Paasch, & Carver, 1996) and young adults (Melby, Fang, Wickrama, Conger, & Conger, 2008), and both educational outcomes and parenting are influenced by family structure. Coming from a non-intact family increases the risk of high school dropout by 67.7% (Rumberger & Thomas, 2000) and decreases expectations to attend college (Heard, 2007). Research suggests that differences in family resources in one- and two-parent families (e.g., financial, parental time, engagement) place children in one-parent families at greater risk for poor academic outcomes (Anguiano, 2004). The risk for children in stepfamilies is often overlooked by policy makers, because they have the advantage of two parents in the household. However, some research shows that children in stepfamilies are at similar risk (Jeynes, 2006) or greater risk for poor outcomes as those from single-parent families (Ham, 2004; Heard, 2007). Especially overlooked are children born to a couple already in a stepfamily who then create a blended family, and many of these “mutual” children are born within two years after remarriage (Downs, 2004; Stewart, 2002). The few studies examining these children show that they, too, are at more risk for poor outcomes than are biological children in two-parent families where no stepchildren reside (Tillman, 2008). In studies, these mutual children are often misclassified as coming from intact families, because family structure is typically determined by the child’s relationship to his/her resident parents (Gennetian, 2005). Further, children who experience multiple transitions (e.g., parental divorce and subsequent remarriage), a precursor to living in a

1 stepfamily in many cases, are at risk for high school dropout (Pong & Ju, 2000), and evidence shows that they are at increased risk for less academic success and lower expectations for college (Tillman, 2008). Given these additional risks for poor outcomes of these children in stepfamilies, the current study examines two-parent families exclusively to focus on the unique risks that children in stepfamilies face which are often overshadowed in studies of both one- and two-parent families. In previous research, differences in parenting failed to explain why children from non- intact families experience a higher risk of dropout (Astone & McLanahan, 1991); however, several important parenting behaviors known to affect educational outcomes and that differ by family structure were not examined, specifically monitoring, parent-child relationship quality, and parent-school involvement. These three aspects of parenting are believed to contribute to children’s social capital (Dika & Singh, 2002). Social capital refers to the social relationships in children’s lives which, in this case, facilitate access to the benefits of parents’ human capital (e.g., their education, knowledge, and skills) (Coleman, 1988). Each of these aspects of social capital has been associated with differences in educational outcomes (e.g., Manning & Lamb, 2003; Woolley & Grogan-Kaylor, 2006) and are, therefore, important to examine in research addressing the relationship between family structure and educational outcomes. The study undertaken here contributes to the literature in three ways. First, few studies have examined how indicators of family structure influence the links between children’s social capital and their educational outcomes. Such information can provide insights into which aspects of social capital are most important for educational intervention. Second, because educational outcomes are typically measured as a single event (e.g., high school dropout, entering college) or in total number of years attained (Gordon & Cui, in press), I examined a number of educational outcomes to determine which are best explained by these indicators of social capital and family structure. For example, educational outcomes included continuous high school dropout versus dropout and subsequent enrollment, total years completed, degree completion including high school diploma or GED, completing some college versus completing a post-secondary degree, total years of education completed, and the highest degree completed. Where the results converge or diverge can be used to promote more effective intervention through the implementation of policies and practices that target adolescents and young adults differentially at points when they are at greatest risk for dropping out or discontinuing their educational pursuits.

Third, examining the combined effects of stepchild status and family structure can provide insights for intervention that targets those in two-parent families who have been ignored in prior research and who have the greatest risk for early dropout and not achieving their academic potential.

Study Purpose

The study had two aims. One aim was to examine both individual- and family-level characteristics, family structure, and parenting behaviors (as indicators of social capital) to determine which factors place adolescents at greatest risk for achieving less education by comparing those in stepfamilies and those in intact two-parent families. Another aim was to test a conceptual model that addressed how family structure influences the effects of children’s social capital, as reflected in parenting behaviors, on various indicators of educational outcomes beyond the effects of the individual- and family-level characteristics. Specifically, I addressed three questions: 1. How do family-level structure and stepchild status influence these educational outcomes (e.g., Are biological children in blended families at greater risk for early exit compared with biological children in intact families?)? 2. How do the effects of social capital (parental monitoring, parent-child relationship quality, and parent-school involvement) on the various educational outcomes differ by family structure and stepchild status? 3. What is the relative effect of social capital and family structure on various educational outcomes?

CHAPTER TWO

REVIEW OF LITERATURE

Here I present three theoretical frameworks underlying both the examination of differences in parenting and parent-child relationships among families of differing structures and the link between parenting behaviors and educational outcomes. Discussions of social capital theory, symbolic interactionism and role identity, and stress theory are presented and followed by findings from the extant literature used to develop hypotheses and the proposed model.

Theoretical Background

The relationship between social capital and educational outcomes is best understood by examining some of the main propositions of Coleman’s (1988) social capital theory. Additionally, symbolic interactionism and role identity are used to explain why social capital is expected to vary among children from different family structures and why the relationship between social capital and educational outcomes is influenced by family structure. Finally, I draw from Boss’ (2002) family stress theory to explain why biological children in blended families specifically are likely to experience decreased social capital compared to biological children in intact two-parent families.

Social Capital Theory

Coleman’s (1988) theory of social capital is used to explain how parents contribute to children. Specific to this study, it is used to explain how parents contribute to children’s educational outcomes. Emerging from classical capital theory (Brewer, 1984) and theories of human capital (Becker, 1964/1993), social capital theory answered the desire to explain the value and benefits inherent in social relationships. Because resources in social networks are gained through participation in relational processes (Lin, 2001), theoretically children benefit from parents’ social networks and from their relations with their parents. Of particular importance to educational outcomes are parental monitoring, the quality of the parent-child relationship, and parent-school involvement. Parental monitoring and the quality of the parent-child relationships reflect the nature of children’s relations with their parents, whereas parent-school involvement reflects the parents’ social network in their interaction with school-related personnel and

4 activities. Thus, these are all indicators of children’s social capital which builds when they are present in the family. Coleman (1988) argued that without strong social capital children cannot access the benefits of their parents’ human capital (parents’ particular knowledge and skills) to improve their educational outcomes. Social capital increases as parents and children make investments in various aspects of their social relationship, and the amount of investment is associated with the expected returns from that relationship (Lin, 2001). Expected differences in social capital for children living with a stepparent and those living with both biological parents can be extrapolated from this theory combined with components of symbolic interactionism and role identity theory as discussed in the next section. Briefly, the expected returns on relational investments with non-biological parents or children may be less than those expected in biological relationships. Research findings consistently support this idea (Dunn, Davies, O’Conner, & Sturgess, 2000; Fisher, Leve, O’Leary, & Leve, 2003; Hetherington & Stanley-Hagan, 2000). Applied here, family structure is expected to moderate the relationship between social capital and educational outcomes, such that the relationship will be weaker for those in a stepfamily. Social capital theory also emphasizes the benefit of a closed versus an open intergenerational social network for children (Coleman, 1988). In an open intergenerational network parents and children are connected through their family relationships, but their networks diverge outside of the home. Thus, parents in an open network are not well connected with their children’s friends, friends’ parents, and teachers. In contrast, parents in a closed intergenerational network would know their children’s friends, are friends with their parents, and form a relationship with other important individuals in their children’s lives such as teachers and other school personnel. Theoretically, closed networks allow for closer monitoring of children and guiding of their behaviors. Applied here, a closed intergenerational network is represented by greater parental monitoring and more parent-school involvement. Moreover, parents’ level of education, one aspect of human capital in families, is widely recognized as an important factor in children’s academic performance, educational aspirations, and attainment (Anguiano, 2004; Henry, Plunkett, & Sands, 2011; Wojtkiewics & Holtzman, 2011). However, social capital theory (Coleman, 1988) posits that parents’ own educational background alone is insufficient in explaining differences in child outcomes. Strong social capital, such as a high quality parent-child relationship and frequent parental involvement in

5 children’s schooling, is needed to confer the benefits gained from a more educated parent. In fact, Coleman suggested that social capital may be more important for child outcomes than human capital, such that children with parents who have less education but who put more effort into teaching and guiding them will experience better outcomes than will children with more educated and less involved parents. It follows that strong social capital may also be affected by other parent and family factors which are known to negatively affect educational outcomes, such as experiencing family structure transitions and living in complicated structures. As such, social capital is likely less present among children in non-intact families and those living with a stepparent. Symbolic Interactionism and Role Identity Symbolic interactionism (LaRossa & Reitzes, 1993) is used here to explain why parenting behaviors may vary between biological parent-child and stepparent-stepchild relationships. The meanings attributed to relationships and interactions among individuals are the central focus of symbolic interactionism (LaRossa & Reitzes, 1993). An assumption of this theory is that individuals’ actions are driven by meanings, and meanings are both influenced and created through interactions with others. Through social interactions, members of groups or families create shared meaning of the roles they hold within the group (Stryker, 1968), and role relationships come together to create an individual’s identities. Some of the role relationships that occur in families include husband-wife, parent-child, and stepparent-stepchild. Attached to these role relationships are the concepts of role salience and centrality, or how meaningful and important one’s roles are to the overall sense of self (Stryker, 1987). The roles of biological parent and biological child, for example, may carry more meaning than those of stepparent and stepchild, because they often had more time to develop (Heatherington & Kelly, 2004; Robertson, 2008), and the expectations of those roles are more clear and socially proscribed (Hetherington & Staley-Hagan, 2000). Less social capital may develop in stepparent-stepchild relationships compared to biological parent-child relationships, because both parties experience their roles in the relationship as less central or important compared to roles that are biologically based and of longer duration (e.g., mother-child, father-child). Role salience contributes to our understanding of a father’s investment in children, wherein those for whom the father role is more salient are more committed to the father-child relationship and, therefore, invest more resources (e.g., time, affection, money) in their children (Fox & Bruce, 2004).

Importantly, the concept of role salience has been extrapolated to create the concept of relationship salience (Umberson & Chen, 1994). Regardless of whether a relationship is negative or positive, those that are more salient to an individual theoretically will have more impact on his or her well-being. Thus, although social capital may be lower for children in stepfamilies, the concept of relationship salience leads to the hypothesis that the effect of social capital on educational outcomes may also be weaker for stepchildren compared to biological children. Family Stress Theory Family stress theory (Boss, 2002) is used to explain why mutual biological children in stepfamilies may have poorer social capital and educational outcomes than biological children in intact, first-marriage families. According to this theory, when families are not able to adequately cope with the stressors they experience, such stressors can negatively affect family members’ well-being. During the early years of stepfamily formation, stress is often high as family members attempt to define their roles, blend family cultures, and establish new relationships (Pasley & Garneau, 2012). Coupled with unrealistic expectations for their new family, members are unprepared to manage these common stressors. The increased stress may take a toll on parenting behaviors, as findings show that couples in newer stepfamilies report lower parenting competency (Hoffman & Johnson, 1998). It is also during these early years that mutual children are likely born into stepfamilies (Coleman & Ganong, 2000), adding to other stressors. Increased family stress and decreased parenting competencies may influence the early development of the parent-child relationship for these mutual children. It may also set up these mutual biological children in blended families for poorer educational outcomes, as experiencing stress early, particularly prenatal stress, can negatively impact later cognitive functioning (Lupien, McEwen, Gunnar, & Heim, 2009). Finally, evidence shows that family stress may continue to be greater in stepfamilies as children age, with adolescents reporting more family conflict in stepfamilies compared to adolescents in both intact and single-parent families (Kurdek & Fine, 1993). Other research shows higher levels of interparental conflict in stepfamilies, which translates into more hostile/rejecting parenting and poorer child outcomes (Shelton, Walters, & Harold, 2008). With greater conflict and stress in stepfamilies, parents may engage in fewer monitoring behaviors, have more strained parent-child relationships, and be less involved in adolescent’s schooling.

Educational Outcomes Educational outcomes have a lasting and significant effect on the quality of life well into adulthood. For example, level of education completed is commonly included in research across disciplines, because of its known effects on physical health outcomes (Mezuk, Eaton, Golden, & Ding, 2002), mental health outcomes (Herzog, Markus, Franks & Holmberg, 1998), later parenting (Hofferth & Anderson, 2003) and co-parenting behaviors (Stright & Bales, 2003), marital stability (Heaton, 2002), marital adjustment (Flouri & Buchanan, 2002), financial success (U.S. Census Bureau, 2010), and productivity in later adulthood (Herzog, Markus, Franks, & Holmberg, 1998). Although educational attainment has been recognized as an important focus of research, much of the research examining the educational attainment of adolescents and young adults emphasizes predicting one outcome, such as high school graduation (e.g., Anguiano, 2004; Ream & Rumberger, 2008), entrance into post-secondary schooling (Brooks-Gunn, Guo, & Furstenberg, 1993; Orthner, Jones-Sanpei, Hair, Moore, Day, & Kaye, 2009), or total number of years of education by a specified age (e.g., Gordon & Cui, in press). Studies which looked more deeply at the influences of complicated family structures or children’s social capital have examined intentions or expectations to attend college (Tillman, 2007) and recognize the need to extend such findings to post-secondary experiences using longitudinal data (Astone & McLanahan, 1991). Perna and Titus (2005) examined the influence of social capital on the likelihood of enrollment in 2- and 4-year postsecondary education programs, yet they did not account for any measure of family structure. Measuring attainment at one point or in terms of expectations is a weakness in this research, because findings fail to highlight how the influence social capital and family structure during adolescence varies depending upon the specific outcome and timing of students’ greatest risk for exiting from education. Thus, the current study examines a sequence of educational outcomes to determine points of greatest risk for exit and to capture a more thorough understanding of how and when family structure and social capital influence education. Dropping out of school becomes a concern in many states when adolescents turn 16 years old and are legally able to discontinue their education (Education Commission of the States, 2010). According to national census data, the majority of students who eventually dropout of high school do so after completing the 11th grade (U.S. Census Bureau, 2011); this includes students from states requiring enrollment until age 18. In a study of high school dropout rates in

North Carolina, where students may legally dropout at age 16, the majority of students who eventually dropped out did so by the end of 9th grade (Stearns & Glennie, 2006). However, each year a student remains enrolled has marked benefits in terms of financial success (e.g., Day & Newburger, 2002). Thus, early exit from high school (prior to graduation) becomes a risk factor for this and other associated poor outcomes later in adulthood. A thorough examination of the impact of social capital and family structure on education must also consider educational behaviors beyond high school completion. Successful completion of high school, though not always a requirement, is a precursor to college attendance for most individuals. Young adults who have dropped out of high school prior to graduation, even when they return or obtain equivalency degrees, are less likely to be employed or attend post-secondary school two years later (Rumberger & Lamb, 2003). Additionally, although roughly 68.1% of high school graduates enter post-secondary education (Bureau of Labor Statistics, 2011), many fall short of degree completion, with only 52.3% of students who enter four-year colleges graduating with a degree within five years (ACT, 2010). Thus, the current study adds to the literature by examining the educational attainment of individuals using a variety of measures of educational outcomes, including continuous high school dropout versus dropout and subsequent enrollment, total years completed, degree completion including high school diploma or GED, completing some college versus completing a post-secondary degree, and the timing of risk for exit from the educational system. A strength of recent research is the emphasis on (a) the cumulative influence of risk factors that precede high school dropout (Alexander, Entwisle, & Kabbani, 2001; Hickman, Bartholomew, & Mathwig, 2008; Jimmerson et al., 2000) and (b) the influence of children’s social capital on educational attainment (Englund, Egeland, & Collins, 2008). For example, researchers found that for many children the trajectory toward high school dropout begins with the lack of available resources and presence of other risk factors (e.g., nontraditional family structures, structure transitions, grade retention, low academic achievement, enrollment in special education) during early childhood that are compounded later (Alexander, Entwisle, & Kabbani, 2001). Hickman and colleagues (2008) also found that trends in differences between dropouts and graduates in their academic performance and absenteeism clearly emerge as early as kindergarten.

Recent research is lacking, however, in two important ways. First, although the impact of family structure and social capital have been extensively examined as independently influential factors of educational outcomes, the literature has failed to account for the effect of family structure on the relationship between social capital and educational outcomes. Second, although individual studies have emphasized high school completion, entrance into post-secondary schooling, or total number of years of education completed, they lack a comprehensive examination of how social capital differentially influences these and other educational outcomes. For example, it may be that social capital is more influential in determining high school completion than college attendance and completion. Examining various indicators of education outcomes can provide more insight into who is at greatest risk, for which outcome, and when. Family Structure and Educational Outcomes Importantly, educational outcomes have been consistently linked to family structure. Compared to intact biological families, children from all other structures are less likely to graduate from high school (Anguiano, 2004; Rumberger & Thomas, 2000). Although most studies compare those from intact families to all other structures, Manning and Lamb (2003) showed specifically that children in stepfamilies were at greater risk than those in intact and single-parent families for poorer GPA and more frequent school suspension, factors known to influence dropout (Christle, Jolivette, & Nelson, 2007; Hickman, Bartholomew, Mathwig, & Heinrich, 2008). Beyond examining basic categories of family structure, findings also show that experiencing parental divorce and multiple family structure transitions, common precursors to stepfamily life, also increase adolescents’ risk for dropout (Cavanagh, Schiller, & Riegle-Crumb, 2006) and decrease their expectations to attend college (Tillman, 2007). Compared to all other family types, those in stepfamilies are the least likely to attend post-secondary schooling (Sandefur, Meier, & Campbell, 2006). Other findings show that children raised by biological mothers in stepfamilies attained more education overall than did stepchildren in these families (Case, Lin, & McLanahan, 2001) and less education than biological children in first-marriage families (Ginther & Pollak, 2004). These findings highlight the importance of considering both children’s overall family structure and individual stepchild status within the family when examining family structure. Using both of these measures of family structure, categories of family structure include intact biological (biological parents and only biological siblings),

10 simple stepfather (biological mother, stepfather, and only biological siblings/children of the mother), simple stepmother (biological father, stepmother, and only biological siblings/children of the father), complex stepfamily (one biological parent, one stepparent, and at least one stepsibling), and blended stepfamily (either both biological parents or one biological parent and a stepparent and at least one half sibling). Applying these measure of family structures, I propose that mutual biological children (who are half-siblings to other children) in blended stepfamilies will have poorer educational outcomes compared to biological children in intact families. Also, within blended families, I expect that stepchildren will have poorer educational outcomes compared to mutual biological children. Indicators of Social Capital and Educational Outcomes Although findings show family structure to be an important factor in children’s long-term educational outcomes, no study to date has examined how family structure affects the link between social capital and educational outcomes. Studies show that several family processes representing components of children’s social capital significantly affect their later educational outcomes (Anguiano, 2004; Stokes, 2008; Teachman, Paasch, & Carver, 2008). Specifically, parental monitoring, parent-child relationship quality, and parent-school involvement (components of social capital) are three of the strongest predictors of educational outcomes. Parental monitoring commonly refers to parents’ knowledge and tracking of children’s activities when they are not at home (Stattin & Kerr, 2000). Research examining parental monitoring has used children’s reports of how much their parents know about their friends, where they are when they are not home, and who they spend time with outside of the home (Fletcher et al., 1995). Interestingly, adolescents who report more monitoring also have lower expectations to attend college, although effect sizes are small so these possible negative effects of monitoring are not considered strong (Manning & Lamb, 2003). More consistently, findings show a positive relationship between monitoring and education. Both mothers’ and fathers’ monitoring was positively related with grades in high school, and this relationship was mediated by adolescent’s increased academic engagement (Plunkett, Behnke, Sands, & Choi, 2009). Woolley and Bowen (2007) examined the influence of several monitoring behaviors, such as “adults monitor whereabouts” and “adults in home know friends,” in a scale measuring overall social capital assets. Results were that greater social capital was a significant predictor of school engagement, beyond the effects of other aspects of risk exposure (e.g., grade retention, low

11 socioeconomic status, unsafe neighborhoods, and delinquent peers); however, it is unclear how much monitoring alone affected these outcomes. Importantly, inconsistencies in these findings may result from changes in causal order. For example, it may be that when greater monitoring is associated with poorer educational outcomes the parents increased their monitoring in response to poor grades or lessening academic engagement. In the current study, I include controls for prior grade retention, and I expect that monitoring will be positively associated with educational outcomes, such that higher levels of monitoring will be linked with later exit from education and, thus, more positive educational outcomes. Parent-child relationship quality is a second aspect of social capital frequently associated with educational attainment. Quality of parent-child relationships is examined in the current literature in terms of several components of the relationship, including closeness (Manning & Lamb, 2003), warmth (Henderson & Taylor, 1999; Hetherington & Jodl, 1994), positivity and negativity (Dunn, Davies, O’Connor, & Sturgess, 2000), and children’s reports of parents’ supportiveness (Henderson & Taylor, 1999; Henry, Plunkett, & Sands, 2011; Hetherington & Stanley-Hagan, 2000). Higher quality parent-child relationships, as measured by parental support and positivity, with even one parent increase the likelihood of high school completion and enrolling in post-secondary school among adolescents (Orthner et al., 2009). When students rate parents as more supportive, they also tend to have greater academic motivation (Henry, Plunkett, & Sands, 2011). In a group of low-income students considered at high risk for high school dropout, a supportive parent-child relationship mitigated risk and increased the likelihood of graduation (Englund, Egeland, & Collins, 2008). Thus, I expect that when students report that their relationship with their parents is of higher quality, they will likely exit from education later and have more positive educational outcomes. Parent-school involvement is also an important predictor of educational attainment and is another indicator of social capital. School-related support is associated with less frequent dropout and more years of education (Flouri, 2006). Also, children with parents who were more engaged in various aspects of their schooling, such as selection of courses, homework, and showing interest in school activities, showed greater academic achievement than those whose parents were less involved (Woolley & Grogan-Kaylor, 2006). Anguiano (2004) found that when parents had greater contact with the school and school personnel, attended parent-teacher meetings, and attended children’s school-related activities, the risk of high school dropout

12 decreased. The likelihood of enrolling in post-secondary education is also positively associated with how often parents volunteered at their children’s school (Perma & Titus, 2005). Englund and colleagues (2008) measured parent-school involvement by asking teachers about parents’ concern regarding children’s school and attendance at school meetings and conferences and found that low-income students with greater parent-school involvement were more likely to graduate from high school. Even having parents who are simply more interested in their children’s schooling increased the likelihood of greater educational attainment into young adulthood, without accounting for actual involvement (Flouri, 2006). Thus, in the current study, I expect that when parents are more involved with their children’s school, the children will achieve more positive educational outcomes. Social Capital and Family Structure Social capital is strongly related to educational attainment; yet, several components of children’s social capital vary by family structure. For example, parents in intact families engage in more monitoring than do those in stepfather families (Fisher, Leve, O’Leary, & Leve, 2003). Stepparents also are less engaged with stepchildren than are parents in intact families, and when both stepchildren and biological children are present in the same household (blended families), fathers engage more with their biological children than their stepchildren (Hofferth & Anderson, 2003). Family structure also affects the quality of parent- or stepparent-child relationships, such that parent-child relationships are characterized by more positivity, less negativity (Dunn, Davies, O’Connor, & Sturgess, 2000), more support (Hetherington & Stanley-Hagan, 2000), and greater warmth (Hetherington & Jodl, 1994) than stepparent-stepchild relationships. Interestingly, Henderson and Taylor (1999) found that biological relatedness was more important than overall family structure in determining mother-child relationship quality; mothers were more supportive and warm toward their biological children regardless of family structure. For fathers, however, monitoring was more frequent among biological fathers in intact families than biological fathers in stepfamilies. Common links exist among family structure, social capital and educational attainment in the extant literature. It may be that social capital mediates the relationship between family structure and educational attainment, wherein lower social capital in stepfamilies is the reason for poorer educational outcomes. However, the link between social capital and education has been consistently demonstrated regardless of family structure. Unique to this study, I examine

13 the possibility that the effects of social capital on educational outcomes vary by family structure. Thus, based on the concepts of relationship and role salience in symbolic interactionism, I expect family structure to moderate the relationship between social capital and educational outcomes, such that the relationship will be weaker for those in each type of stepfamily compared to those in intact families. Also, I hypothesize that stepchild status in blended stepfamilies will moderate the relationship between social capital and educational outcomes, such that the relationship between social capital and educational outcomes will be weaker for stepchildren in these families compared with mutual biological children in them. Additional Factors Affecting Educational Outcomes Beyond family structure and social capital, additional variables influence educational outcomes. At the individual level, increased risk of high school dropout is associated with poor academic achievement (Battin-Pearson, Newcomb, Abbott, Hill, Catalono, & Hawkins, 2000), a greater number of school changes (Swanson & Schneider, 1999), engagement in delinquent activities (Hickman, Bartholomew, Mathwig, & Heinrich, 2008), and early sexual initiation (Sprigs & Halpern, 2008). Other individual factors, such as gender and race, influence dropout; boys are more likely to dropout than are girls (Heckman & LaFontaine, 2007), and Blacks are more likely to drop out than Whites, whereas Hispanics are the most likely to drop out (U.S. Census Bureau, 2010). Rojewski (1999) also found that students who were diagnosed with mild to moderate learning disabilities attain less education than students without disabilities. Further, the influence of social capital on students’ educational attainment may also vary by gender. Flouri (2006) found that the positive relationship between parents’ interest in children’s schooling and their overall educational attainment was stronger for girls than it was for boys. Influential family-level and school factors are also noteworthy. Family-level factors include parents’ education and number of siblings. For example, parents’ educational attainment was the strongest predictor of children’s high school dropout (Janosz, Leblanc, Boulerice, & Tremblay, 1997), and more children in a family increase the risk of attaining less education (Booth & Kee, 2009). At the school level, those who attend schools which are publicly funded are more likely to drop out than those in private schools (Rumberger & Thomas, 2000).

Hypotheses

Guided primarily by social capital theory, I examined how family structure influences the relationship between children’s social capital and a variety of measures of educational outcomes beyond the effects of family- and individual-level characteristics. Specifically, I test the following hypotheses:

1. Participants living in each type of stepfamily (simple stepfather, complex, and blended) during adolescence will have poorer outcomes on all educational behaviors compared to those from intact two-parent families. 2. Participants with weaker social capital (i.e., lower parental monitoring, lower quality parent-child relationship, and less parent-school involvement) during adolescence will have poorer outcomes on all educational behaviors compared to those with stronger social capital. 3. Family structure will moderate the relationship between social capital (i.e., parental monitoring, parent-child relationship quality, and parent-school involvement) and educational behaviors, such that the relationship will be weaker for those in each type of stepfamily (simple stepfather, complex, and blended) compared to those in intact two- parent families. 4. Mutual biological children in blended stepfamilies will have poorer outcomes on all educational behaviors compared to biological children in intact two-parent families. 5. Within blended families, stepchildren will have poorer outcomes on all educational behaviors compared to mutual biological children. 6. Stepchild status in blended stepfamilies will moderate the relationship between social capital (i.e., parental monitoring, parent-child relationship quality, and parent-school involvement) and educational behaviors, such that the relationship between social capital and educational outcomes will be weaker for stepchildren compared with mutual biological children in these families.

CHAPTER THREE

METHODS

Data Source Data are from the National Longitudinal Survey of Youth 1997 (NLSY97), a nationally representative sample of about 9,000 youth between 12 and 16 years old at the end of 1996. Data collection began in 1997 (Round 1) and continued annually. Data are currently available through 2009 (Round 13), when respondents were 24 to 30 years old. Households in this study were randomly selected and include multiple youth respondents from the same household. A parent or guardian of each participant, most often the biological mother when available, provided information on family demographic characteristics, parents' work history, family history, children’s health and schooling history, and family processes through the Parent Questionnaire. Each adolescent participant provided information via the Youth Questionnaire, including schooling history, family processes, substance use and delinquency, dating and sexual history, and attitudes and future expectations. Detailed schooling information included performance on standardized tests and was collected in the School Survey; these data are used here. This data set was selected for use because the majority of the sample entered the study prior to the age at which youth begin to exit the educational system (Education Commission of the States, 2010) and were followed to an age when they had the opportunity to complete a four-year post- secondary degree, capturing a period in which the majority of participants will have exited education, and, thus, their educational outcomes are known. Sample Roughly 9,000 youth were sampled in the NLSY97 during Round 1. Because the aim of this study is to examine a variety of educational outcomes into young adulthood, data from all available rounds were used (1-13). The current study focuses on youth living in two-parent families and with at least one biological parent during adolescence (N = 5,604), so the sample was limited to these youth at Round 1. Included in this number are five participants who were missing reports of their biological relationship to parents, but they had consistent reports of family structure at prior and subsequent time points. Thus, later family structure data were confidently used as a substitute. Excluded were participants with missing and inconsistent family structure data (N = 14). Finally, three participants reported living with both biological parents and also having a stepsibling (who was thus not biologically related to either resident

16 parent) in the home. Because this structure type was a rare-occurring case, these participants were excluded. Of the remaining 5,608 participants, several additional exclusion criteria were applied. To be included, social capital measures were used from Round 1 when participants were between 12 and 14 years old, thus excluding 2,173 participants outside of this age range. Also, findings show that students with mental disabilities, excluding those with “learning disabilities” and “behavioral disabilities,” have poorer educational outcomes, and only 30% go onto complete any post-secondary education or job training (Carson, Frank, & Sitlington, 1992) compared to the almost 70% of high school graduates overall (Bureau of Labor Statistics, 2011). An additional seven participants were excluded, because they were identified as having mental retardation. Finally, two other exclusions were made. Participants who reported being homeschooled in grades K – 12 were excluded (n = 62). Also, those in simple stepmother families (n = 32) were excluded from the analysis due to inadequate sample size. This decision was made after a power analysis was conducted to determine the required group size to detect a small effect size (d = .30) at α = .05 to achieve a power of .80. A small effect size was used based on a comparison of means between the stepmother group and the referent group on several measured outcomes. Results indicated that with a ratio of 83:1, the required size of the simple stepmother group was 73 participants. Demographic information for the final sample (N = 3,334, Round 1) indicated that 46.8% were female, 53.2% were male, and the mean age of participants was 13.98 years (SD = .82; see Table 1). The majority identified as White (58.3%), followed by Hispanic (22.0%), Black (16.4%), and Other (3.3%), including Asian/Pacific Islander, American Indian/Eskimo, or mixed race. The greatest percentage (30.3%) were from families with annual household incomes ranging from $30,000 to $60,000 (M = $57,302; SD = $43,538). Average education attainment was 12.7 (SD = 3.1) years for resident mothers and 12.8 (SD = 3.3) years for resident fathers. The final sample included 2,650 living with both biological parents, 591 living with a biological mother and stepfather, and 93 with a biological father and stepmother. Sibling relationships were then examined to determine the overall structure of participants' families with 2,549 (76.4%) in intact families, 284 (8.5%) in simple stepfather families, 88 (2.6%) in complex stepfamilies, and 413 (12.3%) in blended stepfamilies.

Demographic information was also examined to determine the effect of the exclusion criteria on the representativeness of the final sample. Results from independent sample t-tests indicated that those who were excluded did not differ significantly from those included in the final sample in terms of age, gender, race, household income, and average education attainment by either resident father or mothers.

Measures

Independent Variables

All independent variables were measured at Round 1 when youth were 12-14 years old. Family structure. The complicated nature of family structure was a focus of this study, and so family structure was measured using two variables. First, participants’ relationship to parents was reported, and this measure represented participants’ biological relationship to parents for individual-level structure. Next, this was combined with household roster reports of the types of siblings living together to create a variable for family-level structure. The resulting categories include intact biological (biological parents and only biological siblings), simple stepfather (biological mother, stepfather, and only biological siblings/children of the mother), complex stepfamily (one biological parent, one stepparent, and at least one stepsibling), and blended stepfamily (either both biological parents or one biological parent and a stepparent and at least one half sibling). This allowed mutual children born into a stepfamily with half-siblings to be classified correctly into blended families rather than misclassified into intact families. Additionally, to account for the presence of both mutual biological children and stepchildren in blended families, a dummy variable was created for stepchild status to indicate biological child (0) and stepchild (1). Social capital. Three components of social capital were examined: parental monitoring, parent-child relationship quality, and parent-school involvement. Items assessing parental monitoring were taken from the extant literature (Hetherington, Cox, & Cox, 1982; Maccoby & Mnookin, 1992). Youth reported parent’s monitoring with four items such as “How much does he/she know about your close friends, that is, who they are?” and “How much does he/she know about your close friends’ parents, that is, who they are?” Reponses ranged from knows nothing (0) to knows everything (4). Scores from the four items for resident step/mother and resident step/father were combined to create an overall parental monitoring scale ranging from 0 to 32

18 with higher scores indicating more parental monitoring. Cronbach’s alpha for parental monitoring in the current sample was .79. The quality of the parent-child relationship was measured using two sets of items adapted from the IOWA Youth and Family Project (IYFP) (Conger & Elder, 1994). In the first three items, youth were asked to report for each resident parent/stepparent on items such as “I think highly of him/her,” and “I really enjoy spending time with him/her.” Responses ranged from strongly disagree (0) to strongly agree (4). A second set of five items assessed youth perceptions of parental support from each resident parent/stepparent. Items included, “How often does s/he help you do things that are important to you?” and “How often does s/he praise you for doing well?” Responses ranged from never (0) to always (4) and were summed for all eight items with a possible total score ranging from 0 to 32 for each resident parent; higher scores indicate a more positive relationship. For the current analyses, step/mothers’ and step/fathers' scores were combined to create one measure of overall relationship quality with resident parent/stepparent, and possible scores range from 0 to 64. Cronbach’s alpha for parent- child relationship quality in the current sample was .84. Parent-school involvement was measured using resident step/parents' reports on two items: how often he/she or his/her spouse attended parent-teacher organization (PTO) meetings, and how often one of the resident parents volunteered to help in the child’s classroom in the past three years. Responses were often (1), sometimes (2), and never (3). Items were reverse coded and summed to create an overall score, with higher scores indicating more involvement. Cronbach’s alpha for parent-school involvement in the current sample was 54. This low reliability estimate is likely affected by having only two items. Findings show that reliability estimates for parent-school involvement measures in large national datasets, including items such as volunteering in the classroom, are stronger with a greater number of items (Perna & Titus, 2005; Stewart, 2007), and such indicators of school involvement were significant predictors of educational outcomes. Restrictions of available items in the current data required that only these two items be used. Dependent Variables Educational outcomes. Several educational outcomes were examined. Dummy variables were created using data across all rounds to indicate whether students had a high school diploma /GED (1 = yes, 0 = no), entered post-secondary schooling (1 = yes, 0 = no), completed

19 any post-secondary schooling (1 = yes, 0 = no), completed a 2-year degree (1 = yes, 0 = no), and completed a 4-year degree (1 = yes, 0 = no). Participants also reported the highest degree completed by Round 13 which is used to indicate overall educational attainment. An additional dummy variable were created for participants who dropped out at any point prior to high school graduation to indicate students who dropped out and returned (1) and those who remained unenrolled (0).

Control Variables

Grade retention. Retention was measured based on whether students had ever repeated a grade, as reported by parents during each round of data collection through the end of high school. These variables were used to create a dummy variable indicating those who have ever repeated a grade (1) and never repeated a grade (0). Delinquency. Delinquency was measured during Round 1 by youth reports on items modified from a measure of delinquency and criminality from the National Youth Survey (NYS). Youth were asked to answer yes (1) or no (0) to whether they had engaged in any of 10 behaviors, such as theft, gang activity, damaging property, selling drugs, and arrests. A total count was made to create a Delinquency Index with possible scores ranging from 0 to 10, and higher scores indicate greater engagement in delinquent activities. Early sexual initiation. Early sexual initiation is typically defined as first sexual intercourse at age 15 or younger (Smith, 1997; Spriggs & Halpern, 2008). During each wave, participants reported age at first sex, with those who had not yet engaged in sexual intercourse labeled as a legitimate skip. These responses were combined into one variable to indicate early sexual initiation (1) for those reporting age 15 and younger and later sexual initiation (0) for those reporting age 16 and older. Learning disability. In Round 1 parents reported whether the youth suffered from a learning disability or attention disorder which limited school work or performance. Those who answered affirmatively, then rated how much the youth was limited by the condition; responses were no not currently limited by this condition (0), yes, limited a little (1), or yes, limited a lot (2). Responses were recoded to indicate no learning disability or has a learning disability but not limited (0), and has learning disability and currently limited (1). Family structure transitions. The household roster used to create overall family structure captured only the presence of siblings living in the identified adolescent’s household 20 during Round 1. Changes in family structure across later rounds of data collection were based only on reports of youths’ relationship to parents in the household. This eliminates the influence of an older sibling leaving the household as a normal transition (e.g., to create an independent residence, attend college). Additionally, when participants turned 18 years old during data collection, they had an increased likelihood of missing data or reporting transition from the previous years’ structure to the anything else category, as their relationship to resident parent figures was based on household rosters. Thus, family structure transitions were calculated using data from Round 2 until the round when participants turned 18 years old (Rounds 4-6). Beginning with relationship to resident parents reported in Round 1, each change in family structure between consecutive rounds was counted as 1 transition. The total number of transitions were then summed to create a continuous count variable of family structure transitions experienced over the duration of the study until age 18 and ranged from 0 to 4. School changes. During each round of data collection, participants reported the number of times they had changed schools during the academic year. Additionally, retrospective reports of school changes were reported going back to 1985, the earliest date at which participants may have begun schooling. The number of school changes were summed for all years to create a count variable of total school changes. Number of siblings. Total number of siblings was based on those under 18 reported to be living in the household at Round 1. Parents’ education and household income. During Round 1, parents reported the highest grade completed by each resident parent. Responses range from first grade (1) to eighth year of college or more (20). Household income was reported as gross household income for the most recent year in number of dollars as reported in Round 1. School sector. Participants reported whether the school they attended at Round 1 was designated as publically funded or privately funded. Responses were dummy coded as either public (0) or private (1). Demographic control variables. Several variables were included here. Gender was reported at Round 1, and a dummy variable was created to indicate either male (0) or female (1). Participants' race and ethnicity were indicated by a set of dummy variables with White (non- Hispanic) as the reference group (0) compared to African American, Hispanic, and other (1).

Analytic Strategy Missing Data Multiple imputation was used for all variables with missing data, because it can be used with confidence for variables having up to 30% missing data, when determined to be missing at random (MAR) or missing completely at random (MCAR) without increasing Type I error (Croiseau, Genin, & Cordell, 2007). In multiple imputation, several additional data sets are created, each with a different imputed value on missing variables (Acock, 2005). Parameter estimates are then pooled across each imputation, and the mean estimate of the combined imputations is used in the analyses. Missing-data uncertainty is maintained in the final parameter estimates and standard errors. In the current study, the majority of variables were missing at a rate of less than 5%. Three variables were missing at greater rates with both parent- school involvement items missing at 14% and household income missing at 24%. Preliminary Analyses Prior to conducting primary analyses, sample descriptive statistics were examined for violations to assumptions of normality, assessing kurtosis and skewness. Assumptions of normality are considered violated when values are greater than 10.0 for kurtosis and 3.0 for skewness (Kline, 2011). Although logistic regressions are not influenced by violations of normality compared to other analyses (Leech, Barrett, & Morgan, 2008), the data had kurtosis values outside the acceptable range (3 items ranged from 10.4 to 35.9), so robust estimators were used in all models which adjust for minor violations of normality. Next, bivariate correlations of all variables were examined to test for high correlations indicative of multicollinearity (see Table 2). Relationships among most variables were as expected with educational outcomes negatively associated with being male, non-White, having more siblings, a lower household income, and parents with fewer years of education at p < .05. Additionally, retention, delinquency, early first sexual experience, having a learning disability, experiencing more family transitions and school changes, and attending a public school were associated with poorer educational outcomes at p < .05. Of the family structure variables, living in each stepfamily type was negatively correlated with educational outcomes; yet, correlations were only significant (all at p < .01) for simple stepfathers and blended stepfamilies. Finally, parental monitoring, parent-child relationship quality, and parent-school involvement were all positively associated with educational outcomes.

Regarding multicolinearity, although the strongest correlation was .66 (p < .01) between mothers’ and fathers’ education, variance inflation factors (VIF) and tolerance statistics were also examined. High VIF values (> 10) and low tolerance (< .10) indicate variables which are likely to inflate the variance explained by the model and should be excluded (Chatterjee & Hadi, 2006). Based on these criteria, multicollinearity was not a concern. Confirmatory factor analysis. A multilevel confirmatory factor analysis (CFA) was conducted in Mplus 6.12 to test the construct validity of the three social capital variables (i.e., parental monitoring, parent-child relationship quality, and parent-school involvement). Because these three variables are theoretically connected and examined as factors of social capital both in the extant literature and the current study, social capital was examined as a second-order factor with the purpose of determining its validity as a latent construct and usefulness as a theoretical link. Second-order factors can be included in a CFA when three or more first-order factors are present (Kline, 2011). The sample in these analyses included some siblings within the same family, so multilevel models (Raudenbush & Bryk, 2002) were used to manage issues of dependency in the data. This is particularly important when examining the social capital variables, as they are likely to be highly correlated within families. Because the sample contained unbalanced groups, ranging from 1-3 adolescent respondents per family, the CFA was conducted using Maximum Likelihood Robust (MLR) estimation in Mplus with the CLUSTER variable and COMPLEX analysis commands. This method accounts for dependency within groups and provides adjusted model fit statistics, without providing separate within and between group parameter estimates (Muthen & Muthen, 1998-2010). When using MLR with unbalanced group size, Mplus treats the unbalanced data similar to incomplete data, where the model and likelihood are defined based on raw data; it is able to incorporate random slopes and intercepts into the model (Hox, 2010). Thus, robust estimates of standard errors and chi-square test statistics are used, as these adjust for issues of non-normality and correct for any remaining heterogeneity among groups. The robust chi-square provided in MLR is the Satorra-Bentler scaled chi-square (Satorra & Bentler, 2001). Chi-square difference tests are then conducted using these adjusted scores to compare model fit among nested models (Muthen & Muthen, 2005). When a second-order factor is added to a model with three or more first-order factors, a chi-square difference test and related fit indices are not useful in determining change in model fit, as the chi-square and degrees of freedom

23 remain unchanged. Rather, significant loadings of the first-order factors on the second-order factor indicate the validity of the added factor (Doll, Raghunathan, Lim, & Gupta, 1995). CFA was used because specific items were expected to create the three intended social capital factors, and error variances were expected to be correlated among factors (Thompson, 2004). Even when theory indicates an expected number of factors, Kline (2011) recommends examining models with fewer factors to determine whether a simpler model is a better fit to the data. Thus, one-, two-, and three-factor models were examined in order, followed by the addition of social capital as a second-order factor. Model fit was determined by examining chi- square test statistics, CFI (comparative fit index), RMSEA (root mean square error of approximation), and SRMR (standardized root mean residual). A low chi-square and high p- value indicates that the residual matrix is different from the observed data matrix and that the model is likely a good fit to the data. Because this statistic can incorrectly indicate good fit when samples are large or correlations are high, it is important to examine additional approximate fit indices. The RMSEA is adjusted based on the model degrees of freedom, which makes it a more accurate estimate than chi-square when samples are large. The close-fit hypothesis suggests that RMSEA values of .05 or lower indicate a good fit and .10 or higher indicate poor fit (Kline). The CFI ranges from 0 to 1, with values ≥ .90 indicating an acceptable model and ≥ .95 indicating good fit, or that the model is a significant improvement compared to the baseline model (Hox, 2010). Finally, the SRMR is based on examining the difference between the observed and predicted covariance. Thus, a value of 0 would indicate perfect fit; yet, fit is considered good for values ≤ .08 (Hu & Bentler, 1999). In addition to fit indices, modification indices were examined to determine whether model fit could be improved by making adjustments to the model. Changes suggested by modification indices should only be made when they make theoretical sense and the associated decrease in chi-square is significant (Kline, 2011). When two nested models are both determined to be good fits to the data, a chi-square difference test determines whether the changes improved the model fit in relation to the changes in degrees of freedom, with non-significance indicating better fit to the model with fewer parameters. Results from the CFA examining social capital variables are presented in Tables 3 and 4. To be sure that the proposed second-order model provided the best fit to the data, several models were examined and compared. Model 1 included only one factor with all 14 items. All but one

24 standardized factor loading were acceptable and significant at p ≤ .001. Item 5 of those measuring parent-child relationship was not significant with a standardized loading of .01 and p = .55. Based on model fit indices (see Table 4), this model was a poor fit to the data (χ2 [78] = 13,969.17, p < .001; CFI = .066; RMSEA = .094; SRMR = .280). Model 2 tested two-factors with the eight parent-child relationship items loading on one factor and monitoring and school involvement items loading on a second factor. All standardized factor loadings were acceptable and significant at p ≤ .001. The chi-square was significant, χ2 (77) = 6,904.6, p < .001, indicating poor model fit. Although the fit indices were slightly better, overall the model was a poor fit to the data (CFI = .541; RMSEA = .067; SRMR = .196). Next, a three-factor model was tested with eight parent-child relationship items, 4 monitoring items and 2 parent-school involvement items as predicted. All standardized factor loadings were acceptable and significant at p ≤ .001. The chi-square remained significant, χ2 (75) = 2,300.48, p < .001; yet, other fit indices improved to indicate a close fit, according to the CFI (.850), and a good fit based on the RMSEA (.039) and SRMR (.068). In Model 4 a second- order factor was added to determine if monitoring, parent-child relationship quality, and parent- school involvement factors were explained by a higher-order factor of social capital. When a second-order factor was added to a model with three correlated first-order factors, the model fit indices remain unchanged, because the chi-square and degrees of freedom were unchanged. Standardized loadings were acceptable and significant at p ≤ .001, supporting the added value of including a second-order factor. Further, because the CFI was low, and parent-child relationship and monitoring items shared the same measurement method, modification indices were examined to determine if correlated error variance among items would improve model fit (Kline, 2011). The model was then respecified to include correlated error variances for items as displayed in Figure 1. Following respecification, the chi-square was still significant, χ2 (63) = 277.97, p < .001; yet, additional model fit indices improved with all indicating good model-fit (CFI =.98; RMSEA = .013; SRMR = .032).

Models 4 (χ2 [75] = 2,300.5; p < .001) and 5 (Model 4 with respecification; χ2 [66] = 277.97; p < .001) both had good model fit indices, so a chi-square difference test was conducted using S-B correction to compare the nested models with the following equation (Muthen & Muthen, 2005):

2 χ S-B rescaled = (1) – where the scaling factors, which are the S-B adjustments to chi-square, for Models 4 and 5 were 6.48 and 6.31, respectively. The result of the rescaled chi-square difference test was χ2 (9) = 1,686.57, well above the cutoff of 16.19 for α = .05. However, the additional model fit indices, which are based off of chi-square and model degrees of freedom, showed that the respecified second-order model provided the best overall fit to these data. These results support the use of three separate factors (parental monitoring, parent-child relationship, and parent-school involvement) as valid constructs to examine an overall theoretical conception of adolescents’ social capital. Although linked to a second-order factor, each measure was examined separately to determine the different mechanisms though which social capital may variably influence educational outcomes. Multilevel Logistic and Linear Regressions Various dichotomous educational outcomes were examined using multilevel logistic regression with a random intercept (Rabe-Hasketh & Skrondal, 2008). The random intercept models the combined effect of adolescent-specific covariates which are omitted from the model but may cause some siblings in families to have better educational outcomes than others. Thus, the random effect represents the variation in outcomes for adolescents within families, and the fixed effect estimates represents the variation in outcomes for individuals across families. The estimation of variance components when group size is unbalanced requires the use of an iterative numerical procedure to obtain the most precise estimates (Raudenbush & Bryk, 2002). So, the full maximum likelihood (FML) procedure with numerical integration was used in Stata 12 to achieve the most precise estimates available (Agresti, Booth, Hobert, & Caffo, 2000). Similar to single-level logistic regression, using a logit-link function, odds ratios based on estimated log odds are obtained which indicate the extent to which a one unit change in each covariate influences the likelihood of the outcome variable, holding all other variables constant. Percentage of change of odds is calculated as 100(1 – odds ratio). The residual interclass

26 correlation (ICC) ρ is a parameter which indicates the amount of dependence among individuals within the same family on the outcome measures, yet with binary outcomes this estimate is less useful, due to the heteroscedasicity of level-1 variance (Raudenbush & Bryk, 2002). Also, an F- test, which accounts for uncertainty in the estimated individual-level variance, is used in place of a Wald chi-square test to determine the significance of the overall model, or the null hypothesis that at all predictors in the model have coefficients equal to zero (Rabe-Hesketh & Skrondal, 2005). When the p-value is less than .05, the model is significant with at least one predictor in the model having a coeffiecient not equal to zero, and the null hypothesis is rejected. Because the purpose of including a random intercept in these analyses was to manage violations of the independency assumption rather than to make inferences about variance at the within- and between-family levels, ICC’s are reported in tables only. The log likelihood statistic (-2LL) is a chi-square test that is normally used to compare the change in -2LL across nested models and determine whether adding more predictors increases the ability of the model to predict a binary outcome (Chatterjee & Hadi, 2006). However, this statistic is not available following multiple imputation of missing data, as estimates are approximated for each model parameter separately, and when imputed results are combined, they are no longer likelihood results (Carlin, Galati, & Royston, 2008). Instead, an F- test, similar to the model F-test, is conducted which tests the null hypothesis that the added predictors for each model have coefficients all equal to zero. P-values less than .05 indicate a significant contribution of added predictors. Odds ratios of the fixed effects (variation for individuals across families) in the model are interpreted as the standardized influence of each predictor on the log odds of the outcome, after accounting for the influence of the random effect (within families variation). Beyond the specified demographic and other control variables, dummy variables for each family structure category (with biological children in intact families as the reference group) were created to test Hypothesis 1 (Participants living in stepfamilies during adolescence will have poorer outcomes on all educational behaviors compared to those from intact families). Social capital variables (parental monitoring, parent-child relationship, parent-school involvement) were included to test Hypothesis 2 (Participants with weaker social capital during adolescence will have poorer outcomes on all educational behaviors compared to those with stronger social capital), and two-way interactions among family structure and social capital measures were

27 included to examine Hypothesis 3 (Family structure will moderate the relationship between social capital and educational behaviors, such that the relationship will be weaker for those in each type of stepfamily). The combined individual- and family-level model is represented as: ) (2)

where Xij represents a vector of individual-level covariates, and Zj represents a vector of family- level covariates, is the interaction between individual- and family-level predictors, and

is the family-level variation in overall risk (Hox, 2010). Next, a separate multilevel logistic analysis examined only children living with both biological parents from both intact families and blended stepfamilies to examine Hypothesis 4 (Mutual biological children in blended stepfamilies will have poorer outcomes on all educational behaviors compared to biological children in intact families). All control variables were included in 4 of the 6 logistic regressions, and in the remaining two (completing any post-secondary degree and completing at least a bachelor’s degree), the learning disability variable was omitted due to issues of multicollinearity as evidenced by large standard errors. A dummy variable for living in a blended family was included along with interaction terms between blended family status and each measure of social capital. A final multilevel logistic regression was conducted, including only those in blended families. A dummy for stepchild status was added to test Hypothesis 5 (Within blended families, stepchildren will have poorer outcomes on all educational behaviors compared to mutual biological children), along with social capital variables, and two-way interactions among stepchild status and social capital variables to test Hypothesis 6 (Stepchild status in blended families will moderate the relationship between social capital and educational behaviors, such that the relationship will be weaker for stepchildren compared to mutual biological children). Outcomes examined were: (a) the likelihood of dropping out of high school, (b) the likelihood of return to school following high school dropout, (c) the likelihood of completing high school, (d) the likelihood of entering post-secondary education, and (e) the likelihood of completing any post-secondary degree, and (f) the likelihood of completing at least a bachelor’s degree. Multilevel Linear Regression A multilevel linear regression with a random intercept was conducted using the same set variables for the logistic regressions to examine the influence of predictors on participants total years of education completed and highest degree completed, allowing individuals to vary within

28 families. As in the logistic regressions, the parameter estimates of interest were the fixed effects, with the random effect accounting for violations of independency in the data. Similar to the logistic regressions, the first linear regression tested Hypotheses 1-3 for all participants. A second regression tested Hypothesis 4 for those living with both biological parents, and a second regression tested Hypotheses 5 and 6 for those in blended families only. The combined level-1 and level-2 equation is: (3) where Xij represents a vector of individual-level predictors, Z ij represents a vector of family-level predictors, is the interaction between individual- and family-level predictors, and is the family-level variation in overall risk. As in the logistic analyses, the F-statistic is also reported for the overall model significance, with a p-value < .05, indicating that the model is significant with at least one predictor in the model has a coefficient not equal to zero. The ICCs are reported to indicate the proportion of variance in the model due to the clustering of individuals in families. Models are compared using the F-statistic to test the null hypothesis that the added predictors for each model have coefficients all equal to zero. Finally, the fixed effects (variation for individuals across families) in the model are reported as the standardized coefficients of influence of each predictor on the outcome after accounting for the influence of the random effect (within family variation). Sample Weights The sampling design of the NLSY97 included an over-representation of Black and Hispanic adolescents. Although weights are available to adjust for this, it is recommended that they not be used when examining data from multiple rounds (Moore, Pedlow, Krishnamurty, & Wolter, 2000). Therefore, no weights were included in these analyses. Sample Size and Statistical Power To determine the likelihood of committing a Type II error, a power analysis was conducted to determine the sufficiency of the sample size. First, the following formula was applied to detect the standard error needed for a power of .80 to detect a medium effect size of r =.30 (Cohen, 1992) at p ≤ .05: (4)

With standard error of (.30/2.48 =) .12 (Hox, 2010). The formula for calculating standard error is

) (5) Thus, the minimum sample size required for a single level analysis would be (1/.12)2 + 3 = 72.4 √ individuals. However, determining needed sample size to achieve desired statistical power in multilevel models requires estimation of the design effect (Gulliford, Ukoumunne, & Chinn, 1999). Using the average level-two (family) group size and the ICC, the design effect is calculated as: design effect = 1 + (n – 1)ρ (6) where n is average group size, and ρ is the ICC of the outcome variable. The ICC represents the amount of variance that exists between clusters in proportion to total true variance (Hox).The ICC estimate for the highest grade completed in the current analysis was .36. Thus, the design effect for a main outcome variable in the current study is: design effect = 1 + (1.13 – 1) .36 = 1.047. (7) Finally, multiplying the design effect by the effective sample size calculated equation 5 results in a required sample size of only (73*1.047) = 75.9 for clustered data, well below 3,334. Thus, the current sample yielded a power of at least .80 for an effect size of r =.30.

CHAPTER FOUR STUDY RESULTS

In this chapter, I report the results for each of the six hypotheses tested in this study. First, I examined the effect of living in a stepfamily on various educational outcomes, followed by the influence of social capital on the same outcomes. Next, I examined the interactions between family structure and social capital to test the moderation hypothesis that the relationship between social capital and educational attainment is weaker for those in stepfamilies compared with those in intact families. I then compared stepchildren and mutual biological children in blended families, including interactions between stepchild status and social capital to test the moderation hypothesis that the relationship between social capital and educational outcomes is weaker for stepchildren in blended families compared to mutual biological children in blended families. The goal of each analysis was to examine the relationships among variables. The significant contribution of several control variables varied across the outcomes of interest; yet, non-significant variables were retained to allow for direct comparisons of models across all outcomes. Results from Descriptive Analyses Characteristics were reported for the entire sample to examine the trend in educational outcomes across all participants (see Table 5). Approximately 21% of participants dropped out of school at least once prior to completing a high school diploma or GED. Of those, 38.4% eventually enrolled in school at a later date, and 10.7% of the sample had not yet earned a high school diploma or GED. Over half of the sample (61.7%) enrolled in post-secondary school for at least one semester, and just over half of them (33.7% of the total sample) eventually earned at least a 2-year degree. Similar to reports from U.S. Census data (2010), 27.4% of participants earned at least a bachelors’ degree. When measured as total years of education completed, the average educational attainment by ages 25-27 was 14.27 (SD = 2.44) years (see Table 6). Hypotheses Testing Family-level structure. Hypothesis 1, that those from each type of stepfamily would fare worse on educational outcomes compared to those from intact two-parent families, was partially supported by the data. First, differences in demographic characteristics and educational outcomes were examined comparing individuals by overall family structure type using crosstabulations (Table 5) and ANOVAs (Table 6). Gender did not differ significantly by group.

Significant racial differences were found, χ2(3, 3,334) = 76.64, p <.00, with a higher percentage of Black adolescents living in simple stepfather (26.8%) and blended stepfamilies (25.4%) than intact two-parent families (13.8%) or complex stepfamilies (15.9%). A lower percentage of White participants were in simple stepfather and blended stepfamilies (55.2% and 50.8% respectively), and a lowest percentage of Hispanic participants were in simple stepfather families (16.9%). Those in complex stepfamilies were more likely to have two or more siblings in the home χ2(3, 3,334) = 76.64, p <.00. In addition, results from the ANOVAs (Table 6) also showed that groups differed in the household income reported during Round 1, F (3, 2,499) = 11.80; p < .00, with post hoc comparisons indicating that income was significantly higher in intact two-parent families compared to simple stepfather and blended stepfamilies. Significant differences were also found for mothers’ (F[3, 3,207] = 4.90; p < .05) and fathers’ education (F [3, 3,218] = 5.92; p < .05), with mothers in intact two-parent families completing significantly more years of education than those in blended stepfamilies, and fathers in intact families and simple stepfather families completing more years of education than those in blended stepfamilies Regarding educational outcomes, significant differences were found among family structure types for seven of the eight outcomes. Simple stepfather and blended stepfamilies had higher percentages of high school dropouts (33.4% and 36.3%), χ2(3, 3,334) = 112.17, p <.00, and lower percentages of high school completion (84.9% and 83.8%), χ2(3, 3,334) = 30.75, p < .00, than either intact two-parent families or complex stepfamilies (see Table 5). All stepfamily types had lower percentages of post-secondary school enrollment, χ2(3, 3,334) = 90.66, p < .00, than intact two-parent families. Again, simple stepfather and blended stepfamilies had the lowest percentages for completion of a post-secondary degree with 21.8% and 13.8% completing any degree, χ2(3, 3,334 = 120.01, p < .00, and 13.7% and 9.6% completing at least a bachelor’s degree, χ2(3, 3,334) = 118.6, p < .00. Differences among family structure groups for total number of years completed were also significant, F(3, 3295) = 51.76, p < .00 (see Table 6). Post hoc comparisons indicated that those in simple stepfather families completed fewer years than those in complex stepfamilies p = .08, and those in complex stepfamilies completed significantly fewer years than those in intact families at p = .06. Finally, groups also differed significantly in terms of highest degree completed, F(3, 3,330) = 53.62, p < .00. Post hoc comparisons showed that

32 only complex stepfamilies (M = 2.25, SD = 1.43) and intact families (M = 2.54, SD = 1.30) differed at p = 09. Next, a series of random intercept logistic and linear regressions were used to examine 6 of the 8 educational outcomes. Compared to those in intact two-parent families, the odds of dropping out of high school were higher for those in simple stepfather families by 167% (p < .01) and for those in blended stepfamilies by 85% (p < .01) after controlling for demographic factors and additional factors known to influence educational outcomes (Table 7). Family structure type was not associated with the odds of returning to high school after dropping out; yet, the odds of completing high school or a GED were lower for those living in blended stepfamilies by 37% (p < .05) compared to those in intact two-parent families. The likelihood of enrolling in post-secondary school was 44% (p < .01) lower for those in simple stepfather families and 43% (p < .01) lower for those in blended families. The odds of completing any post- secondary education were 47% lower for those in stepfather families and 64% lower for those in blended stepfamilies. Those in stepfather families were 62% (p < .01) less likely to complete at least a bachelor’s degree, and those in blended families were 71% (p < .01) less likely to do so. Living in these two family types was also negatively associated with the total number of years of education completed (β = -.08, p < .05 for stepfather families and β = -.09, p < .01 for blended families) (Table 8). Coming from a complex stepfamily was not significantly associated with any difference in educational outcomes compared to those in intact two-parent families. Social capital and educational outcomes. Parental monitoring, parent-child relationship quality, and parent-school involvement were measured separately as three distinct aspects of adolescent’s social capital. Hypothesis 2, that greater social capital would be associated with more positive educational outcomes was partially supported (see Tables 7 and 8). Parental monitoring and parent-school involvement were not significant predictors of any of the educational outcomes for the full sample. However, having a higher quality parent-child relationship was associated with four of the eight outcomes. Specifically, a higher quality parent-child relationship was associated with a greater likelihood of completing any post- secondary degree by 4% (p < .05) and completing a bachelors’ degree or higher by 2% (p < .05). Higher parent-child relationship quality also was associated with higher total years completed (β = .11; p < .01) and highest degree completed (β = .07; p < .01). When only those living with both biological parents in intact or blended families were examined (see Tables 9 and 10), a higher

33 quality parent-child relationship was positively associated with the total number of years of education completed (β = .07; p < .01), highest degree completed (β = .03; p < .05), and a 5% (p < .05) increase in odds of earning at least a bachelor’s degree. Within blended families only (see Table 11), greater average parent-school involvement was associated with a 33% (p < .05) increase in the odds of completing a bachelor’s degree or higher. Family structure as a moderator. Hypothesis 3, that family structure would moderate the relationship between social capital and educational outcomes, was partially supported (see Tables 7 and 8), as significant effects were found only for select outcomes. Living in a blended family moderated the relationship between parent-child relationship quality and completing high school/GED (OR = .87; p < .05), such that it was stronger for those in intact two-parent families (OR = 1.03; p < .00) than in blended families (OR = .99; p = .60). The relationship between parent-school involvement and completing at least a bachelor’s degree was also moderated by living in a blended family (OR = 1.36; p < .05). Further examination showed that the relationship was stronger for those in biological families (OR = .96; p < .01) than for those in blended families (OR = 1.04; p = .73). This is contrary to expectations, as greater parent-school involvement was associated with a decrease in the total number of years completed for those in intact families. Similarly, when comparing biological children in intact families to mutual children in blended families (see Tables 9 and 10), living in a blended family moderated the relationship between parent-child relationship quality and the likelihood of high school dropout (OR = 1.27; p < .05), such that a higher quality relationship decreased the odds of dropout by 5% (p < .01) for those in intact families and had no significant effect for those in blended families (OR = 1.04; p = .12). Living in a blended family also moderated the influence of parent-school involvement on the likelihood of completing at least a bachelor’s degree (OR = 1.53; p < .05). Further examination indicated that the relationship was stronger for those in intact families (OR = .96; p < .01) compared to those in blended families (OR = 1.09; p = 6.15). Thus, a greater involvement decreased the likelihood of completing a bachelor’s degree by 4%. Comparing stepchild and mutual biological children within blended families. Within blended families, educational outcomes were examined to test Hypothesis 5, that stepchildren in blended families will fare worse than mutual biological children in blended families, and Hypothesis 6, that stepchild status would moderate the relationship between social capital and

34 educational outcomes (see Tables 11 and 12). After accounting for the control variables, the influence of stepchild status and interactions among stepchild status and social capital measures across all eight educational outcomes were not significant; thus, these hypotheses were not supported. Additionally, within blended families only, no significant relationships were found between any of the social capital variables and educational outcomes. Comparing biological children in intact and blended families. The final set of analyses examined only those living with both biological parents in intact two-parent families and blended families to test Hypothesis 4, that biological children in blended families will have poorer educational outcomes compared to biological children in intact two-parent families. Results were that mutual children were 54% (p < .05) less likely to complete any post-secondary degree and 60% (p < .05) less likely to complete at least a bachelor’s degree (see Table 9). Additionally, being a mutual child in a blended family was negatively associated with the highest degree completed (β = -.03; p < .05) (see Table 10).

CHAPTER FIVE

DISCUSSION

Given the significant impact that educational outcomes have on success and well-being throughout one’s lifespan, and the established relationships between both family structure and social capital and educational outcomes, there were two aims in the current study. The first aim was to examine individual- and family-level characteristics, family structure, and social capital to determine which factors place adolescents at risk for poor educational outcomes. The next aim was to test how family structure influences the relationship between social capital and educational outcomes. These aims were examined specifically for those residing in two-parent families to highlight any specific risk of living in a stepfamily, and particularly the risk for those youth in blended stepfamilies who are biologically related to both parents (mutual children). Important to this study is the inclusion of multiple indicators of educational outcomes, as well as a glimpse into differences between biological children in intact two-parent families and blended families across a variety of outcome indicators. The central hypotheses in this study addressed the influence of family structure on the relationship between social capital and educational outcomes. This relationship was expected to be weaker for those living in stepfamilies than those in intact families based on the concepts of role and relationship salience (Stryker, 1968; Umberson & Chen, 1994). Social capital arises within relationships between parents and children. When those relationships are steprelationships, the magnitude of their effects on an individual should decrease. Few findings supported this hypothesis; however, those that did indicated that the influence of a higher quality parent-child relationship was important for decreasing the likelihood of high school dropout and increasing the likelihood of high school completion for biological children in intact two-parent families only. Further, greater parent-school involvement was associated with a decrease in the likelihood of completing a bachelor’s degree, again only for those in intact two-parent families. Two possible conclusions can be drawn from this finding. First, because theory suggests that biological parents may invest more heavily in their biological children, they may also hold higher expectations for their success and do more to ensure success. Thus, the meaning of parents’ involvement with their children’s school may be greater and linked with additional indicators of social capital not measured here, such as parental expectations and school-related

36 discussion. Also, previous studies have focused primarily on the influence of parent-school involvement for educational outcomes such as high school dropout and completion, events which occur while individuals are still living in their parents’ homes. Having parents heavily involved in one’s schooling during high school may decrease feelings of competency and responsibility for one’s own education, thus leaving students unprepared for the challenges of meeting educational requirements in a less structured environment. Interestingly, social capital during early adolescence had little influence on the educational outcomes in this study. The link between social capital and education has been well- established in previous literature across a variety of specific outcomes. It may be that the items used to measure social capital were not adequate indicators of capital for this sample. For example, rather than parent’s volunteering and attendance at school meetings, measures that tap parents’ engagement with adolescents in relation to their schooling (e.g., helping with homework, transporting adolescents to school-related activities) may exert more influence on future educational outcomes. Moreover, the items used here do not address important aspects of parenting, such as overall engagement, accessibility, and responsibility which represent indicators of paternal involvement known to influence child outcomes (Pleck, 2010). Social capital theory (Coleman, 1988) emphasizes that parents’ greater financial and human capital is transferred within the context of the parent-child relationship to influence children’s future outcomes. Rather than using a measure of parental monitoring which examines how much parents’ know about children’s social circle, adolescents may gain more through access to parents’ human capital when they are more aware of and engaged in their parents’ social circles. Additionally, the lack of findings for long-term outcomes suggests that any influence of social capital during early adolescence is not enduring, and perhaps measuring social capital later on using items appropriate for capturing the nature of parenting emerging adults would better assess this relationship. The findings presented here provide a comprehensive illustration of how the effects of family structure and social capital on educational outcomes during early adolescence vary by a particular outcome. For example, coming from a simple stepfather or blended family increased the likelihood of high school dropout and decreased the likelihood of completing any college or completing at least a bachelors’ degree, and was associated with fewer total years of school; yet, these family contexts did not significantly influence high school completion or post-secondary

37 enrollment. This finding is similar to that of previous research in which the effects of social background factors at age 16 on decisions to continue education vary across educational outcomes as early as the transition into high school (Mare, 1980; Hauser & Andrew, 2006). The difference in influence across outcomes may be due to the differences in timing of certain educational outcomes. These findings suggest that the influence of living in a simple stepfather or blended family in early adolescence is greater for longer term outcomes. As expected, those from simple stepfather families and blended stepfamilies had poorer educational outcomes compared to those in intact two-parent families. Interestingly, no significant effects were found for those in complex stepfamilies. Demographically, complex stepfamilies were similar to intact two-parent families in terms of race, income, and parents’ education than to simple stepfather and blended families. It may be that this stepfamily type is unique from other stepfamilies in other ways. For example, complex stepfamilies are formed when both parents bring biological children into a remarriage and elect not to have a mutual child/ren. It maybe that complex stepfamilies include older adults with older children at the time of remarriage, so having mutual children is less attractive. On the other hand, they may be of shorter duration compared to other stepfamily types, so they not yet had mutual children. Data were not available in the current study to test this assumption. Additionally, compared to simple stepfather families, complex stepfamilies are different, because they include fathers who have joint or full physical custody of their children. Only some blended stepfamilies included fathers with residential children from a prior relationship. Fathers bringing children into complex stepfamilies may differ from those who do not do so. When fathers have at least joint physical custody, it can mean that the father sought such an arrangement or that the mother elected not to be a full-time parent. Findings from a qualitative study (Hamer & Marchioro, 2002) showed lower-income custodial fathers with fulltime custody arrangements had often assumed this role by default when mothers chose not to maintain custody. However, there is no data of determine whether this is true of the current sample. The income and other demographic factors of those in complex stepfamilies resembled those of intact two-parent families. Findings show that fathers with a higher income are more likely to have joint custody, whereas those with lower income are more likely to have full custody (Cancian & Meyer, 1998). Joint custody is also more likely when fathers own their home, have all boys, and no children from a previous relationship. Full custody fathers are

38 linked with mothers who are economically disadvantaged and children who are older. In addition, other selection factors may affect whether or not fathers have joint- or full-custody, such as more interest in parenting, better parenting skills, or having children who are less difficult to parent. Thus, it may be through both certain demographic characteristics and other selection factors that those in complex stepfamilies experience less risk for poor outcomes compared to those in other stepfamily types. Results comparing mutual biological children in blended families with biological children in intact two-parent families showed no significant differences for early educational outcomes. However, mutual children in blended families were less likely to complete a post-secondary degree. These children often go unrecognized as being at any risk for poorer outcomes, because they reside with both biological parents (Gennetian, 2005). These findings add to recent research taking a closer look at the added risk associated with living with a stepparent and also living with half-siblings (Tillman, 2008). Although mutual biological youth are no less likely to enroll in post-secondary education, completion of higher degrees is most influential to securing future employment and increasing lifetime earnings (U.S. Census Bureau, 2010). Thus, even though they reside with both biological parents, something about living in a blended family may impact the long-term success for mutual children. By definition mutual children represent higher-order births in their families; they are born into families where children already reside. Findings show that birth order is associated with educational attainment at both the high school and postsecondary levels, with later born children attaining less education than first and earlier born children (Black, Devereux, & Salvanes, 2005; Kantarevic & Mechoulan, 2006). Therefore, the greater risk of poorer education outcomes associated with being a mutual child may also be related to being a second- or later-born child. As with family structure, when social capital was associated with outcomes, its influence varied by the indicator of social capital and the outcome assessed. Again, none of the social capital variables were significantly associated with high school completion; yet, a higher quality parent-child relationship increased the odds of enrolling in post-secondary school, completing any post-secondary school, and completing at least a bachelors’ degree – all of which typically follow high school completion. Because most students attain at least a high school education (U.S. Census Bureau, 2010), the risk for not doing so may be smaller than later educational outcomes (e.g., attending and completing college), which are less frequent occurrences. Thus,

39 less parental monitoring or poor parent-child relationship quality may not influence the likelihood of highly frequent events for the average adolescent. Rather, these factors may be more influential among those with additional risk factors which were not examined in the current analyses (e.g., neighborhood crime)levels (Woolley & Grogan-Kaylor, 2006), immigrant status (White & Glick, 2000), or living in a rural area (Lesley & Dianne, 2001). Parent-school involvement, though not a significant predictor of outcomes alone, influenced several outcomes when moderated by living in a blended family. Rather than increasing the likelihood of positive educational outcomes, greater parental involvement was associated with poorer outcomes for those in biological families compared to those in blended families. As is the case with parallel findings that parental monitoring decreased expectations of attending college (Manning & Lamb, 2003), causal order cannot be implied here. Parents may become more involved in schooling through participation in school organizations or volunteering as a response to the adolescent’s already low academic engagement, success, or college expectations. In addition, no differences were found between stepchildren and mutual biological children in blended families for any of the educational outcomes. The examination of differences between stepchildren and mutual children in blended families is relatively new in stepfamily research. Such studies show that stepchildren tend to fare worse in these types of families (e.g., Case, Lin, & McLanahan, 2001). However, Ginther and Pollak (2004) examined a similar set of outcomes, and they found no significant mean differences between stepchildren and mutual children in blended families. An important difference in these studies is that the later exclusively focused on those in stable blended families, and this may account for the discrepant findings. Stability in family structure was partially controlled in the current study by including the number of family transitions through later adolescence. Also, those who were identified as mutual children in blended families were likely in stable families, as they reported living with both biological parents at Round 1 when they were at least 12 to 14 years old. It may be that children in blended families of longer duration do not experience deleterious effects in the same way as those in less stable blended families. Lastly, it may be that social capital and other family processes mediate the relationship between family structure and educational outcomes. For example, family stress theory (Boss, 2002) suggests that living in a stepfamily may lead to greater family conflict and higher levels of

40 stress. In turn, the probable nature of stepfamily life, especially early on, can negatively affect educational outcomes. However, it was impossible to test such hunches with the current data. Limitations Several limitations are noteworthy. Due to limitations in the NLSY97 dataset, parent- school involvement was measured with only two items that emphasize such involvement occurring in the school setting and often during school hours. This may bias the results in terms of socio-economic status, race, and ethnicity. Zellman and Waterman (1998) found that even though Black parents are less likely than White parents to volunteer in the classroom, they are no less likely to attend school events or participate in school governance -- two additional indicators of school involvement. Also, the reliability for the composite score used in the analyses was low (α = .58), decreasing confidence in the measure and suggesting the need to examine the items separately in the future. Also, inclusions of additional indicators of involvement are warranted but not possible with the current data set, and it may be helpful to examine them separately in future research. Other limitations are associated with social capital. The items used to assess parental monitoring and parent-child relationship quality. These items may have failed to capture the aspects of social capital most important to the outcomes of interest. A more comprehensive examination of social capital including items that assess parental engagement, accessibility, and time spent with children may have led to more significant findings. Also, the measure of social capital was collected only during early adolescence, the nature of social capital may vary in late adolescence and emerging adulthood, as well as exert more impact when measured closer to the various outcomes. The cumulative influence of social capital and other factors beginning in early childhood is important for later outcomes in adolescence and young adulthood (Alexander, et al., 2001; Jimerson, et al., 2000); yet, it was impossible to assess this with the current data. Further, the direction of causality between social capital and educational outcomes cannot be inferred from these results. It remains unclear whether greater parent-school involvement or a higher quality parent-child relationship is the cause of or a response to educational outcomes. Another limitation was the inability to account for family structure stability prior to Round 1. Although some participants retrospectively reported family structure for several points throughout childhood, these data were too inconsistent to be included in the analyses. Findings show that family structure at one point, multiple family structure transitions overtime, and the

41 timing of such transitions together are important (Cavanagh, et al., 2006; Sun & Li, 2009). The current study only accounted for family structure at ages 12-14 and structural transitions from this time to age 18. Also, because only Round 1 measures of the independent variables and control variables were used, inference from these results are derived solely from the influence of these factors during early adolescence on later educational outcomes. Unfortunately, NLSY97 focuses primarily on the individual adolescent, following him/her into young adulthood, and many of the family-related variables are not included once participants leave home. Thus, the availability of family-level control variables and family process measures was sporadic in later rounds of data collection. Due to inadequate sample size, those in simple stepmother families were excluded, and this is another limitation of the study. It is possible that the children in this group represents a population at higher risk for poor educational outcomes, particularly as stepmothers come into this family type with no prior parenting experience and are often expected to take on an active parenting role (Coleman, Troilo, & Jamison, 2008). Finally, although several hypotheses were supported based on findings at a significance level of p < .05, the magnitude of the effect size for most of these differences were small. Thus, significant differences found here do not necessarily imply substantive difference. Future examination of these hypotheses using other samples may result in finding more substantive differences. Implications and Conclusions This was the first study to examine how family structure influences the relationship between social capital and educational outcomes, so it provides a basis for further investigations with other data. Although findings provided limited support for the moderation hypothesis, future research should continue to examine this, accounting for the influence of structure and social capital at various developmental stages. It may be that earlier educational outcomes, such as high school completion, are most influenced by structure and parenting behaviors during childhood. It may also be that the trajectory of such outcomes is already set in motion by the time students reach early adolescence. In fact, Hickman and colleagues (2008) found that the path to eventual high school dropout can be traced back to kindergarten (e.g., more days absent from school, poor performance on standardized tests). It may also be that other indicators of social capital not included here are more influential to later educational outcomes.

The construct validity of the measure of social capital measures used in this study was confirmed through a CFA, and they were associated with a higher-order factor. As noted, future analyses may benefit from a more thorough investigation of social capital beyond the three examined here. For example, inclusion of parental expectations, parental interest and involvement in children’s activities, children’s involvement in parents’ social networks, and additional indicators of parent-school involvement (e.g., helping with homework, having discussions about school) might prove to be better indicators or social capital. Longitudinal research that provides a thorough history of family processes and educational experiences are needed to address the issues associated with the casual direction of effects between social capital and outcomes. For example, it is important to determine whether parental monitoring and involvement increase as a result of poor educational outcomes, or if they lead to poor outcomes. If increases in these parenting behaviors negatively impact children’s future education, then schools could focus on encouraging parents of adolescence to engage in monitoring or involvements that are more appropriate for their child’s age. The inclusion of developmentally appropriate measures of social capital in a longitudinal study would also tap into the continued influence of parenting and parent-child relationships as adolescents become young adults. Future research examining differences in family structure should continue to employ the methods of this study when placing participants into categories. Findings show that the influence of family structure on educational outcomes goes beyond a simple comparison of intact two-parent families and stepfamilies, and that coming from each distinct type of stepfamily increases the risk of poor outcomes at different times and at different magnitudes. Although detailed information on parents’ marital status was not available, findings also show this is an important factor to consider. Specifically, the negative impact of living in a cohabiting stepfather family on academic outcomes is significantly stronger than that of living in a married stepfather family (Tillman, 2008).

APPENDIX A

TABLES

Table 1. Youth Demographic Characteristics (N = 3,334) Demographic Characteristics N % M SD Gender Male 1,773 53.2 Female 1,561 46.8 Age at Baseline (in years) 13.98 .82 Race Non-Hispanic White 1,944 58.3 African American 547 16.4 Hispanic 733 22.0 Other 110 3.3 Family Structure at Baseline Two-parent Intact 2,594 76.5 Simple Stepfather 284 8.5 Complex Stepfamily 88 2.6 Blended Stepfamily 413 12.4 Number of Siblings at Baseline 0 16.1 1 40.9 2 26.3 3 11.2 4 + 5.5 Household Income at Baseline $57,302 $43,538 Resident Step/Mother’s Education (in years) 12.7 3.1 Resident Step/Father’s Education (in years) 12.8 3.3

Table 2. Bivariate Correlations of Study Variables (N = 3,334) Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 1. Female - 2. Age -.01 - 3. Non-white .01 .00 -

4. # of Siblings .01 -.07** -.14** - 5. Income .01 .03 -.28** -.11** -

6.Mother’s education -.01 -.00 .33** -.21** .44** -

7.Father’s education .02 .02** .33** -.16** .47** .66** - 8. Retention -.11** .05** .13** .04* .17** .14** .16** - 9. Delinquency -.21** .13** -.01 .01 -.04** -.03** -.05** .10** -

10. Early first sex -.05** -.12** -.11** .01 -.14** -.11** -.14** .12** .15** - 11. Learning dis. -.12** -.00 .08** -.01 .02 .01 -.01 .15** .09** .03† - 12.Transitions .02 -.12** -.06** .02 -.14** -.07** -.09** .09** .09** .20** .03† -

13. School change -.05** -.20** -.08** .05** -.11** -.10** -.12** .21** .11** .18** .08** .22** - 14. Public school .01 .01 .08** -.05** .15** .13** .12** -.06** -.03† -.07** -.01 -.03 -.06** 15. Simple stepfather .02 -.02 -.02 -.14** -.05* -.01 -.00 .03* .03† .06** .02 .18** .07**

16. Complex -.01 .01 -.01* .08** -.01* .01 .00 .06** .04* .02 .01 .11** .03* 17. Blended -.01 -.03† -.06** .18** -.09** -.06** -.07** .09** .12** .09** -.01 .16** .11** 18. Stepchild -.01 -.01 .03 .05* -.11** -.03† -.03† .10** .13** .11** .02** .30** .16**

18. P Monitoring .04* -.12** -.11** -.03 .11** .09** .13** -.06** -.32** -.12** -.05 -.11** -.10** 19. P-C relationship -.01 -.11** -.06** -.02 .10** .09** .12** -.08** -.21** -.11** -.05** -.12** -.08** 20. P-S Involvement .01 -.11** -.03† -.09** .19** .21** .20** -.12** -.11** -.03 -.04† - .07** - .07**

21. Highest grade .03* .03† -.06** -.01 .10** .15** .15** -.12** -.10** -.12** -.02 -.06** -.08** 22.Highest degree .13** .07** -.19** -.12** .34** .38** .40** -.30** -.21** -.28** -.11** -.20** -.27** 23. HS/GED .05** .03 -.08** -.09** .15** .23** .21** -.21** -.10** -.17** -.06** -.12** -.14**

24. Post-secondary .12** .00 -.09** -.11** .27** .30** .32** -.27** -.15** -.21** -.11** -.15** -.20** 25. Associates .12** .06** -.18** -.08** .29** .32** .34** -.22** -.15** -.21** -.10** -.14** -.20** 26. Bachelors .12** .07** -.18** -.08** .30** .33** .35** -.21** -.16** -.20** -.09** -.14** -.19**

*p ≤ .05; **p ≤ .01; † p ≤ .10 46

Table 2. (cont.) Variables 14 15 16 17 18 19 20 21 22 23 24 25 26 27 14. Private school - 15. Simple stepfather -.04* - 16. Complex -.01 -.05** -

17. Blended -.06** -.12** -.06** -

18. Stepchild -.06** .60** .32** .51** - 19. P monitoring -.11** -.09** -.02 -.11** -.14** - 20. P-C relationship .05** -.07** -.02 -.12** -.14** .58** -

21. P-S Involvement -.21** .06** .02 .11** .13** -.17** -.14** - 22. Highest grade .02 -.02 -.02 -.07** -.07** .09** .11** .09** - 23. Highest degree .11** -.11** -.01 .18** -.20** .18** .18** .17** .31** - 24. HS/GED .06** -.05** -.01 -.08** -.09** .07** .07** .10** .20** .56** -

25. Post-secondary .10** -.07** -.03† -.13** -.16** .13** .14** .18** .26** .61** .41** - 26. Associates .13** -.08** -.02 -.16** -.17** .15** .16** .15** .28** .84** .24** .56** - 27. Bachelors .15** -.09** -.01 -.15** -.17** .15** .14** .16** .28** .83** .20** .47** .86** -

*p ≤ .05; **p ≤ .01; † p ≤ .10; Note:

Table 3. Results of Confirmatory Factor Analysis: Social Capital Items Standardized Coefficients M1 M2 M3 M4 Factors SC M R M R I M R I SC Parental monitoring Knows about your close friends, that is, who they are? .08 .08 .88 .88 Knows about your close friends’ parents, that is, who they are? .12 .12 .79 .79 Knows about who you are with when you are not at home? .08 .08 .61 .61 Knows about who your teachers are and what you are doing in school? .14 .15 .54 .54 Parent-child relationship I think highly of him/her. .07 .74 .73 .73 S/he is a person I want to be like. .07 .81 .81 .81 I really enjoy spending time with him/her. .06 .81 .80 .80 Praises you for doing well. .09 .69 .71 .71 Criticizes you or your ideas.* .01 .37 .38 .38 Helps you do things that are important to you. .10 .72 .74 .74 Blames you for his/her problems.* .07 .44 .45 .45 Makes plans with you and cancels for no good reason.* .05 .40 .41 .41 Parent-school involvement Attended meetings of the parent-teacher organization at youth’s school? 1.00 1.00 1.00 1.00 Volunteered to help at the school or in the classroom? -.36 .36 .36 .36 Social Capital (Second-order) Parental monitoring .67 Parent-child relationship quality .85 Parent-school involvement .14 Note: M1 = One-factor Model; M2 = Two-factor Model; M3 = Three-factor Model; M4 = Second-order Factor Model; SC = Social Capital; M = Parental Monitoring; R = Parent- child Relationship; I = Parent-school Involvement *Reverse coded prior to analysis.

Table 4. Summary of Fit Indices for One-, Two-, Three-, and Second-order factor Models of Social Capital Model χ2 df P CFI RMSEA SRMR One-factor 7.14 78 .000 .066 .094 .280 Two-factor 6.66 77 .000 .541 .067 .196 Three-factor 6.49 75 .000 .850 .039 .068 Second-order factor 6.49 75 .000 .850 .039 .068 Second-order respecified 6.29 67 .000 .986 .013 .032

χ2 = 6.29; df = 63 CFI = .986 RMSEA = .013 SRMR = .032

.985** .728** .144**

.681** .726** .602** .674** .372** .446** 1.0 .361**

.596** .690** .699** .759** .798** .410**

Figure 1. Second-order Social Capital CFA Model with Correlated Measurement Error

Table 5. Crosstabulations by Gender, Race, and Educational Outcomes for Full Sample and Family Structure Type Intact Simple Complex Blended Full Sample Two-parent Stepfather Stepfamily Stepfamily (N = 3,334) (n = 2,549) (n = 284) (n = 88) (n = 413) Variables N % N % N % N % N % Χ2 df p Gender Male 1,775 53.2 1,361 53.4 141 49.6 50 56.8 223 54.0 Female 1,559 46.8 1,188 46.6 143 50.4 38 43.2 190 46.0 2.04 3 .563 Race White 1943 58.3 1,522 59.7 157 55.2 54 61.4 210 50.8 Black 547 16.4 352 13.8 76 26.8 14 15.9 105 25.4 Hispanic 733 22.0 581 22.8 48 16.9 17 19.3 87 21.1 Other 111 3.3 72 2.8 3 1.1 3 3.4 11 2.6 76.64 15 .000 Number of Siblings 0 535 16.1 416 16.3 91 32.0 3 3.4 25 6.1 1 1,364 40.9 1,126 44.2 107 37.7 15 17.0 116 28.1 2 878 26.3 648 25.4 61 21.5 40 45.5 129 31.2 3 372 11.2 233 9.1 18 6.3 21 23.9 10 2.4 4 + 185 5.5 126 5.0 7 2.5 9 10.2 43 10.4 254.06 12 .000 Educational Outcomes Dropout prior to high school completion 701 21.0 432 16.9 95 33.4 24 27.3 150 36.3 112.17 3 .000 Enroll in any school following high school 269 38.4 178 41.2 33 34.7 6 25.0 52 34.7 4.68 3 .197 dropout (not including GED) Complete sigh school/GED 2,978 89.3 2,341 91.8 241 84.9 77 87.5 346 83.8 30.75 3 .000 Enroll in post-secondary schooling 2,058 61.7 1,685 66.1 142 50.0 46 52.3 185 44.8 90.66 3 .000 Complete any post-secondary 1,125 33.7 981 38.5 62 21.8 25 28.4 57 13.8 120.01 3 .000 degree

Complete at least a bachelor’s 912 27.4 811 31.8 39 13.7 22 25.0 40 9.6 118.60 3 .000 degree Note: The percentages are for within the family structure types. Percentage of “Enroll in any school following high school dropout” is out of the category total for “Dropout prior to high school completion”

Table 6. Analysis of Variance Results by Family Structure Type (N = 3,334) Intact Two-Parenta Simple Stepfatherb Complex Stepfamilyc Blended Stepfamilyd (n = 2,549) (n = 284) (n = 88) (n = 413) Educational Outcomes M SD M SD M SD M SD F p Household income 60,038.34bd 45,098.97 50,081.78a 39,689.51 53,972.97 34,600.73 45,842.68c 34,735.44 11.80 .000 Mother’s education 12.81d 2.91 12.63 2.69 12.90 2.93 12.22a 2.59 4.90 .004 Father’s education 12.94d 3.50 12.80d 12.80 12.85 2.82 12.17ab 2.78 5.92 .001 Total years completed 14.27c 2.43 13.16c 2.23 13.70ab 2.69 12.90 2.12 51.76 .000 Highest degree 2.54c 1.30 1.94 1.76 2.25a 1.43 1.79 1.12 53.62 .000 completed Note: Subscripts represent groups significantly different within row at p < .10 in the Tukey’s HSD post hoc comparison.

Table 7. Random-Intercept Logistic Regressions for Educational Outcomes – Full Sample (N = 3,334 ) Complete High Enter Post-secondary Complete Any Post- High School Dropout Enroll Following Dropout Complete BA or Higher School/GED School secondary Degree B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR Fixed-effects variables Female -.26 .13 .77† .22 .20 1.25 .08 .15 1.07 .50 .11 1.64** .55 .11 1.73** .64 .13 1.89** Age .11 .08 1.11 -.20 .13 .82 .03 .09 1.03 -.11 .07 .89† .17 .07 1.18** .25 .07 1.28** Non-White -.26 .14 .77† .29 .21 1.34 .26 .17 1.30 .42 .12 1.52** -.22 .12 .81* -.21 .13 .81† # of Siblings .18 .06 1.20** -.05 .08 .95 -.09 .06 .92 -.09 .05 .92† -.03 .05 .97 -.02 .05 .98 Income -.06 .02 .94** .05 .04 1.06† .03 .03 1.03 .06 .02 1.06** .04 .01 1.04** .05 .02 1.04** Mother’s education -.11 .03 .89** .06 .04 1.06 .17 .03 1.19** .15 .03 1.16** .14 .03 1.15** .19 .03 1.20** Father’s education -.12 .03 .88** .07 .04 1.07† .09 .04 1.09* .13 .02 1.13** .14 .02 1.15** .16 .03 1.17** Retention 1.36 .18 3.89** -.37 .21 .69† -1.09 .18 .33** -1.19 .17 .30** -1.28 .21 .28** -1.56 .27 .21** Delinquency .24 .04 1.27** -.01 .05 .99 -.12 .04 .90† -.09 .04 .91* -.15 .04 .86** -.21 .05 .81** Early first sex 1.24 .16 3.45** -.09 .19 .92 -.77 .16 .46** -.70 .13 .50** -.83 .15 .44** -.88 .17 .41** Learning disable .30 .24 1.35 .12 .31 1.12 -.26 .26 .77 -.77 .22 .46** -1.04 .27 .35** -1.16 .31 .31** Structure .39 .11 1.47** -.10 .13 .90 -.23 .11 .80* -.21 .09 .81* -.11 .11 .90 -.18 .13 .83 transitions School changes .53 .07 1.70** .13 .06 1.14* -.14 .06 .87* -.21 .05 .81** -.27 .06 .76** -.27 .07 .76** Private school .15 .25 1.16 -.03 .39 .97 .33 .33 1.39 .39 .19 1.48* .49 .17 1.63** .68 .19 1.98** Family structure Stepfather .98 .23 2.67** -.20 .30 .82 -.49 .24 .61† -.57 .19 .56** -.60 .21 .53** -.96 .26 .38** Complex .01 .39 1.02 -.94 .71 .39 .03 .42 1.02 -.34 .33 .71 -.15 .36 .97 .06 .37 1.06 Blended .61 .20 1.85** -.54 .28 .58† -.47 .20 .63* -.57 .16 .57** -1.09 .21 .36** -1.22 .25 .29** Social capital Monitoring -.01 .02 .99 .04 .03 1.05 .00 .03 1.00 .01 .02 1.00 .00 .02 1.00 .03 .02 1.03 PC relationship -.03 .02 .98 -.01 .03 .99 .01 .02 1.01 .03 .01 1.03† .04 .02 1.04* .02 .02 1.02* PS involvement .01 .03 1.01 .08 .05 1.08 .04 .04 1.04 .00 .03 1.00 -.01 .03 .99 .00 .03 1.00 Interactions Stepf*Monitor .09 .08 1.09 .15 .11 1.16 -.04 .10 .96 -.03 .07 .97 .07 .08 1.07 .16 .10 1.18 Stepf*Relate .07 .06 1.07 .06 .09 1.06 .03 .07 1.03 .04 .05 1.04 -.01 .06 .99 -.03 .06 .97 Stepf*Involve -.11 .11 .89 -.20 .16 .82 .22 .14 1.24 .09 .32 1.09 -.02 .11 .98 .05 .13 1.04 Complex* -.06 .15 .94 .02 .21 1.02 .05 .18 1.05 .01 .12 1.01 .04 .14 1.04 .02 .16 1.02 Monitor

Table 7 - continued Complete High Enter Post-secondary Complete Any Post- High School Dropout Enroll Following Dropout Complete BA or Higher School/GED School secondary Degree B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR Complex*Relate .17 .10 1.18 .17 .19 1.19 .02 .12 1.02 -.02 .09 .98 -.06 .11 .95 -.06 .12 .96 Complex*Involve -.06 .20 .95 .61 .43 1.85 .22 ..23 1.25 .28 .68 1.32† .22 .18 1.24 .31 .20 1.36 Blended*Monitor .04 .06 1.04 .03 .10 1.03 -.01 .07 1.00 .01 .06 1.01 -.04 .07 .96 -.02 .09 .98 Blended*Relate .09 .05 1.09 -.14 .07 .87* -.07 .05 .93 -.01 .04 .99 -.01 .05 .99 -.03 .06 .97 Blended*Involve .01 .09 1.01 -.17 .13 .85 .10 .11 1.11 .03 .35 1.02 .14 .10 1.15 .30 .12 1.36* Constant -1.70 1.31 -.13 1.90 -.10 1.50 -.03 -1.37 1.07 .43 -7.18 1.29 1.87** -9.60 1.49 1.87** Random effects

σu0 1.29 .28 .73 .61 .96 .39 1.17 .26 1.07 .26 1.07 .26 ICC (ρ) .33 .10 .14 .20 .23 .14 .30 .09 .26 .09 .26 .09 Model F-test 4.81** .84 3.41** 4.84** 5.41** 4.49** ** p < .01; *p < .05; † p < .10

Table 8. Random-Intercept Linear Regressions for Educational Outcomes – Full Sample (N = 3,334) Total Years Completed Highest Degree Completed B SE B β B SE B β Fixed-effects variables Female .36 .07 .07** .19 .04 .07** Age .15 .04 .05** .07 .02 .04** Non-White .14 .08 .03† -.02 .04 -.01 # of siblings -.05 .03 -.02* -.03 .02 -.03 Income .03 .01 .05** .02 .05 .07** Mother’s education .15 .02 .19** .07 .01 .17** Father’s education .12 .02 .17** .06 .01 .15** Retention -1.04 .10 -.15** -.53 .06 -.15** Delinquency -.12 .02 -.08** -.08 .01 -.09** Early first sex -.64 .08 -.12** -.39 .05 -.13** Learning disable -.59 .14 -.06** -.27 .08 -.05** Structure transitions -.15 .07 -.04* -.11 .03 -.05** School changes -.19 .03 -.09** -.11 .02 -.10** Private school .49 .12 .06** .23 .07 .05** Family structure Stepfather -.66 .13 -.08* -.37 .07 -.08** Complex -.14 .24 -.01 .00 .12 .00 Blended -.69 .12 -.09** -.33 .06 -.08** Social capital .00 Monitoring .00 .01 .00 .00 .01 .00 PC relationship .03 .01 .11** .01 .01 .07* PS involvement .00 .02 .00 .00 .01 .00 Interactions Stepf*Monitor .07 .05 .03 .01 .03 .01 Stepf*Relate -.01 .04 -.01 -.05 .02 -.05 Stepf*Involve .03 .25 .00 .04* .04 .01 Complex*Monitor .16 .09 .03† .04 .05 .01 Complex*Relate -.08 .06 -.02 -.03 .04 -.02* Complex*Involve -.12 .53 .00 .15 .06 .01 Blended*Monitor -.05 .04 -.02 -.02 .02 -.02 Blended*Relate -.02 .03 -.01 -.02 .02 -.02 Blended*Involve .00 .25 .00 .05 .03 .02 Constant 8.52 .74 -.08 .40 Random effects σu0, σe0 1.17 1.62 .52 .93 ICC (ρ) .35 .26 Model F-test 51.52** 51.66** ** p < .01; *p < .05; † p < .10; Note: β represents calculated standardized coefficients

Table 9. Random-Intercept Logistic and Linear Regressions for Educational Outcomes – Biological Children Only (N =2,650 ) Complete High Enter Post-secondary Complete Any Post- High School Dropout Enroll Following Dropout Complete BA or Higher School/GED School secondary Degree SE B SE B OR B SE B OR B OR B SE B OR B SE B OR B SE B OR B Fixed-effects variables Female -.15 .17 .86 .30 .24 1.35 -.02 .18 .98 .46 .12 1.58** .57 .13 1.77** .61 .14 1.84** Age .05 .10 1.05 -.14 .15 .87 .15 .11 1.17 -.06 .07 .93 .17 .07 1.19* .23 .08 1.26** Non-White -.11 .18 .90 .20 .26 1.23 .16 .20 1.18 .28 .13 1.32* -.29 .13 .75* -..29 .14 .75* # of siblings .18 .07 1.20* -.02 .09 .98 -.08 .07 .92 -.07 .05 .93 .01 .05 1.01 .00 .06 1.00 Income -.05 .03 .95 .06 .05 1.06 .01 .03 1.01 .06 .02 1.06* .04 .02 1.04* .04 .02 1.04* Mother’s education -.12 .04 .88** .02 .05 1.02 .14 .04 1.15** .11 .03 1.12** .14 .03 1.15** .20 .04 1.22** Father’s education -.11 .04 .90** .10 .05 1.10† .11 .04 1.12** .13 .03 1.15** .15 .03 1.16** .15 .03 1.16** Retention 1.58 .24 4.87** -.45 .26 .64† -1.18 .27 .30** -1.25 .19 .29** -1.25 .24 .29** -1.58 .31 .20** Delinquency .21 .06 1.23** -.08 .08 .93 -.16 .06 .85** -.11 .04 .89* -.14 .05 .87** -.20 .06 .82** Early first sex 1.32 .20 3.76** -.09 .24 .91 -.68 .19 .51** -.66 .14 .52** -.85 .16 .43** -.82 .18 .44** Learning disable .70 .30 2.00* .20 .36 1.22 -.37 .31 .68 -1.01 .24 .37** -1.31 .31 .27** -1.34 .35 .26** Structure .43 .16 1.54** -.08 .19 .93 -.32 .15 .72* -.22 .12 -.33 .15 .72* -.29 .17 transitions .80† .75† School changes .58 .09 1.78** .17 .08 1.18* -.07 .07 .93 -.17 .06 .85** -.30 .07 .74** -.28 .07 .76** Private school .09 .30 1.10 .18 .49 1.20 .66 .43 1.94 .45 .21 1.57* .56 .19 1.75** .68 .20 1.96** Blended family .22 .39 1.25 -.17 .55 .85 -.16 .39 .85 -.02 .30 .98 -.78 .36 .46* -.91 .43 .40* Social capital Monitoring -.02 .03 .98 .03 .04 1.03 .00 .03 1.00 .00 .02 1.00 -.02 .03 .98 -.01 .03 1.00 PC relationship -.02 .02 .98 -.02 .04 1.00 .00 .02 1.00 .02 .02 1.02 .04 .02 1.05* .03 .02 1.03 PS involvement .04 .05 1.04 .09 .07 1.09 -.01 .05 .99 -.03 .03 .97 -.01 .03 .98 .00 .04 1.00 Interactions Blended*Monitor .09 .14 1.10 .08 .21 1.08 .01 .14 1.01 -.02 .11 .98 -.11 .14 .89 -.017 .16 .84 Blended*Relate .24 .12 1.27* -.07 .24 .93 -.11 .11 .89 .03 .08 1.03 .01 .10 1.01 .02 .11 1.02 Blended*Involve .17 .20 1.18 -.10 .26 .91 -.11 .21 .90 -.06 .15 .94 .15 .18 1.16 .43 .21 1.53* Constant .09 .14 -.35 2.4 -1.34 1.74 -1.64 1.18 -7.28 1.39 -9.18 1.60 Random effects

σu0 1.47 .33 .75 .73 .94 .50 1.09 .28 1.16 .28 1.24 .31 ICC (ρ) .39 .11 .15 .24 .21 .18 .27 .10 .29 .10 .32 .11

Table 9 - continued Complete High Enter Post-secondary Complete Any Post- High School Dropout Enroll Following Dropout Complete BA or Higher School/GED School secondary Degree SE B SE B OR B SE B OR B OR B SE B OR B SE B OR B SE B OR B Model F-test 4.68** .73 3.71** 5.89** 5.93** 5.04** ** p< .01; *p< .05; †p< .10

Table 10. Random-Intercept Linear Regressions for Educational Outcomes – Biological Children Only (N = 2,650) Total Years Completed Highest Degree Completed B SE B β B SE B β Fixed-effects variables Female .33 .08 .07** .18 .04 .07** Age .17 .05 .06** .09 .03 .06** Non-White .08 .09 .02 -.07 .05 -.03 # of siblings -.02 .04 -.01 -.02 .02 -.02 Income .02 .01 .04† .01 .01 .03* Mother’s education .13 .02 .17** .06 .01 .15** Father’s education .13 .02 .19** .06 .01 .16** Retention -1.13 .12 -.16** -.55 .07 -.14** Delinquency -.13 .03 -.07** -.08 .02 -.08** Early first sex -.59 .10 -.10** -.35 .05 -.11** Learning disable -.79 .17 -.08** -.35 .09 -.07** Structure transitions -.21 .09 .00* -.17 .05 .00** School changes -.20 .04 -.09** -.11 .02 -.10** Private school .52 .14 .07** .26 .07 .06** Blended family -.28 .22 -.02 -.22 .11 -.03* Social capital Monitoring -.01 .02 -.01 .00 .01 .00 PC relationship .04 .01 .07** .01 .01 .03* PS involvement -.01 .02 .00 -.01 .01 .00 Interactions .00 .00 Blended*Monitor -.07 .08 -.02 -.04 .04 -.02 Blended*Relate .02 .06 .01 -.02 .03 -.01 Blended*Involve .00 .11 .00 .04 .06 .01 Constant 8.07 .81 -.36 .44 Random effects (σ) σu0, σe0 1.59 2.49 .32 .84 ICC (ρ) .39 .27 Model F-test 51.80** 51.73** ** p < .01; *p < .05; †p < .10 Note: β represents calculated standardized coefficients

Table 11. Random-Intercept Logistic Regressions for Educational Outcomes – Within Blended Families (N = 413 ) Complete High Enter Post-secondary Complete Any Post- High School Dropout Enroll Following Dropout Complete BA or Higher School/GED School secondary Degree B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR Fixed-effects variables Female -.24 .26 .79 2.05 2.16 7.78 .53 .35 1.70 1.05 .73 2.87 .66 .34 1.93* 1.05 .42 2.87* Age .08 .17 1.09 -2.77 1.43 .06* -.05 .22 .96 -.50 .41 .63 -.07 .21 .93 .06 .25 1.06 Non-White -.73 .28 .48** 4.81 2.14 123.32* .90 .38 2.46* .78 .68 2.19 -.30 .34 .74 -.13 .41 .88 # of siblings .03 .12 1.03 -1.91 1.08 .15† -.10 .15 .91 -.12 .26 .89 -.01 .15 .99 .05 .17 1.06 Income -.06 .05 .94 .27 .34 1.30 .03 .06 1.03 .12 .10 1.13 .09 .05 1.09 .12 .06 1.13* Mother’s education -.05 .06 .95 .90 .47 2.44 .22 .08 1.24* .42 .21 1.52* .18 .09 1.20* .12 .10 1.13 Father’s education -.13 .07 .88† -.15 .42 .86 -.04 .07 .96 .03 .13 1.03 -.02 .08 .98 .02 .09 1.02 Retention .83 .32 2.30** .88 2.07 2.42 -.89 .39 .41* -2.52 1.22 .08* -.79 .53 .45 -.56 .62 .57 Delinquency .21 .07 1.24** .56 .47 1.74 .03 .08 1.03 -.12 .18 .89 -.26 .13 .77* -.29 .16 .74† Early first sex 1.20 .29 3.33** -.91 1.86 .41 -1.02 .39 .36** -1.60 .84 .20* -1.20 .43 .30** 1.72 .65 .18** Learning disable -.20 .53 .82 -5.87 3.89 .00 -.34 .59 .71 -.82 1.40 .44 ------Structure .46 .18 1.58* .81 1.18 2.24 -.11 .22 .89 -.11 .40 .90 .15 .24 1.16 -.30 .35 .74 transitions School changes .41 .10 1.51** .73 .64 2.07 -.24 .12 .79* -.17 .23 .83 -.11 .14 .89 -.12 .17 .88 Private school .06 .61 1.06 -9.20 7.02 .00 -.38 .70 .68 .90 1.38 2.46 -.98 .89 .37 -.41 .91 .66 Stepchild status .33 .36 1.39 -1.69 2.69 .18 .00 .43 1.00 -1.17 .77 .31 -.36 .38 .69 -.24 .46 .78 Social capital Monitoring .01 .09 1.01 .63 .76 1.88 .04 .06 1.04 -.05 .10 .94 -.04 .11 .96 -.04 .12 .96 PC relationship .14 .07 1.15* -.87 .78 .42 -.08 .07 .92 .08 .12 1.08 .01 .07 1.01 .02 .07 1.00 PS involvement .14 .12 1.15 -.55 .72 .58 .25 .36 1.27 -.05 .60 .96 .10 .12 1.10 .29 .15 1.33* Interactions Step*Monitor -.50 .11 .95 -.40 .83 .67 -.01 .16 .99 .14 .22 1.15 .02 .15 1.02† .10 .16 1.10 Step*Relate -.14 .08 .87† -.14 .85 1.15 .03 .01 1.03 -.10 .15 .91 .01 .10 1.01 -.04 .10 .96 Step*Involve -.22 .17 .80 -.14 1.09 .87 -.32 .86 .72 -.58 1.41 .56 -.03 .19 .97 -.17 .22 .84 Constant -5.53 3.12 43.45 27.18 2.03 4.13 7.66 .39 6.09 1.48 -2.57 3.16 -5.66 4.24 Random effects

σu0 .02 2.64 8.21 2.28 .53 1.76 3.21 1.52 .00 .29 .01 .57 ICC (ρ) .00 .04 .95 .02 .08 .48 .76 .17 .00 .00 .00 .00

Table 11 - continued Complete High Enter Post-secondary Complete Any Post- High School Dropout Enroll Following Dropout Complete BA or Higher School/GED School secondary Degree B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR B SE B OR Model F-test 2.48** .98 .78 .35 1.56* 1.67* ** p < .01; *p < .05; †p < .10

Table 12. Random-Intercept Linear Regressions for Educational Outcomes – Blended Sample (N = 413) Total Years Completed Highest Degree Completed B SE B β B SE B β Fixed-effects variables Female .33 .20 .08† .27 .10 .11* Age .09 .13 .03 -.03 .07 -.02 Non-White .34 .21 .08 .16 .11 .07 # of siblings -.13 .09 -.11 -.02 .05 .05 Income .05 .04 .10 .03 .02 .04 Mother’s education .15 .04 .18** .07 .02 .16** Father’s education .02 .05 .03 .01 .02 -.17 Retention -.92 .25 -.19** -.36 .13 -.14** Delinquency -.13 .05 -.12* -.07 .03 -.14** Early first sex -.64 .21 -.15** -.48 .11 -.20** Learning disable -.17 .42 -.02 -.13 .22 -.03 Structure transitions -.11 .14 -.04 -.11 .07 -.07 School changes -.09 .07 -.06 -.10 .04 -.12** Private school .58 .46 .06 -.13 .24 -.02 Stepchild status -.37 .26 -.07 -.13 .13 -.04 Social capital Monitoring -.05 .05 -.08 -.01 .02 -.03 PC relationship .03 .04 .13 -.02 .01 -.08 PS involvement -.02 .09 -.04 .01 .0 .58 Interactions Step*Monitor .02 .07 .02 .01 .04 .04 Step*Relate -.03 .06 -.05 .01 .03 -.03 Step*Involve -.00 .12 -.34 .01 .06 -.17 Constant 10.36 2.15 2.26 1.12 .11 Random effects σu0, σe0 1.02 1.64 .15 .98 ICC (ρ) .28 .28 .03 Model F-test 5.29** 5.80** ** p < .01; *p < .05; †p < .10; Note: β represents calculated standardized coefficients

APPENDIX B

SOCIAL CAPITAL ITEMS

Social Capital Items Included in Confirmatory Factor Analysis Type Variable Question Stem Responses Latent Parent-child Please tell us whether you strongly 1. I think highly of him/her. 0 = strongly disagree Continuous* Relationship disagree, disagree, are neutral, 2. S/he is a person I want to be like. 1 = disagree Quality agree, or strongly agree with the 3. I really enjoy spending time with him/her. 2 = neutral following statements about {each 3 = agree parent}. 4 = strongly agree

Now we are going to list some 4. Praise you for doing well? 0 = never things that might describe 5. Criticize you or your ideas? 1 = rarely {parent}. Please tell us how often 6. Help you do things that are important to you? 2 = sometimes he/she does these things. How 7. Blame you for his/her problems? 3 = usually often does s/he: 8. Make plans with you and cancel for no good reason? 4 = always

Latent Parental Now we are going to list some 1. Know about your close friends, that is, who they are? 0 = knows nothing Continuous* Monitoring things that might describe 2. Know about your close friends’ parents, that is, who 1 = knows just a little {parent}. Please tell us how often they are? 2 = knows some things he/she does these things. How 3.How much does s/he know about who you are with 3 = knows most things much does s/he: when you are not at home? 4 = knows everything 4.How much does s/he know about who your teachers are and what you are doing in school?

Latent Parent-School 1 .In the last three years have you or your 1 = never Continuous* Involvement [spouse/partner] attended meetings of the parent-teacher 2 = sometimes organization at [youth’s] school? 3 = often 2. In the last three years have you or your 1 = never [spouse/partner] volunteered to help at the school or in 2 = sometimes the classroom? 3 = often

*= Reverse coded for ease of interpretation

APPENDIX C

HUMAN SUBJECTS APPROVAL LETTER

Office of the Vice President For Research Human Subjects Committee Tallahassee, Florida 32306-2742 (850) 644-8673 · FAX (850) 644-4392

EXEMPTION MEMORANDUM

Date: 2/27/2012

To: Chelsea Garneau

Dept.: FAMILY & CHILD SCIENCE

From: Thomas L. Jacobson, Chair

Re: Use of Human Subjects in Research Family Structure, Social Capital, and Educational Attainment: Progression from Secondary to Post-Secondary Schooling and Beyond

The application that you submitted to this office in regard to the use of human subjects in the proposal referenced above have been reviewed by the Secretary, the Chair, and one member of the Human Subjects Committee. The proposed research protocol is Exempt from human subjects regulations as described in per 45 CFR § 46.101(b)4.

The Human Subjects Committee has not evaluated your proposal for scientific merit, except to weigh the risk to the human participants and the aspects of the proposal related to potential risk and benefit. This memorandum does not replace any departmental or other approvals, which may be required.

The Committee expects that all relevant subject protection measures and ethical standards will be followed, as outlined in your proposal. No continuing review is required unless the nature of the project changes and it would affect the project exemption status.

You are advised that any change in protocol for this project that would affect the exemption status must be reviewed and approved by the Committee prior to implementation of the proposed change in the protocol. A protocol change/amendment form is required to be submitted for approval by the Committee. In addition, federal regulations require that the Principal Investigator promptly report, in writing any unanticipated problems or adverse events involving risks to research subjects or others.

By copy of this memorandum, the Chair of your department and/or your major professor is reminded that he/she is responsible for being informed concerning research projects involving human subjects in the department, and should review protocols as often as needed to insure that the project is being conducted in compliance with our institution and with DHHS regulations.

This institution has an Assurance on file with the Office for Human Research Protection. The

Assurance Number is FWA00000168/IRB number IRB00000446.

Cc: Kay Pasley, Advisor HSC No. 2012.7849

REFERENCES

ACT. (2010). 2010 Retention/completion summary tables. Retrieved April 10, 2012 From: http://www.act.org/research/policymakers/pdf/10retain_trends.pdf

Acock. A. C. (2005). Working with missing values. Journal of Marriage and Family Therapy, 67, 1012-1028.

Alexander, K. L., Entwisle, D. R., & Kabbani, N. S. (2001). The dropout process in life course perspective: Early risk factors at home and school. Teachers College Record, 103, 760- 822.

Anguiano, R. P. V. (2004). Families and schools: The effect of parental involvement on high school completion. Journal of Family Issues, 25, 61-85. doi:10.1177/0192513X03256805

Astone, N. M., & McLanahan, S. S. (1991). Family structure, parental practices, and high school completion. American Sociological Review, 56, 309-320. doi: 10.2307/2096106

Becker, G. S. (1964/1993). Human capital. Chicago: University of Chicago Press.

Black, S. E., Devereux, P. J., & Salvanes, K. G. (2005). The more the merrier? The effect of family size and birth order on children’s education. Quarterly Journal of Economics, 120, 669-700.

Booth, A. L., & Kee, J. H. (2009). Birth order matters: The effect of family size and birth order on educational attainment. Journal of Population Economics, 22, 367-397.

Brewer, A. (1984). A guide to Marx’s capital. Cambridge, MA: Cambridge University Press.

Brooks-Gunn, J., Guo, G., & Furstenberg, Jr., F. F. (1993). Who drops out of and who continues beyond high school? A 20-year follow-up of black urban youth. Journal of Research on Adolescence, 3, 271-294.

Buchner, A., Erdfelder, E., Faul, F., & Lang, A. (2009). G*Power (Version 3.1.2)[Computer program]. http://www.psycho.uni-duesseldorf.de/aap/projects/gpower/

Bureau of Labor Statistics (BLS). (2011). College enrollment and work activity of 2010 high school graduates. Retrieved April 11, 2012 from: http://www.bls.gov/news.release/hsgec.nr0.htm

Cancian, M., & Meyer, D. R. (1998). Who gets custody? Demography, 35, 147- 157.

Carlin, J. B., Galati, J. C., & Royston, P. (2008). A new framework for managing and analyzing multiply imputed data in Stata. Stata Journal, 8, 49-67.

Carson, R., Frank, A. R., & Sitlington, P. L. (1992). Adult adjustment among high school

graduates with mild disabilities. Exceptional Children, 59, 221 – 233.

Case, A., Lin, I., & McLanahan, S. (2001). Educational attainment of siblings in stepfamilies. Evolution and Human Behavior, 22, 269-289. doi: 10.1016/S1090-5138(01)00069-1

Cavanagh, S. E., Schiller, K. S., & Riegle-Crumb, C. (2006). Marital transitions, parenting, and schooling: Exploring the link between family-structure history and adolescents’ academic status. Sociology of Education, 79, 329-354. doi: 10.1177/07419325070280060201

Coleman, J. S. (1988). Social capital in the creation of human capital. The American Journal of Sociology, 94, S95-S120. doi: 10.1086/228943

Cheadle, J. E., & Goosby, B. J. (2010). Birth weight, cognitive development, and life chances: A comparison of siblings from childhood into early adulthood. Social Science Research, 39, 570-584.

Chatterjee, S., & Hadi, A. S. (2006). Regression analysis by example (4th ed.). Hoboken, NJ: John Wiley & Sons.

Christle, C. A., Jolivette, K., & Nelson, C. M. (2007). School characteristics related to high school dropout rates. Remedial and Special Education, 28, 325-339. doi: 10.1177/07419325070280060201

Crawford, J. R. & Henry, J. D. (2003). The Depression Anxiety Stress Scales: Normative data and latent structure in a large non-clinical sample. British Journal of Clinical Psychology, 42, 111-131.

Croiseau, P., Genin, E., & Cordell, H. J. (2007). Dealing with missing data in family-based association studies: A multiple imputation approach. Human Heredity, 63, 229-238.

Day, J. C., & Newburger, E. C. (2002). The big payoff: Educational attainment and synthetic estimates of work-life earnings. Current Population Reports, P23-210, Washington, DC: U.S. Census Bureau.

Dika, S. L., & Singh, K. (2002). Applications of social apital in educational literature: A critical synthesis. Review of Educational Research, 72, 31-60.

Downs, K. J. M. (2004). Family commitment, role perceptions, social support, and mutual children in remarriage. Journal of Divorce & Remarriage, 40(1/2), 35-53.

Dunn, J., Davies, L. C., O’Connor, T. G., & Sturgess, W. (2000). Parents’ and partners’ life course and family experiences: Links with parent-child relationships in different family settings. Journal of Child Psychology and Psychiatry, 41, 955-968. doi: 10.1111/1469- 7610.00684

Education Commission of the States (ECS). (2010). Compulsory school age requirements. Retrieved April 11, 2012 from: http://www.ncsl.org/documents/educ/ECSCompulsoryAge.pdf

Englund, M. M., Egeland, B., & Collins, W. A. (2008). Exceptions to high school dropout predictions in a low-income sample: Do adults make a difference? Journal of Social Issues, 64, 77-93.

Fisher, P. A., Leve, L. D., O’Leary, C. C., & Leve, C. (2003). Parental monitoring of children’s behaviors: Variation across stepmother, stepfather, and two-parent biological families. Family Relations, 52, 45-52. doi: 10.1111/j.1741-3729.2003.00045.x

Fletcher, A. C., Darling, N., & Steinberg, L. (1995). Parental monitoring and peer influences on adolescent substance use. In J. McCord (Ed.), Coercion and punishment in ;ong-term perspectives (pp. 259-271). Cambridge, MA: Cambridge University Press.

Flouri, E. (2006). Parental interest in children's education, children's self-esteem and locus of control, and later educational attainment: Twenty-six year follow-up of the 1970 British birth cohort. British Journal of Educational Psychology, 76, 41-55. doi: 10.1348/000709905X52508

Flouri, E., & Buchanan, A. (2002). What predicts good relationships with parents in adolescence and partners in adult life: Findings from the 1958 British birth cohort. Journal of Family Psychology, 16, 186-198.

Fox, G. L. & Bruce, C. (2004). Conditional fatherhood: Identity theory and parental investment theory as alternative sources of explanation of fathering. Journal of Marriage and Family, 63, 394-403.

Gennetian, L. A. (2005). One or two parents? Half or step siblings? The effect of family structure on young children’s achievement. Journal of Population Economics, 18, 415-436. doi: 10.1007/s00148-004-0215-0

Ginther, D. K., & Pollak, R. A. (2004). Family structure and children’s educational outcomes: Blended families, stylized facts, and descriptive regressions. Demography, 41, 671-696. doi: 10.1353/dem.2004.0031

Gordon, M.S. & Cui, M. (in press). The effect of school-specific parenting on academic achievement in adolescence and young adulthood. Family Relations.

Gulliford, M. C., Ukoumunne, O. C., & Chinn, S. (1999). Components of varlance and intraclass correlations for the design of community-based surveys and intervention studies. American Journal of Epidemiology, 149, 876-883.

Guo, S., & Fraser, M. W. (2010). Propensity score analysis: Statistical methods and applications. Thousand Oaks, CA: Sage.

Ham, B. D. (2004). The effects of divorce and remarriage on the academic achievement of high school seniors. Journal of Divorce & Remarriage, 42, 159-178.

Hamer, J., & Marchioro, K. (2002). Becoming custodial dads: Exploring parenting among low- income and working-class African American fathers. Journal of Marriage and Family, 64, 116-129.

Hauser, R. M., & Andrew, M. (2006). Another look at the stratification of educational transitions: The logistic response model with partial proportionality constraints. Sociological Methodology, 36, 1-26.

Heard, H. E.(2007). Fathers, mothers, and family structure: Family trajectories, parent gender, and adolescent schooling. Journal of Marriage and Family, 69, 435-450. doi: 10.1111/j.1741-3737.2007.00375.x

Heaton, T. B. (2002). Factors contributing to increasing marital stability in the United States. Journal of Family Issues, 23, 392-409.

Hetherington, E. M., Cox, M., & Cox, R. (1985). Effects of divorce on parents and children. In M. E. Lamb (Ed.), Nontraditional families (pp. 233-288). Hillsdale, NJ: Erlbaum.

Hetherington, E. M., & Jodl, K. M. (1994). Stepfamilies as settings for child development. In A. Booth & J. Dunn (Eds.), Stepfamilies: Who benefits? Who does not? (pp. 55-79). Hillsdale, NJ: Lawrence Erlbaum.

Hetherington, E. M., & Stanley-Hagan, M. (2000). Diversity among stepfamilies. In D. H. Demo, K. R. Allen, & M. A. Fine (Eds.), Handbook of family diversity (pp. 173-196). New York, NY: Oxford University Press.

Herzog, A. R., Markus, H. R., Franks, M. M., & Holmberg, D. (1998). Activities and well-being in older age: Effects of self-concept and educational attainment. Psychology and Aging, 13, 179-185.

Hickman, G. P., Bartholomew, M., & Mathwig, J. (2008). Differential developmental pathways of high school dropouts and graduates. The Journal of Educational Research, 102, 3-14.

Hofferth, S. L., & Anderson, K. G. (2003). Are all dads equal? Biology versus marriage as a basis for paternal investment. Journal of Marriage and Family, 65, 213-232. doi: 10.1111/j.1741-3737.2003.00213.x

Hong, G., & Yu, B. (2008). Effects of kindergarten retention in children’s development: An application of propensity score method to multivariate, multilevel data. Developmental Psychology, 44, 407-421. doi: 10.1037/0012-1649.44.2.407

Hox, J. J. (2010). Multilevel analysis: Techniques and applications (2nd ed.). New York, NY:

Routledge.

Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.

Janosz, M., LeBlanc, M., Boulerice, B., & Tremblay, R. E. (1997). Disentangling the weight of school dropout predictors: A test on two longitudinal samples. Journal of Youth and Adolescence, 26, 733 – 762.

Jeynes, W. H. (2006). The impact of parental remarriage on children. Marriage & Family Review, 40(4), 75-102. doi: 10.1300/J002v40n04_05

Jimmerson, S., Egeland, B., Sroufe, L. A., & Carlson, B. (2000). A prospective longitudinal study of high school dropouts examining multiple predictors across development. Journal of School Psychology, 38, 525-549.

Kantarevic, J., & Mechoulan, S. (2006). Birth order, educational attainment, and earnings: An investigation using the PSID. Journal of Human Resources, 41, 755-777.

Kline, R. B. (2011). Principles and practice of structural equation modeling (3rd ed.). New York, NY: Guilford Press.

LaRossa, R. & Reitzes, D. C. (1993). Symbolic interactionism and family studies. In P.G. Boss, W.J. Doherty, R. Larossa. W.R. Schumm, & S.K. Steinmetz (eds.), Sourcebook of family theories and methods: A contextual approach (pp.135-177). New York, NY: Plenum Press.

Lesley, A., & Dianne, L. E. (2001). Rurality and capital: Educational expectations and attainments of rural, urban/rural, and metropolitan youth. Canadian Journal of Higher Education, 31(2), 1–45.

Lupien, S. J., McEwen, B. S., Gunnar, M. R., & Heim, C. (2009). Effects of stress throughout the lifespan on the brain, behavior, and cognition. Nature Reviews Neuroscience, 10, 434- 445. doi:10.1038/nrn2639

Maccoby, E. E., & Mnookin, R. H. (1992). Dividing the child: Social and legal dilemmas of custody. Cambridge, MA: Harvard University Press.

Manning, W. D., & Lamb, K. A. (2003). Adolescent well-being in cohabiting, married, and single-parent families. Journal of Marriage and Family, 65, 876-893. doi: 10.1111/j.1741-3737.2003.00876.x

Mare, R. D. (1980). Social background and school continuation decisions. Journal of the American Statistical Association, 75, 295-305.

Melby, J. N., Fang, S-A, Wickrama, K. A. S., Conger, R. D., & Conger, K. J. (2008). Adolescent

family experiences and educational attainment during early adulthood. Developmental Psychology, 44, 1519-1536. doi: 10.1037/a0013352

Mezuk, B., Eaton, W. W., Golden, S. H., & Ding, Y. (2002). The influence of educational attainment on depression and risk of type 2 diabetes. American Journal of Public Health, 98, 1480-1485

Molnar, B. E., Browne, A., Cerda, M., & Buka, S. L. (2005). Violent behavior by girls reporting violent victimization. Archives of Pediatric Adolescent Medicine, 159, 731-739. doi: 10.1001/archpedi.159.8.731

Moore, W., Pedlow, S., Krishnamurty, P., & Wolter, K. (2000). National longitudinal survey of youth1997 (NLSY1997) technical sampling report. Chicago, IL: National Opinion Research Center.

Muthen, B. (1989). Latent variable modeling in heterogeneous populations. Psychometrika, 54, 557-585.

Muthen, L. K., & Muthen, B. O. (1998-2010). Mplus user’s guide (6th ed.). Los Angeles: Muthen & Muthen.

Muthen, L. K., & Muthen, B. O. (2005). Chi-square difference testing using the S-B scaled chi- square. Retrieved from http://www.statmodel.com/chidiff.shtml

Orthner, D. K., Jones-Sanpei, H., Hair, E. C., Moore, K. A., Day, R. D., & Kaye, K. (2009). Marital and parental relationship quality and educational outcomes for youth. Marriage & Family Review, 45, 249-269. doi: 10.1080/01494920902733617

Papernow, P. L. (1993). Becoming a stepfamily: Patterns of development in remarried families. San Francisco: Jossey-Bass.

Perna, L. W., & Titus, M. A. (2005). The relationship between parental involvement as social capital and college enrollment: An examination of racial/ethnic group differences. The Journal of Higher Education, 76, 486-518.

Peugh, J. L. (2010). A practical guide to multilevel modeling. Journal of School Psychology, 48, 85-112.

Pong, S., & Ju, D. (2000). The effects of change in family structure and income on dropping out of middle and high school. Journal of Family Issues, 21, 147-169. doi: 10.1177/019251300021002001

Plunkett, S. W., Behnke, A. O., Sands, T., & Choi, B. Y. (2009). Adolescents’ reports of parental engagement and academic achievement in immigrant families. Journal of Youth and Adolescence, 38, 257-268.

Rumberger, R. W., & Thomas, S. L. (2000). The distribution of dropout and turnover rates among urban and suburban high school. Sociology of Education, 73, 39-67. doi:10.2307%2F2673198

Teachman, J. D., Paasch, K., & Carver, K. (1996). Social capital and dropping out of school early. Journal of Marriage and Family, 58, 773-783. doi: 10.2307/353735

Rabe-Hasketh, S., & Skrondal, A. (2008). Multilevel and longitudinal modeling using Stata (2nd ed.). College Station, TX: Stata Press.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models (2nd ed.). Thousand Oaks, CA: Sage.

Sandefur, G. D., Meier, A. M., & Campbell, M. E. (2006). Family resources, social capital, and college attendance. Social Science Research, 35, 525-553.

Satorra, A. & Bentler, P. M. (2001). A scaled difference chi-square test statistic for moment structure analysis. Psychometrika, 66, 507-514.

Sitlington, P. L., Frank, A. R., & Carson. R. (1992). Adult adjustment among high school graduates with mild disabilities. Exceptional Children, 59, 221-233.

Smith, C. A. (1997). Factors associated with early sexual activity among urban adolescents. Social Work, 42, 334-346.

Stattin, H., & Kerr, M. (2000). Parental monitoring: A reinterpretation. Child Development, 71, 1072-1085.

Stearns, E., & Glennie, E. J. (2006). When and why dropouts leave high school. Youth Society, 38, 29-57.

Stewart, E. B. (2007). School structural characteristics, student effort, peer associations, and parental involvement: The influence of school and individual-level factors on academic achievement. Education and Urban Society, 40, 179-204. doi: 10.1177/0013124507304167

Stewart, S. D. (2002). The effect of stepchildren on childbearing intentions and births. Demography, 39, 181-197.

Stright, A. D., & Bales, S. T. (2003). Coparenting quality: Contributions of child and parent characteristics. Family Relations, 52, 232-240.

Stryker, S. (1968). Identity salience and role performance: The relevance of symbolic interaction theory for family research. Journal of Marriage and the Family, 30, 558-564.

Stryker, S. (1987). Identity theory: Developments and extensions. In K. Yardley & T. Honess

(Eds.), Self and identity: Psychosocial perspectives (pp. 89-103). New York, NY: Wiley.

Sun, Y., & Li, Y. (2009). Postdivorce family stability and changes in adolescents’ academic performance: A growth-curve model. Journal of Family Issues, 30, 1527-1555.

Swanson, C. B., & Schneider, B. (1999). Students on the move: Residential and educational mobility in America’s schools. Sociology of Education, 72, 54-67.

Tillman, K. H. (2007). Family structure pathways and academic disadvantage among adolescents in stepfamilies. Sociological Inquiry, 77, 383-424. doi: 10.1111/j.1475- 682X.2007.00198.x

Tillman, K. H. (2008). “Non-traditional” siblings and the academic outcomes of adolescents. Social Science Research, 37, 88-108. doi:10.1016/j.ssresearch.2007.06.007

Thompson, B. (2004). Exploratory and confirmatory factor analysis. Washington, DC: American Psychological Association.

Umberson, D., & Chen, M. D. (1994). Effects of a parent’s death on adult children: Relationship salience and reaction to loss. American Sociological Review, 59, 152-168.

U.S. Census Bureau. (2010). Current population survey: Detailed years of school completed by people 25 years and over by sex, age groups, race, and Hispanic origin: 2010. Retrieved January 20, 2012 from: http://www.census.gov/hhes/socdemo/education/data/cps/2010/tables.html

U.S. Census Bureau. (2011). Current population survey: Educational attainment of the population 18 years and over, by age, sex, race and Hispanic origin: 2011. Retrieved April 11, 2012 from: http://www.census.gov/hhes/socdemo/education/data/cps/2011/tables.html

Wayman, J. C. (2002). The utility of educational resilience for studying degree attainment in school dropouts. Journal of Educational Research, 95, 167-178.

Wojtkiewics, R. A., & Holtzman, M. (2011). Family structure and college graduation: Is the stepparent effect more negative than the single parent effect? Sociological Spectrum: Mid-South Sociological Association, 31, 498-521. doi: 10.1080/02732173.2011.574048

Woolley, M. E., & Grogan-Kaylor, A. (2006). Protective family factors in the context of neighborhood: Promoting positive school outcomes. Family Relations, 55, 95-104.

Zellman, G. L., & Waterman, J. M. (1998). Understanding the impact of parent school involvement on children’s educational outcomes. Journal of Educational Research, 91, 370-380.

BIOGRAPHICAL SKETCH

Chelsea Garneau received her B. A. degree in Psychology from Michigan State University, her M. A. degree in Marriage and Family Therapy from the Adler School of Professional Psychology, and finally her Ph.D. in Family Relations from The Florida State University. Chelsea’s research interests include studying the influence of complicated family structure on a variety of outcomes for adolescents and young adults, such as romantic relationship quality, sexual behaviors, and educational outcomes. She also is interested in examining relationship dynamics of couples in repartnered and remarried families, particularly the influence of beliefs on co-parenting, financial management, and relationship satisfaction.