WENDY D. MANNING AND KATHLEEN A. LAMB Bowling Green State University

ࡗ

Adolescent Well-Being in Cohabiting, Married, and Single-Parent Families

Cohabitation is a family form that increasingly status was perhaps more appropriate when rela- includes children. We use the National Longitu- tively few children lived in cohabiting unions. Re- dinal Study of Adolescent Health to assess the cent estimates indicate that two fifths of children well-being of adolescents in cohabiting parent are expected to spend some time in a cohabiting stepfamilies (N ϭ 13,231). Teens living with co- parent family (Bumpass & Lu, 2000), and 41% of habiting stepparents often fare worse than teens cohabiting unions have children present (Fields & living with two biological married parents. Ado- Casper, 2000). Despite this shift in children’s ex- lescents living in cohabiting stepfamilies experi- perience in cohabitation, research on the impli- ence greater disadvantage than teens living in cations of cohabitation for children’s lives is rel- married stepfamilies. Most of these differences, atively sparse. however, are explained by socioeconomic circum- In this paper we examine the well-being of ad- stances. Teenagers living with single unmarried olescents in cohabiting stepparent families. We use mothers are similar to teens living with cohabiting the term cohabiting stepfamily to indicate living stepparents; exceptions include greater delin- with one biological parent and the parent’s partner quency and lower grade point averages experi- (cohabiting stepfamily). We address three key ques- enced by teens living with cohabiting stepparents. tions in this paper. First, do teenagers in cohabiting Yet mother’s marital history explains these differ- stepparent families have similar academic and be- ences. Our results contribute to our understanding havioral outcomes as teenagers living with two of cohabitation and debates about the importance married biological parents? We begin with this of marriage for children. question because over half of the children in the United States live with two married biological par- An extensive literature exists that examines the ents (Fields, 2001), and most research on family importance of family structure (defined by marital structure contrasts how children in specific family status) for child well-being. Marital status acts as types fare compared with children living with mar- an indicator of the potential number of caretakers ried, two-biological-parent families. Second, do and may imply certain characteristics or qualities children residing with cohabiting stepparents fare of the child’s family life. This emphasis on marital better or worse than children living with single mothers? We focus on children living with unmar- Department of Sociology and Center for Family and De- ried mothers and determine how their cohabitation mographic Research, Bowling Green State University, status influences child well-being. Third, do ado- Bowling Green, OH 43403 ([email protected]). lescents in cohabiting stepfather families fare as Key Words: adolescence, child well-being, cohabitation, well as adolescents living in married stepfather family structure, marriage, stepfamilies. families? We test whether children living with step-

876 Journal of Marriage and Family 65 (November 2003): 876–893 Adolescent Well-Being 877 fathers fare better when their mother is married, Thus, cohabitation for adolescents (unlike for rather than cohabiting. For each question, we eval- young children) represents a family that is struc- uate whether the effects of parental cohabitation are turally similar to a stepfamily. explained by socioeconomic circumstances, parent- ing, and family instability. Cohabitation and Family Life This paper builds on prior research and moves beyond previous work in several key ways. First, Children in cohabiting parent families experience by employing a large data source (National Lon- family life that differs from those raised with mar- gitudinal Study of Adolescent Health), our anal- ried or single parents. Children raised in cohabit- yses are based on a relatively large number of ing couple families may experience different de- adolescents in cohabiting stepfather families. Sec- velopmental outcomes, in part because of the ond, the rich nature of the data allows us to in- family environment or context in which children clude potentially important factors that represent are raised. We discuss three potential contextual family processes and may help account for some mechanisms through which family structure, and observed effects of family structure. Third, we are specifically cohabiting parent families, may influ- not limited to a single indicator of well-being and ence child well-being: economic circumstances, focus on multiple measures of well-being that are instability, and parenting. appropriate for teenagers. Finally, to better under- stand the implications of cohabitation on child Economic status. Children raised in families with well-being, we focus on family-type comparisons higher socioeconomic status experience more pos- based on similar household structure (stepfather itive cognitive and social developmental indica- presence; cohabiting stepfather vs. married step- tors of well-being (e.g., Carlson & Corcoran, father) or mother’s marital status (unmarried 2001; Duncan & Brooks-Gunn, 1997; McLanahan mothers; cohabiting mother vs. single mother). & Sandefur, 1994). Indicators of both family in- come and mother’s education exert positive effects on child development, but income rather than BACKGROUND mother’s education seems to have a stronger influ- ence on child outcomes (Duncan & Brooks- Cohabitation As a Family Structure Gunn). It appears that income typically does not Children in the United States are increasingly like- explain the effects of family structure on child ly to spend some of their lives residing in a co- well-being, but for some outcomes, it does reduce habiting parent family. Indeed, two fifths of co- the effect of family structure (Carlson & Corco- habiting households include children (Fields & ran; Duncan & Brooks-Gunn; Hill, Yeung, & Casper, 2000). In 1999, 6% of children were liv- Duncan, 2001; McLanahan & Sandefur). On av- ing with a cohabiting parent (Acs & Nelson, erage, children raised in cohabiting parent families 2001). Bumpass and Lu (2000) estimate that two experience economic situations that are better than fifths of children in the United States are expected those of children in single-parent families (e.g., to experience a cohabiting parent family at some greater parental education and family earnings), point during their childhood, and children born but more stressful economic situations than chil- during the early 1990s will spend 9% of their lives dren in married couple families (e.g., greater pov- living with parents who are in cohabiting unions. erty and food insecurity; Acs & Nelson, 2002; Adolescents in cohabiting parent families typ- Manning & Lichter, 1996). ically are living with their mother and her cohab- iting partner. Based on the 1996 Survey of Income Family stability. Family stability is positively re- and Program Participation, half (54%) of the chil- lated to child and young adult behavior (Hao & dren in cohabiting parent families lived with one Xie, 2001; Hill et al., 2001; Wu & Martinson, biological parent (Fields, 2001). Given the insta- 1993). At times family stability has a stronger in- bility of cohabiting unions for children, older chil- fluence on child outcomes than family structure. dren in cohabiting parent families primarily live It is argued that the stress of family change hin- with their mother and her partner who is not their ders normal developmental transitions among biological parent (Manning, Smock, & Majumdar, children (Hao & Xie; Hill et al.; Wu & Martin- in press). Brown (2002) reports that almost all son). Family stability may be particularly impor- children over the age of 12 in cohabiting parent tant in assessments of the effect of cohabitation families are living with only one biological parent. because children born to cohabiting parents ex- 878 Journal of Marriage and Family perience higher levels of instability than children habiting parent families with children living with born to married parents (Manning et al., in press). two biological married parents. The focus of most of these studies is not specifically on cohabitation Parenting. Parental monitoring is important for but more broadly on how family structure influ- keeping children’s behavior on task and ensuring ences child well-being. The results of these studies that children meet their individual responsibilities. indicate that children in cohabiting parent families Empirical evidence supports the notion that pa- fare worse than their counterparts in married, two- rental monitoring has positive effects on children. biological-parent families. For example, McLanahan (1997) reports lack of A limitation of this approach is that it con- supervision by parents is associated with poor founds the effects of marriage and living with two school performance among children in single and biological parents. Research on family structure stepparent families. Another core feature of par- recognizes the importance of adults’ biological enting is parental support, which is positively re- ties to children and argues that children in two- lated to desirable outcomes for children and ado- biological-parent families fare better than children lescents (e.g., Baumrind, 1991). For instance, living with a stepparent (see Coleman, Ganong, & interacting with children in positive ways has been Fine, 2000). Following this logic, the biological shown to raise grade point averages and decrease relationship of cohabiting partners should be con- externalizing behaviors (e.g., O’Connor, Hether- sidered in the analysis of child well-being. Many ington, & Clingempeel, 1997). Parent-child rela- of the children who are living in cohabiting parent tionships that cross household boundaries also families, particularly older children, are not living influence children’s development. Evidence sug- with their biological father, making the traditional gests that closeness to nonresident fathers is pos- married stepparent family a more appropriate itively associated with child well-being (Amato & comparison group. To better understand the influ- Gilbreth, 1999; White & Gilbreth, 2001). ence of cohabitation, we argue that comparisons Parenting in cohabiting unions may have be- should be made across families who share either come easier as cohabitation moves toward social the same biological relationships to parents (two acceptance, but cohabiting unions with children biological parents or stepfamilies) or parental present still do not benefit from legal and social marital status (married or unmarried), and differ recognition (e.g., Durst, 1997; Mahoney, 2002). in terms of the presence or absence of a cohabiting Thus the responsibilities of cohabiting partners to partner (Manning, 2002). children are not specified, creating sources of par- The findings from empirical work suggest that enting ambiguity in terms of obligations and rights teenagers and children in cohabiting parent step- of cohabiting partners to their partner’s children. families sometimes fare worse in terms of behav- Research that distinguishes parenting behaviors of ior problems and academic performance than chil- cohabitors from married couples or single parents dren in married stepparent families (Brown, 2001; supports the notion that slightly more negative Buchanan, Maccoby, & Dornbusch, 1996; parenting practices occur among cohabiting par- Morrison, 2000; White & Gilbreth, 2001). Other ents (Brown, 2002; Dunifon & Kowaleski-Jones, research suggests that adolescents and children in 2000; Hofferth & Anderson, 2003; Thomson, cohabiting stepparent families share similar levels McLanahan, & Curtin, 1992). Yet parenting in- of behavior problems and academic achievement dicators do not explain the effect of parental co- as children in married stepparent families (Brown; habitation on child well-being (Dunifon & Ko- Morrison, 1998, 2000). The findings seem to de- waleski-Jones; Thomson, Hanson, & McLanahan, pend on the gender and age of the child as well 1994; White & Gilbreth, 2001). as the specific dependent or outcome variable (e.g., math scores vs. verbal scores or internalizing vs. externalizing behavior). Cohabitation and Child Outcomes Only a few studies contrast the well-being of To date, a limited but growing number of studies children in unmarried mother families who have examine the social well-being of children living a cohabiting parent with those who do not. Anal- in cohabiting parent families (e.g., Brown, 2001; ysis of the 1999 National Survey of American DeLeire & Kalil, 2002; Dunifon & Kowaleski- Families (NSAF) suggests teenagers living in sin- Jones, 2002; Hao & Xie, 2001; Nelson, Clark, & gle-mother and cohabiting stepparent families Acs, 2001; Thomson et al., 1994). Often these re- share similar levels of behavior problems (Acs & searchers contrast the well-being of children in co- Nelson, 2002). Work using longitudinal data and Adolescent Well-Being 879 multivariate, fixed effects models finds that teen- Jones). Also, nonresident biological fathers are of- agers living with cohabiting mothers and unmar- ten ignored. Rarely have relationships with non- ried mothers share similar levels of behavior prob- resident fathers been considered in assessments of lems (Morrison, 1998). how children living with cohabiting parents fare, Two shortcomings of prior work are limited despite the fact that this relationship may be ad- samples and a narrow range of covariates. First, a vantageous to the child’s well-being (White & few studies are restricted only to children of di- Gilbreth, 2001). vorce (Buchanan et al., 1996; Morrison, 1998, 2000). The implications of cohabitation may differ CURRENT INVESTIGATION among children who have lived with married bi- ological parents compared with children who have Three broad questions are addressed in this paper. never lived with their biological father. In addi- First, the literature shows that children are gen- tion, other data sources (such as the National Sur- erally better off when they live with two biolog- vey of Families and Households [NSFH]) have ical, married parents (e.g., Brown, 2002; Mc- small numbers of children in cohabiting, two-bi- Lanahan & Sandefur, 1994). In addition, in 1996 ological-parent and cohabiting stepparent families, over 50% of the children in the United States were and sample sizes become even smaller when two living in married, two-biological-parent families waves of data are used (e.g., Hao & Xie, 2001; (Fields, 2001). Therefore, a basic starting point is White & Gilbreth, 2001). Finally, data sources to demonstrate whether teenagers living with co- such as the National Longitudinal Survey of habiting stepparent families fare the same or Youth (NLSY) include less than optimal measures worse than children living with two married, bi- of parental cohabitation. Parental cohabitation is ological parents. measured annually, so research using these data is Given the vast literature that supports the rel- biased toward longer term cohabiting unions ative strength of the married, two-biological-par- (more than 1 year; Dunifon & Kowaleski-Jones, ent family, of greater interest in this analysis will 2002; Morrison, 2000). Thus, analyses using the be other family structure comparisons. Our second NLSY may be underestimating the negative ef- question is whether cohabitation provides any ad- fects of cohabitation because only longer term vantage for children living with unmarried moth- unions are included in the data. ers. Based on both social control and economic A second shortcoming is that some research deprivation perspectives, children in single-parent includes only a narrow set of independent vari- families may fare worse than children in cohabi- ables. Thus, prior studies cannot explore potential tation because they lack the benefits of income explanations about why children in cohabiting and parenting that a cohabiting partner may pro- parent families fare differently than children in vide. As a result, we anticipate that children in other family types, disentangling the effects of cohabiting-parent families will fare better than family structure from other factors. First, a few children in single-mother families. A competing studies include only socioeconomic indicators, hypothesis is that children experience some dis- such as gender, parental education, and poverty advantages by living with a mother’s unmarried (Hanson, McLanahan, & Thomson, 1997; Nelson partner who may not be a fully integrated family et al., 2001). Second, other research does not in- member and may compete for their mother’s time clude measures of family instability or indicators and attention. Family roles may not be as clearly of relationship quality (Acs & Nelson, 2002; established in cohabiting stepfamilies, perhaps Thomson et al., 1994). The NSAF does not in- creating confusion over parenting responsibilities clude questions about duration of the parents’ re- and weak child-stepparent relationships. This hy- lationship or relational history (Acs & Nelson; pothesis is consistent with the role ambiguity per- Brown, 2001; Nelson et al., 2001). Other studies spective used to understand stepfamilies. In this that include measures of family stability do not case, adolescents in cohabiting stepfamilies would incorporate measures of the resident parents’ re- fare worse than adolescents in single-mother fam- lationship quality (DeLeire & Kalil, 2002; Duni- ilies. Finally, we may find no effect of cohabita- fon & Kowaleski-Jones, 2002; Hao & Xie, 2001). tion as the benefits and costs of a cohabiting par- Third, many studies do not include measures of ent outweigh one another. The bulk of research on parenting strategies when evaluating the effects of stepfamilies indicates that children in stepfamilies parental cohabitation on well-being (exceptions and single-mother families share similar devel- include Brown, 2001, and Dunifon & Kowaleski- opmental outcomes (Coleman et al., 2000). Thus 880 Journal of Marriage and Family we may find that adolescents who live in cohab- Tests, grades in school, and college expectations. iting stepfamilies fare as well as children who re- As any one measure may suffer some shortcom- side with a single mother. ings, taken together we have indicators of well- Third, do children experience any advantage being that tap several dimensions of adolescent by living in a married (or traditional) rather than behavior and academic well-being. in a cohabiting stepparent family? We determine Fourth, we are able to include key variables whether children in married stepparent families that may explain some of the effects of family fare as well as children in cohabiting stepparent structure on child outcomes. We include measures families. Marriage provides the socioeconomic of parenting characteristics (closeness to mother benefits and stability that cohabitation does not and nonresident father, as well as monitoring); so- offer. Moreover, family roles may be clearly de- cioeconomic status (mother’s education and fam- fined and child-stepparent relationships more for- ily income); and family stability (number of moth- malized in married than in cohabiting stepparent er’s marriages and duration of relationship). Most families. We expect children in married stepfam- prior work has accounted for one or more of these ilies to have better developmental outcomes than measures, but no study has accounted for all of children in cohabiting stepfamilies. Once we ac- these factors simultaneously. count for the parent’s relationship with the child, In addition to our measures of socioeconomic family stability, and socioeconomic characteris- status, family stability, and parenting, we control tics, however, these differences according to mar- for a number of sociodemographic and child char- ital status may no longer exist. These findings may acteristics, including race and ethnicity, mother’s suggest that marriage itself does not create the ad- age, child’s age and sex, number of children in the vantage experienced by children in married step- household, and importance of religion to the child. parent families. If differences persist, then such Although residing in a cohabiting or single-parent findings would indicate that some feature of co- family is increasingly common for all children, it habitation itself (i.e., role ambiguity) may have is a more common feature of the life experiences negative consequences for children in this type of of Black and Hispanic children (Bumpass & Lu, family structure. 2000). We also control for mother’s age; older Previous work provides some initial evidence mothers may be more skilled at parenting, which about the effects of cohabitation on child well- in turn may result in increased attentiveness to chil- being. In this project we build on previous studies dren’s needs. The number of one’s siblings is re- in four key ways. First, many of the previous stud- lated negatively to academic achievement (e.g., ies do not distinguish between adolescents and Carlson & Corcoran, 2001), presumably because younger children. Our focus on adolescents limits more children in the household means parents pos- our conclusions to one stage of childhood, but at sess fewer instrumental and emotional resources to the same time allows us to detail the effects of invest in each child individually. In terms of the family structure for a critical period of develop- characteristics of the adolescent, boys tend to ex- ment. We examine outcomes that are most salient perience more behavior problems than girls, and for adolescents. girls tend to have higher academic achievement Second, most adolescents in cohabiting parent than boys (Carlson & Corcoran). We control for families are living with only one biological parent child’s age, as older children may experience fewer (Brown, 2001). Thus, answers to questions about behavior problems as a function of maturity. We the effects of cohabitation require being specific also control for the importance of religion to the about the family type contrasts. The traditional ap- adolescent, as involvement with an institution that proach is to compare the well-being of all children encourages adherence to particular moral standards in cohabiting families with those in married, two- may act as an agent of social control to discourage biological-parent families. Yet, contrasting the deviant behavior in young people. Families who well-being of adolescents in married and cohab- encourage religious attendance may also more iting stepfamilies is more appropriate because closely monitor the actions of their children. these families share the same basic structure (bi- ological mother and her cohabiting partner). METHOD Third, we include a range of indicators of well- being. For example, we do not rely on a single Data measure to indicate academic achievement. We in- We draw on the first wave of the National Lon- clude measures of Peabody Picture Vocabulary gitudinal Adolescent Study of Adolescent Health Adolescent Well-Being 881

(Add Health). The Add Health is based on inter- adolescents living with unmarried mothers (sin- views with students in grades 7 through 12 and gle-mother vs. cohabiting-mother families) so we their parents in 1995. These data are based on a can estimate the effect of cohabitation among un- sample of 80 high schools and 52 middle schools married mothers. Second, we focus on teenagers from the United States. We use the contractual living with stepfathers (married stepfather families data that include in-home interviews administered vs. cohabiting stepfather families) so we can de- to 18,924 students with a response rate of 78.2% termine the influence of formal marital status (Udry, 1998). These sample schools were selected among children living with stepfathers. Our anal- with unequal probability of selection. Once design ysis of teens living with single mothers and step- effects are taken into account, these data are na- fathers is based on 5,504 respondents. tionally representative of adolescents in the Unit- ed States (see Bearman, Jones, & Udry, 1997). We Dependent Variables use procedures in a software package, STATA, to ensure our results are nationally representative We include a range of indicators of well-being. with unbiased estimates (Chantala & Tabor, 1999). The indicators of behavior problems are ever hav- In this paper we use the first wave of the Add ing been expelled or suspended from school, ex- Health data. This cross-sectional analysis provides periencing trouble in school, and self-reported de- a basic starting point for understanding whether linquency scores. The suspension or expulsion parental cohabitation is associated with indicators measure is a dichotomous measure simply indi- of child well-being. Researchers often emphasize cating whether the respondent ever received an how changes in family structure influence child out of school suspension from school or an ex- outcomes without understanding whether and how pulsion from school. This is coded such that 1 ϭ specific family structures are associated with child yes and 0 ϭ no. Unlike the other outcomes, ex- outcomes. Furthermore, fixed effects models do pulsion or suspension may occur prior to the for- not allow for the analysis of how core, fixed, so- mation of the current family, but provides a rough ciodemographic variables such as race or gender indicator of problem behavior. The second mea- influence adolescent outcomes. sure, problems in school, assesses the respondent’s The Add Health is appropriate because it con- difficulty in the school context. The four items tains a large number of adolescents living in co- comprising the scale indicate the degree, since the habiting parent families, includes key measures of start of the school year, the respondent has had consequential adolescent outcomes, and has rich problems getting along with teachers, paying at- measures of family processes that may explain tention in school, getting homework done, and some of the observed differences in family struc- getting along with other students. (All items are ture. Other data sources, such as the National Sur- coded such that 0 ϭ never, 1 ϭ just a few times, vey of American Families and Current Population 2 ϭ about once a week, 3 ϭ almost every day, Survey, provide information only about the cur- and 4 ϭ every day.) The responses are summed so rent family situation and no details about family the scores may range from 0 to 16. This measure stability. Yet the Add Health data do not include has a Cronbach ␣ reliability of .69. The delin- details about family structure histories. quency scale is composed of 15 items asking the Our analytic sample depends on the question frequency that respondents engaged in a series of that we address. Dividing the sample is necessary delinquent acts over the past 12 months, including because not all of the predictors used for analyses painting graffiti or signs on someone else’s prop- of married, two-biological-parent families can be erty or in a public place; deliberately damaging applied to the unmarried and stepparent families property; lying to parents or guardian about whom (e.g., number of mother’s prior marriages and respondent had been with; taking something from nonresident father closeness). We begin by con- a store without paying for it; getting into a serious trasting the well-being of children in cohabiting physical fight; hurting someone badly enough to stepparent families to those living in married, two- need medical care; running away from home; biological-parent families, including all possible driving a car without the owner’s permission; family types. Our analytic sample consists of stealing something worth more than $50; going 13,231 adolescents. Our next analysis is limited into a house or building to steal something; using to teens living in stepfamilies and single-mother or threatening to use a weapon to get something families. We make two sets of specific family from someone; selling marijuana or other drugs; comparisons. First, we examine the well-being of stealing something worth less than $50; taking 882 Journal of Marriage and Family part in a fight where a group of friends was live in two-biological-parent cohabiting families. against another group; or being loud, unruly, or This is consistent with findings from other data rowdy in a public place. Responses (scored such (Brown, 2002). Thus, we limit our analyses of co- that 0 ϭ never, 1 ϭ one or two times, 3 ϭ three habitation to adolescents living with their biolog- or four times, 3 ϭ five or more times) were ical mother and her cohabiting partner (cohabiting summed such that the scores ranged from 0 to 45. stepfather families). Our family structure catego- After the items were summed, cases were omitted ries include two married biological parents, single from analysis when less than 75% (11 items) of mother, married stepfather, and cohabiting step- the items had valid responses. Cases where 75% father. Table 1 shows the distribution of the in- or more of the items had valid data were given dependent variables according to each family the mean of the scale on any items with missing type. Among adolescents living in stepfamilies, data. This strategy allows us to retain respondents one third live with cohabiting parents and two in our sample and base delinquency scores on a thirds live with married parents. Among adoles- minimum of 11 items. The delinquency measure cents living with unmarried mothers, 13% are liv- has a high Cronbach ␣ reliability of .85. ing with their mother and her cohabiting partner. Measures of cognitive development or academ- The unmarried mothers may be never married, di- ic achievement and expectations are student-re- vorced, or widowed. These findings mirror those ported grade point average, Peabody Picture Vo- reported in the NSAF and NSFH (Brown, 2002; cabulary Test, and college expectations. Only one Bumpass, 1994). measure may not be an adequate indicator of ac- The remaining independent variables are divid- ademic achievement. Low grade point average is ed into three categories: sociodemographic, par- a dichotomous measure indicating whether, of enting or socialization variables, and family sta- four subject areas in school (English, mathemat- bility. The distribution for each of the independent ics, history or social studies, and science), the re- variables is provided in Table 1. spondent received two or more grades of D or lower. Respondents receiving two or more Ds or Sociodemographic. Race and ethnicity respon- Fs were coded as 1, and respondents receiving dents is based on their own response and coded one or no Ds or Fs were coded as 0. We use this into four categories: Black, White, Latino, and measure of poor academic performance because Other. The ‘‘other’’ category includes groups that grading systems vary considerably across schools, are too small to distinguish in analyses. In both and student grades depend on the types of classes stepparent and unmarried mother families, the ma- students attend (e.g., advanced placement courses jority of the adolescents are White, whereas 15% vs. a general curriculum). The second indicator is are Black and 12% Latino. The family income an abbreviated version of the Peabody Picture Vo- measure is logged and the family income values cabulary Test. We use the age-standardized scores are higher among teens in married stepparent fam- with a mean of 100 and a standard deviation of ilies than in the other family types. A shortcoming 15. This is considered a measure of verbal cog- of the Add Health data is that a considerable share nitive ability or development. The third indicator (23%) of the sample has missing data on income. measures expectations for college. Respondents To avoid deleting all of these cases, respondents were asked how much they want to go to college with missing income are coded to the mean value (responses ranging from 1 ϭ low to 5 ϭ high). of income and a dummy variable is included in The mean response on this question was high with the model that indicates which respondents were a value of 4. missing on income. Mother’s age is coded as a continuous variable, and the mean value is 32. Mother’s education is coded on an ordinal scale Independent Variables (1 ϭ eighth grade or less;2ϭ more than eighth Family structure. The key independent variable is grade, but did not graduate from high school;3 family structure. Cohabitation family status is es- ϭ went to a business, trade, or vocational school tablished by the adolescent response in the house- in place of high school;4ϭ received a GED;5 hold roster question and by the parent’s response ϭ high school graduate;6ϭ went to college but to relationship questions. If either the adolescent did not graduate;7ϭ graduated from a college or the parent reports that the parent has a cohab- or university;8ϭ had professional training be- iting partner, then the family is coded as a cohab- yond college). On average, single mothers have a iting parent family. We find very few adolescents high school education, and mothers in married Adolescent Well-Being 883

TABLE 1. DISTRIBUTION OF INDEPENDENT VARIABLES, BY FAMILY STRUCTURE (N ϭ 13,231) Married Two Biological Single Married Cohabiting Parents Mother Stepfather Stepfather Sociodemographic Race White .75 .49 .73 .56 (.02) (.04) (.02) (.04) Black .07 .33 .11 .19 (.01) (.04) (.02) (.03) Hispanic .11 .13 .11 .19 (.02) (.02) (.02) (.03) Other .07 .05 .06 .07 (.01) (.01) (.01) (.01) Log family income 3.75 3.01 3.63 3.19 (.03) (.04) (.03) (.05) Missing income (1 ϭ yes) .12 .21 .08 .15 (.01) (.01) (.01) (.03) Mother’s age 41.2 39.15 38.19 37.53 (.17) (.22) (.23) (.28) Mother’s education 5.49 5.04 5.43 4.89 (.09) (.10) (.09) (.13) Child’s age 15.28 15.35 15.33 15.20 (.12) (.14) (.13) (.17) Child’s sex (1 ϭ male) .52 .47 .51 .54 (.01) (.01) (.02) (.03) Importance of religion to child 3.34 3.33 3.31 3.21 (.02) (.02) (.03) (.04) Number of children in household 1.24 1.28 1.45 1.41 (.03) (.06) (.05) (.09) Family Stability Number of mother’s marriages 1.01 1.45 2.12 2.16 (.25) (.03) (.03) (.06) Duration of relationship 15.20 — 6.67 4.44 (.20) (.23) (.27) Parenting Monitoring by parents 1.93 1.70 1.97 1.82 (.06) (.07) (.08) (.10) Closeness to mother 4.56 4.58 4.63 4.49 (.02) (.02) (.02) (.05) Closeness to nonresident father — 3.06 3.13 3.11 (.03) (.05) (.07) Missing closeness to nonresident — .25 .26 .27 father (1 ϭ yes) — (.01) (.02) (.03) N 7,727 3,593 1,352 559 Note: From the National Longitudinal Survey of Adolescent Health. stepfamilies have the highest levels of education. Family stability. Indicators of family stability in- Religiosity is measured by responses to questions clude mother’s relationship history and duration about the importance of religion in the life of the of current relationship. The number of mother’s adolescent. The responses range from 1 to 4, with prior marriage-like relationships is included as a 1 indicating not at all important and 4 indicating control variable. These relationships are asked very important. The mean response is 3.3, indi- about in reference to the 18-year period prior to cating religion is considered fairly important. The Wave I, or from 1977–1995, so these refer to mean age of the child is 15 and the ages range changes in mother’s relationships during the from 11 to 21. The sample is evenly split between course of the child’s lifetime. Single mothers have boys and girls. On average, about one other child been in, on average, only one marriage-like rela- lives in the household. tionship, and cohabiting and married mothers in 884 Journal of Marriage and Family this sample have been in, on average, two rela- Our analytic strategy is to estimate a series of tionships. The following indicator of stability is models for each outcome. We first estimate a zero- applied only to the stepfamily analysis. Stability order or bivariate model that includes only the of the stepfamilies is measured in terms of the family structure variable. The second model we duration of the parental relationship. The mean present adds the remaining factors, including so- duration of the cohabiting stepfamilies is 4.4 cioeconomic, parenting, and family stability mea- years, and the mean duration of the married step- sures. We also enter variables separately to assess families is 6.7 years. This is consistent with find- how they contribute to the fit of the models, but ings from the NSFH (Hao & Xie, 2001). because of space constraints, we do not present the results in the tables. Parenting. The parenting measures focus on con- trol and support. Parental control is based on a RESULTS seven-item scale with high values indicating high control. The questions are coded dichotomously Distribution of Adolescent Outcomes (0 ϭ yes and 1 ϭ no) and then summed. Adoles- cent respondents are asked whether parents let Table 2 presents the mean and median values of them make their own decisions about the time the dependent variables according to each family they must be home on weekend nights, the people type. This provides information about the basic they hang around with, what they wear, how much levels of the well-being indicators and shows the TV they watch, which TV programs they watch, range of values for the measures of well-being. what time they go to bed on week nights, and Most teenagers, regardless of family type, were what they eat. The ␣ reliability of the scale is .64. not expelled or suspended from school. Two fifths The mean level of control is 1.83, indicating a of the adolescents in single-mother and cohabiting fairly low level of parental supervision. stepfather families were expelled or suspended, Closeness to resident mother is an individual and three tenths of teens living in married step- item, asking teens how close they feel to their father families experienced school suspension or mothers, coded 1 ϭ not at all, 2 ϭ very little, 3 expulsion. Delinquency levels range from 0 to 45, ϭ somewhat, 4 ϭ quite a bit, 5 ϭ very much. The so those reported are quite low, and the mean val- average closeness to mothers ranges between ues are highest for teens living in cohabiting step- quite a bit to very much. Unfortunately, the data father families. In terms of school problems, the do not include questions about closeness to co- values fall within a narrow range from 3.95 to habiting stepfathers. For those respondents who 4.79, suggesting that the majority of teenagers report having a nonresident biological father, the have just a few troubles in school. The measure same question is included as a predictor. The av- of academic achievement shows that the vast ma- erage value is somewhat close. We also include a jority of teens in each family type have not re- dummy variable measuring whether responses ceived Ds or Fs in two or more subjects. The Pea- were missing on closeness to nonresident father. body Picture Vocabulary Test is an indicator of This strategy allows us to retain the variable in cognitive development, and the scores range from our analyses; approximately one quarter of the 98 to 104, with adolescents in married, two-bio- sample is missing on the indicator of closeness to logical-parent families scoring best. Finally, most nonresident father. teens possess high expectations for attending col- lege, and there appears to be only slight variation according to family type. Design We correct for design effects and the unequal Cohabiting Stepparent and Married, probability of selection using STATA (Chantala & Two-Biological-Parent Families Tabor, 1999). The analytic method depends on the nature of the dependent variables. Logistic regres- Our first aim is to contrast the well-being of chil- sion is used for analyses of dichotomous depen- dren in cohabiting stepfamilies to children living dent variables, whether the adolescent was ex- in married, two-biological-parent families (refer- pelled or suspended from school and whether the ence category in Table 3). The inclusion of the teen received low grades. Ordinary least square entire sample for these analyses prevents us from regressions are estimated for all remaining out- using the couple-level indicators (duration, rela- comes. tionship quality); number of mother’s prior mar- Adolescent Well-Being 885

TABLE 2. MEANS (STANDARD ERRORS) OF OUTCOME VARIABLES (N ϭ 13,231) Married Two Biological Unmarried Step Step Dependent Variables Parents Single Mother Married Cohabiting Suspension/expulsion M .18 (.01) .39 (.02) .30 (.02) .41 (.30) Median 0 0 0 0 Delinquency M 3.76 (.10) 4.67 (.15) 4.29 (.18) 5.44 (.33) Median 3 3 3 3 School problems M 3.95 (.06) 4.52 (.09) 4.60 (.11) 4.79 (.19) Median 3 4 4 4 Low grade point average M .09 (.01) .15 (.01) .14 (.01) .19 (.02) Median 0 0 0 0 PPVT M 103.87 (.56) 98 (.78) 102 (.62) 98 (1.02) Median 104 97 101 98 College expectations M 4.50 (.03) 4.37 (.03) 4.42 (.04) 4.28 (.07) Median 5 5 5 5 Note: From the National Longitudinal Survey of Adolescent Health. Means are weighted using Wave I grand sample weight. PPVT ϭ Peabody Picture Vocabulary Test. riages; and relationship with nonresident fathers of child well-being. Girls appear to fare better in the models. We highlight the findings related to than boys. Younger children more often have the well-being of teenagers living in cohabiting higher levels of delinquency, school problems, stepparent families. Notably, adolescents living in low GPA, and lack college expectations. Religi- married, two-biological-parent families generally osity often is associated with higher levels of child fare better than teenagers living in any other fam- well-being. Teenagers who are closer to their ily type. mothers have fewer behavioral and academic The first three columns show that teens who problems. reside in cohabiting stepfather families experience 122% (exponential value of 0.80) higher odds of Cohabiting Stepparent, Married Stepparent, and being expelled from school, greater levels of de- Single-Mother Families linquency, and more school problems than teen- agers residing with two married, biological par- The first row of Table 4 shows the effect of living ents. The next three columns indicate that with married rather than cohabiting stepparents on adolescents living with cohabiting stepfathers are adolescent problem behaviors. These sets of find- more likely to have a low grade point average or ings reflect the importance of formal marital sta- experience 90% (exponential value of 0.64) great- tus. The second row presents the effect of living er odds of low grades and score worse on the with a single mother rather than cohabiting step- vocabulary test. Teenagers living with cohabiting parents on teenage problem behaviors. These re- stepfathers have similar expectations of going to sults indicate how mothers’ cohabitation influenc- college as teenagers living with two married, bi- es teenage well-being among unmarried mothers. ological parents. At the bivariate level, college ex- The first model shows the zero-order or bivariate pectations are lower among teens living with co- effects, and the second model presents the effects habiting stepfathers than teens living with two of family structure, net of the other variables. We biological married parents. The effects of the oth- present the family structure effects for each model er covariates vary across adolescent outcomes. We and then discuss the effects of the remaining co- find that higher levels of family income and moth- variates. er’s education are typically related to higher levels The first column shows that at the bivariate 886 Journal of Marriage and Family

TABLE 3. REGRESSION COEFFICIENTS ESTIMATES OF ADOLESCENT BEHAVIORAL AND ACADEMIC OUTCOMES (N ϭ 13,231) School Low College Suspend/Expela Delinquency Problems GPAb PPVT Expectations Family structure (Married, two biological) Cohabiting stepfather .80*** 1.32** .76*** .64*** Ϫ2.36** Ϫ.10 (.13) (.32) (.17) (.17) (.70) (.06) Married stepfather .56*** .61** .69*** .52*** Ϫ.93 Ϫ.05 (.08) (.21) (.12) (.12) (.47) (.04) Single mother .62*** .95*** .66*** .38*** Ϫ.85* Ϫ.04 (.09) (.19) (.09) (.10) (.40) (.03) Sociodemographic characteristics Race (White) Black .99*** .22 Ϫ.20 0.005 Ϫ9.09*** .10** (.11) (.18) (.14) (.12) (.68) (.04) Hispanic .17 1.02 Ϫ.27 .19 Ϫ7.10*** .08 (.13) (.24) (.17) (.14) (.74) (.05) Other .03 .72 Ϫ.003 Ϫ.18 Ϫ3.42*** .16** (.13) (.26) (.17) (.15) (.86) (.05) Log family income Ϫ.25*** Ϫ.03 .01 Ϫ.20*** 2.16*** .10*** (.06) (.11) (.05) (.05) (.27) (.02) Missing income (no) .01 Ϫ.35 Ϫ.0001 .24* Ϫ1.59** Ϫ.02 (.07) (.20) (.11) (.12) (.51) (.04) Mother’s age Ϫ.01* .005 0.002 Ϫ.01 .03 .01** (.005) (.01) (.006) (.01) (.03) (.002) Mother’s education Ϫ.14*** Ϫ.006 Ϫ.03 Ϫ.13*** 1.44*** Ϫ.07*** (.02) (.03) (.02) (.02) (.11) (.01) Child’s age .11*** Ϫ.12** Ϫ.04 Ϫ.03 Ϫ.30* .07*** (.03) (.04) (.03) (.03) (.12) (.01) Child’s sex (female) .97*** 1.62*** .81*** .51*** 1.36*** Ϫ.17*** (.06) (.11) (.07) (.08) (.30) (.03) Importance of Ϫ.15*** Ϫ.75*** Ϫ.27*** Ϫ.25*** Ϫ.29 .10*** religion to child (.04) (.09) (.06) (.05) (.24) (.02) Number of children in .04 Ϫ.001 .02 0.002 Ϫ.79*** Ϫ.01 household (.03) (.05) (.03) (.03) (.15) (.01) Parenting Monitoring .002 Ϫ.11* Ϫ.04 .01 Ϫ1.14*** Ϫ.03* (.02) (.05) (.03) (.03) (.14) (.01) Closeness to mother Ϫ.17*** Ϫ1.29*** Ϫ.64*** Ϫ.19*** Ϫ.75*** .11*** (.04) (.10) (.05) (.04) (.20) (.02) Intercept Ϫ.45 12.4** 7.80*** 1.36* 91.79*** 3.55*** (.58) (1.07) (.57) (.58) (2.38) (.21) F-valuec Ϫ6585.5*** 25.7*** 26.7*** Ϫ4396.2*** 77.9*** 29.1*** R2d .13 .09 .06 .05 .25 .07 Note: Reference category for variables is presented in parentheses. Unstandardized coefficients are presented, and standard errors are shown in parentheses. PPVT ϭ Peabody Picture Vocabulary Test. aLogistic regression was used for suspended or expelled, 1 ϭ yes. bLogistic regression was employed for low grade point average (1 ϭ low grades). cThe log likelihood is shown for the models predicting suspension or expulsion and low grade point average. dThe R2 is the pseudo R2 for the models predicting suspension or expulsion and low grade point average. *p Ͻ .05. **p Ͻ .01. ***p Ͻ .001.

level, teenagers living in married stepparent fam- parenting variables (closeness to mother and mon- ilies have significantly lower odds of being sus- itoring) reduce the effect of marital status. Thus pended or expelled from school than teens resid- in the multivariate model teens living in married ing in cohabiting stepparent families. The second and cohabiting stepparent families share similar model shows that this family structure effect can odds of being suspended or expelled from school. be explained by the other covariates. No single We shift the reference category to single mothers factor explains the effect of family structure: So- and find that children living in married stepfather ciodemographic variables in conjunction with the families have similar odds of being suspended or Adolescent Well-Being 887

TABLE 4. REGRESSION COEFFICIENTS ESTIMATES OF ADOLESCENT BEHAVIORAL OUTCOMES (N ϭ 5,504)

Suspension/Expulsiona Delinquency School Problems Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Family structure (Cohabiting stepfather) Married stepfather Ϫ.52*** Ϫ.21 Ϫ1.15** Ϫ.68* Ϫ.19 Ϫ.10 (.14) (.15) (.36) (.35) (.22) (.20) Single mother Ϫ.11 Ϫ.06 Ϫ.76* Ϫ.06 Ϫ.27 0.005 (.12) (.14) (.35) (.37) (.19) (.20) Sociodemographic characteristics Race (White) Black .97*** .23 Ϫ.30 (.12) (.25) (.18) Hispanic .11 1.17** Ϫ.33 (.17) (.39) (.21) Other .15 1.00* .16 (.18) (.47) (.30) Log family income Ϫ.22*** .05 .07 (.06) (.15) (.07) Missing income (no) .15 Ϫ.31 .01 (.11) (.27) (.14) Mother’s age Ϫ.01 .01 .02 (.01) (.02) (.01) Mother’s education Ϫ.16*** Ϫ.04 Ϫ.05 (.02) (.05) (.03) Child’s age .07* Ϫ.21*** Ϫ.09* (.03) (.06) (.04) Child’s sex (female) .96*** 2.01*** .94*** (.09) (.22) (.12) Importance of religion to Ϫ.17** Ϫ.72*** Ϫ.25** child (.05) (.15) (.08) Number of children in .05 0.001 .04 household (.03) (.08) (.05) Family stability Number of mother’s .16*** .39** .15* marriages (.04) (.15) (.07) Parenting Monitoring Ϫ.02 Ϫ.13 Ϫ.06 (.02) (.10) (.05) Closeness to mother Ϫ.21*** Ϫ1.18*** Ϫ.55** (.05) (.16) (.07) Closeness to nonresident father Ϫ.06* Ϫ.29*** Ϫ.13** (.03) (.08) (.06) Missing closeness to .02 Ϫ.16 Ϫ.06 nonresident father (no) (.09) (.24) (.12) Intercept Ϫ.35** .73 5.44*** 13.97*** 4.79** 7.95*** F-valueb Ϫ3591.45 Ϫ3225.41 4.84* 11.95*** 2.03 7.73*** R2c .01 .11 .00 .09 .00 .06 Note: Reference category for variables is presented in parentheses. Unstandardized coefficients are presented, and standard errors are shown in parentheses. aLogistic regression was used for suspended or expelled, 1 ϭ yes. b The log likelihood is shown for the models predicting suspension or expulsion. cThe R2 is the pseudo R2 for the models predicting suspension or expulsion and low grade point average. *p Յ .05. **p Յ .01. ***p Յ .001. expelled as their counterparts living in single- pended from school as adolescents living with mother families (results not shown). The next row their mother and her cohabiting partner. This is indicates that adolescents living with single moth- true in both the bivariate and multivariate models. ers have similar odds of being expelled or sus- In terms of delinquency, teens living in married 888 Journal of Marriage and Family stepfather families have significantly lower levels Table 5 shows the effects of cohabitation on than teens living in cohabiting stepfather families. academic well-being, and the table format mirrors The results in the next column suggest that the Table 4. The first column of Table 5 shows that inclusion of the remaining covariates reduces but teenagers living in married stepfather families does not fully explain the marital status effect. have lower odds of earning low grades than teens The multivariate model indicates that teenagers in cohabiting stepfather families. Yet the inclusion living in married rather than cohabiting stepparent of the remaining covariates (income in particular) families have significantly lower delinquency explains this difference. We also do not find sta- scores. We also find that teenagers living with tistical differences between teens living in married married stepfathers have lower levels of delin- stepfamilies and single-mother families (results quency than teens living with single mothers (re- not shown). The next row shows that adolescents sults not shown). living with unmarried mothers who are cohabiting Delinquency is significantly lower among ad- have higher odds of having low grades than teens olescents living with just their mother than those living with single mothers. The inclusion of the living with their mother and her cohabiting part- remaining covariates shifts the relationship be- ner. Yet the next column includes all of the co- tween family structure and grades such that teens variates and shows that these differences are no in cohabiting stepparent and single-mother fami- longer statistically significant. The effect of family lies share similar odds of having low grades. The structure on delinquency is primarily explained by family structure differences are explained by our the number of mother’s marriages. indicator of family stability, the number of moth- The last two columns in Table 4 present the er’s marriages. effects of the covariates on school problems. The The next two columns present the effects of bivariate and multivariate model results show that family structure on verbal ability. At the bivariate teenagers in cohabiting and married stepfather level, adolescents in married stepfather families families have similar levels of school problems. score higher on the vocabulary test than teens in Further analyses indicate that married stepfathers cohabiting stepfather families. The effect of co- and single mothers have similar school problems habitation is reduced with the inclusion of the ex- (results not shown). The next row shows teenagers planatory variables; however, the family effect is living with single mothers and cohabiting partners marginally significant (p ϭ .06). In contrast, teen- share similar levels of trouble in school. agers living in married stepfather and single- The remaining covariates in Table 4 operate in mother families share similar levels of verbal abil- the expected direction and vary somewhat de- ity (results not shown). Adolescents living in pending on the particular outcome. Younger teen- unmarried mother families without cohabiting agers and boys consistently are more likely to ex- partners and with cohabiting partners have statis- perience problems. The indicator of importance of tically similar verbal ability scores, suggesting religion is also negatively associated with problem that teens’ mother’s cohabitation status is not re- behaviors. The greater the number of mother’s lated to cognitive development. marriages, the higher the incidence of problem be- The last two columns focus on college expec- haviors. Closeness to mother as well as closeness tations. The bivariate results demonstrate that ad- to nonresident father are associated with fewer olescents living in married stepfather families pos- problem behaviors. sess higher college expectations than adolescents Further analyses of only teenagers living in living in cohabiting stepfamilies. The final col- stepfamilies reveal that duration of the parental umn, however, shows that these family structure relationship is usually not associated with adoles- differences no longer persist when the remaining cent behavior problems (results not shown). We covariates are included. The positive effect of also tested whether the effects of family type dif- marriage on college expectations reduces to non- fer according to the duration of the parental rela- significance when income or mother’s education tionship. Analyses of interaction effects indicate is included in the model. Similarly, teenagers liv- that the effects of family type differ according to ing with married stepfathers and single mothers duration for only one outcome, school problems do not differ in terms of college expectations (re- (results not shown). The effect of marital status sults not shown). In both bivariate and multivari- on school problems is greater early in the rela- ate models, youth living in cohabiting stepfather tionship and then diminishes at later union dura- families and single-mother families share similar tions. expectations for college. Among children living Adolescent Well-Being 889

TABLE 5. REGRESSION COEFFICIENTS ESTIMATES OF ADOLESCENT ACADEMIC OUTCOMES (N ϭ 5,504) Low Grade Point Peabody Picture Averagea Vocabulary Test College Expectations Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Family structure (Cohabiting stepfather) Married stepfather Ϫ.38* Ϫ.11 4.21*** 1.65 .13* .06 (.18) (.19) (.99) (.86) (.06) (.07) Single mother Ϫ.33* Ϫ.20 .36 1.29 .09 .04 (.16) (.18) (.98) (.80) (.07) (.06) Sociodemographic characteristics Race (White) Black .03 Ϫ8.62*** .11* (.15) (.77) (.05) Hispanic .20 Ϫ6.17*** .03 (.18) (.98) (.08) Other Ϫ.03 Ϫ2.91* .14* (.23) (1.12) (.07) Log family income Ϫ.20** 1.97*** .09*** (.06) (.33) (.02) Missing income (no) .37* Ϫ2.39*** Ϫ.09 (.18) (.64) (.06) Mother’s age Ϫ.01 .03 0.004 (.01) (.05) (.003) Mother’s education Ϫ.09** 1.51*** .05*** (.03) (.14) (.01) Child’s age Ϫ.04 Ϫ.49** Ϫ.09*** (.03) (.16) (.01) Child’s sex (female) .48*** 1.64*** Ϫ.20*** (.11) (.46) (.04) Importance of religion Ϫ.16* Ϫ.58 .11*** to child (.08) (.39) (.03) Number of children Ϫ.03 Ϫ1.00*** Ϫ.01 in household (.04) (.19) (.02) Family stability Number of mother’s .13** Ϫ.37 Ϫ.03 marriages (.05) (.31) (.02) Parenting Monitoring Ϫ.04 Ϫ1.29*** Ϫ.02 (.03) (.20) (.02) Closeness to mother Ϫ.22** Ϫ.96*** .08*** (.07) (.30) (.02) Closeness to nonresident Ϫ.09* Ϫ.08 .04* father (.04) (.23) (.02) Missing closeness to Ϫ.12 Ϫ1.80*** Ϫ.10 nonresident father (no) (.12) (.49) (.05) Intercept Ϫ1.43*** 1.75* 97.74*** 95.39*** 4.28** 4.20*** F-valueb Ϫ1778.52 Ϫ2245.05 .11 45.68*** 2.03 12.54*** R2c .08 .04 .01 .26 .001 .07 Note: Reference category for variables is presented in parentheses. Unstandardized coefficients are presented, and standard errors are shown in parentheses. aLogistic regression was employed for low grade point average (1 ϭ low grades). bThe log likelihood is shown for the models predicting low grade point average. cThe R2 is the pseudo R2 for the models predicting suspension or expulsion and low grade point average. *p Յ .05. **p Յ .01. ***p Յ .001. 890 Journal of Marriage and Family with unmarried mothers, the cohabiting parent ditional data about the relationship between does not appear to improve or worsen adolescents’ cohabiting and married stepfathers’ relationships school aspirations. to their wives and partners’ children may help to In terms of the remaining covariates, we find explain this marriage advantage. We lack mea- minority youth more often have lower academic surement of role ambiguity, which may serve to outcomes than Whites. Mother’s education, family distinguish parenting roles in cohabiting and mar- income, and religiosity are associated with higher ried stepfamilies. Married stepfathers may have a academic achievement. Boys have lower college more clearly defined obligation to their stepchil- expectations and grades than girls. Closeness to dren than cohabiting stepfathers (Hofferth & An- mothers and nonresident fathers is related to high- derson, 2003). The act of remarriage may carry er college expectations and grades. Additional with it a more pronounced expectation of stepfa- analyses of just teenagers in stepfamilies show ther involvement (e.g., spending time with step- that the quality of parental relationships and du- children and contributing financially to their up- ration of parental relationship are not associated bringing) that has positive consequences for child with most adolescent academic outcomes. One ex- well-being. ception is that duration of stepparent’s relationship The results from this paper suggest that teen- is positively tied to adolescent college expecta- agers living with unmarried mothers do not seem tions. to benefit from the presence of their mother’s co- habiting partner. We argued at the outset that it may be important to distinguish between unmar- DISCUSSION ried mothers who are cohabiting and those living Recent debates have emerged about the advantage alone. In terms of adolescent outcomes, we do not of marriage for adults and children (e.g., Waite & appear to gain much by distinguishing between Gallagher, 2000). Adolescents in married, two-bi- cohabiting stepfather and single-mother families. ological-parent families generally fare better than We do find differences at the bivariate level, how- children in any of the family types examined here, ever, in terms of delinquency and low grades in including single-mother, cohabiting stepfather, and school. Thus, as found in the stepfamily literature married stepfather families. The advantage of (e.g., Coleman et al., 2000), men’s presence alone marriage appears to exist primarily when the child seems neither sufficient nor necessary to create is the biological offspring of both parents. Our positive outcomes for children. Indeed, our results findings are consistent with previous work, which show that stepfathers (married or cohabiting) pro- demonstrates children in cohabiting stepparent vide limited benefit when contrasted with single- families fare worse than children living with two mother families. Our findings suggest that neither married, biological parents (e.g., Acs & Nelson, parental cohabitation nor marriage to a partner or 2002; Brown, 2001; DeLeire & Kalil, 2002; Hao spouse who is not related to the child (stepfamily & Xie, 2001). formation) is associated with uniform advantage Researchers argue that we need to expand our in terms of behavioral or academic indicators to traditional understanding of stepfamily life to in- teenagers living in single-mother families. These clude cohabiting stepfamilies (Stewart, 2001). The results are consistent with research focusing on marital status of men in stepfamilies appears to behavior problems (Acs & Nelson, 2002; Morri- influence adolescent well-being. Among adoles- son, 1998). Our findings are not consistent with cents living in stepfamilies, those living with mar- Nelson et al. (2001) who reported negative effects ried rather than cohabiting mothers are sometimes of parental cohabitation. One explanation may be advantaged, although this is not consistent across that we explain our negative effects of parental all outcomes. At the bivariate level, teenagers liv- cohabitation on delinquency and grade point av- ing with married stepfamilies experience more erage by mother’s marriage history, a variable that positive behavioral and academic outcomes (ex- is not included in the data set used by Nelson et cept school problems), than teens living in cohab- al. iting stepfamilies. Yet, at the multivariate level, We attempt to capture the fluidity and stability many of the observed family structure differences of families. Our core measure of family stability, can be explained by the covariates in our models. the number of the mother’s prior marriage-like re- Differences in delinquency attributable to cohab- lationships (during the child’s lifetime), is an im- itation and marital status, however, cannot be ex- portant contributor to children’s well-being. Moth- plained by the factors included in our model. Ad- er’s relationship history is related to many Adolescent Well-Being 891 adolescent outcomes. In fact, this measure ex- structure of families, but also for the nature of plains differences in delinquency and low grade relationships that exist within and across house- point average among teenagers living with cohab- holds. Another measure, which could be consid- iting stepfathers and single mothers. This is con- ered to be part of family life socialization, is re- sistent with researchers who emphasize the im- ligiosity, and we observe similar levels across portance of family stability rather than family family types. We find that the teens who were structure for predicting child well-being (Hao & more religious than other teens fared better in Xie, 2001; Hill et al., 2001; Wu & Martinson, terms of behavior and academic outcomes, but 1993). We also evaluate whether family structure this variable does not explain the effects of family effects differ according to duration of the rela- structure. tionship. In stepfamilies, duration of the current This paper suffers from several shortcomings. relationship is only related to college expectations. First, we employ cross-sectional data, so our find- Perhaps the stability of the relationship reflects the ings are suggestive because longitudinal analyses stepfather’s willingness to provide financial assis- are necessary to accurately evaluate how parental tance for college. This is similar to findings re- cohabitation or marriage causes changes in an ad- ported by Hao and Xie (2001), that time spent in olescent’s well-being (see Hao & Xie, 2001). For the current union is not associated with child mis- example, we may find that mothers with children behavior. We find that family structure effects do who have greater behavior problems and poor not differ according to duration of the stepparent’s school performance are more likely to cohabit relationship, except for school problems. This than marry. Thus, there could be selection into suggests that the effect of cohabitation is typically family types based on the adolescent behaviors. In similar when stepfamilies first form and during a similar vein, the causal nature of the covariates later years. is not clearly specified in our models. Our covar- We try to account for economic status (moth- iates represent factors that may be related to entry er’s education and family income), and similar to into specific types of families (e.g., education or prior studies find that economic circumstances are religiosity) as well as effects of family structure associated with adolescent well-being (e.g., Dun- (e.g., income). We are not able to account for se- can & Brooks-Gunn, 1997). These factors are par- lection in our models, but we believe that we have ticularly important for understanding differences provided important baseline information about pa- in the effect of cohabitation in stepfamilies. Most rental cohabitation and adolescent well-being. of the bivariate differences based on parental mar- Second, some potentially important variables are ital or cohabitation status in stepfamilies are ex- omitted from our analyses. Measures that tap into plained by socioeconomic factors (e.g., family in- stepfamily processes, such as relationships with come, race or ethnicity, mother’s education, cohabiting stepfathers or parenting problems in child’s sex and age). Thus, the higher levels of stepparent families, are not available in the Add mother’s education and family income observed Health. As discussed above, stepfathers who are in married stepfather families explains some of the cohabiting may face quite different parenting cir- differences in child outcomes in stepfather fami- cumstances than stepfathers who are married. An- lies. other factor that is associated with child well-be- Our findings also speak to how parenting and ing and found to be important among cohabiting the complexity of family influence children’s families is maternal depression (Brown, 2001). lives. Parental control is not uniformly associated Unfortunately, measures of maternal depression with better teenage outcomes, but this measure is are not included in the Add Health. Finally, our not capturing early adolescent parenting and fo- measure of economic circumstances is far from cuses narrowly on limit setting. With regard to ideal. There is a high level of missing data on parental support, we find that closeness of teens family income in the Add Health. We hoped to to their biological mothers and nonresident fathers alleviate this limitation by accounting for mother’s is positively related to many indicators of adoles- education, but acknowledge it is not a substitute cent well-being and is more often a significant for income. predictor of adolescent outcomes than parental The issue of cohabitation and child develop- monitoring. Hence, our findings appear to be more ment has become more important as cohabitation consistent with attachment than with social control has become an increasingly large part of chil- theories of child development. Our work suggests dren’s family experiences (Bumpass & Lu, 2000; that it is important to account not only for the Graefe & Lichter, 1999). The findings from this 892 Journal of Marriage and Family paper represent an initial step toward understand- Brown, S. (2002). Child well-being in cohabiting fam- ing the implications of parental cohabitation on ilies. In A. Booth & A. Crouter (Eds.), Just living together: Implications of cohabitation for children, adolescent well-being. Research that focuses on families and social policy (pp. 173–188). Mahwah, younger children and examines the well-being of NJ: Lawrence Erlbaum. children born into cohabiting parent families is Buchanan, C. M., Maccoby, E. E., & Dornbusch, S. M. warranted. Future efforts must consider potential (1996). Adolescents after divorce. Cambridge, MA: Harvard University Press. selection issues from both the adult’s and child’s Bumpass, L. (1994,December). The declining signifi- perspective as well as model the fluid nature of cance of marriage: Changing family life in the United cohabiting unions. States. Paper presented at the Potsdam International Conference, The Netherlands. Bumpass, L., & Lu, H. (2000). Trends in cohabitation NOTE and implications for children’s family contexts. Pop- This research was supported in part by the Center for ulation Studies, 54, 29–41. Family and Demographic Research, Bowling Green Carlson, M., & Corcoran, M. (2001). Family structure State University, which has core funding from the Na- and children’s behavioral and cognitive outcomes. tional Institute of Child Health and Human Develop- Journal of Marriage and Family, 63, 779–792. ment (R21 HD042831-01). This paper was presented at Chantala, K., & Tabor, J. (1999). Strategies to perform the annual meeting of the American Sociological As- a design-based analysis using the Add Health data. sociation in Annaheim, California, August 2001. The Chapel Hill, NC: Carolina Population Center, Uni- authors have benefited from comments provided by par- versity of North Carolina. ticipants at the Ohio State University Initiative in Pop- Coleman, M., Ganong, L., & Fine, M. (2000). Reinves- ulation Research; the University of Chicago Alfred P. tigating remarriage: Another decade of progress. Sloan Center on Parents, Children, and Work; the Office Journal of Marriage and the Family, 62, 1288–1307. of Population Research at Princeton University; and the DeLeire, T., & Kalil, A. (2002). Good things come in University of Texas Population Research Center. In ad- 3s: Single-parent multigenerational family structure dition, Susan Brown, Larry Bumpass, Steven Demuth, and adolescent adjustment. Demography, 39, 393– Peggy Giordano, Monica Longmore, Laura Sanchez, 342. Pamela Smock, and Susan Stewart have provided valu- Duncan, G., & Brooks-Gunn, J. (Eds.). (1997). Conse- able comments and suggestions. This research uses data quences of growing up poor. New York: Russell Sage from Add Health, a program project designed by J. Foundation. Richard Udry, Peter S. Bearman, and Kathleen Mullan Dunifon, R., & Kowaleski-Jones, L. (2002). Who’s in Harris, and funded by grant P01-HD31921 from the Na- the house? Race differences in cohabitation, single tional Institute of Child Health and Human Develop- parenthood and child development. Child Develop- ment, with cooperative funding from 17 other agencies. ment, 73, 1249–1264. Special acknowledgment is due Ronald R. Rindfuss and Durst, R. (1997). Ties that bind: Drafting enforceable Barbara Entwisle for assistance in the original design. cohabitation agreements. New Jersey Law Journal, Contact Add Health, Carolina Population Center, 123 147, S13–S14. West Franklin Street, Chapel Hill, NC 27516-2524 Fields, J. (2001). Living arrangements of children. Cur- (www.cpc.unc.edu/addhealth/contract.html) to obtain rent Population Reports, 70–74. the data. Fields, J., & Casper, L. (2000). America’s families and living arrangements. Current Population Reports, 20–537. REFERENCES Graefe, D. R., & Lichter, D. T. (1999). Life course tran- Acs, G., & Nelson, S. (2001). Honey, I’m home. Chang- sitions of American children: Parental cohabitation, es in living arrangements in the late 1990s (New Fed- marriage, and single parenthood. Demography, 36, eralism National Survey of America’s Families B- 205–217. 38). Washington DC: Urban Institute. Hanson, T., McLanahan, S., & Thomson, E. (1997). Acs, G., & Nelson, S. (2002). The kids alright? Chil- Economic resources, parental practices, and chil- dren’s well-being and the rise in cohabitation (New dren’s well-being. In G. J. Duncan & J. Brooks-Gunn Federalism National Survey of America’s Families B- (Eds.), Consequences of growing up poor (pp. 190– 48). Washington DC: Urban Institute. 238). New York: Russell Sage Foundation. Amato, P. R., & Gilbreth, J. G. (1999). Nonresident fa- Hao, L., & Xie, G. (2001). The complexity and endo- thers and children’s well-being: A meta-analysis. genity of family structure in explaining children’s Journal of Marriage and the Family, 61, 557–573. misbehavior. Social Science Research, 31, 1–28. Baumrind, D. (1991). Parenting styles and adolescent Hill, M., Yeung, W., & Duncan, G. (2001). Childhood development: In R. Lerner, A. C. Petersen, & J. family structure and young adult behaviors. Journal Brooks-Gunn (Eds.), Encyclopedia of adolescence of Population Economics, 14, 271–299. (pp. 746–759). New York: Garland. Hofferth, S., & Anderson, K. (2003). Are all dads Bearman, P., Jones, J., & Udry, R. (1997). The National equal? Biology versus marriage as a basis for paternal Longitudinal Study of Adolescent Health. Chapel Hill, investment. Journal of Marriage and Family, 65, NC: University of North Carolina. 213–232. Brown, S. (2001,March). Child well-being in cohabiting Mahoney, M. (2002). The economic rights and respon- unions. Paper presented at the Annual Meeting of the sibilities of unmarried cohabitants. In A. Booth & A. Population Association of America, Washington, DC. Crouter (Eds.), Just living together: Implications of Adolescent Well-Being 893

cohabitation on families, children, and social policy two-parent family: How teenagers fare in cohabiting (pp. 247–254). Mahwah, NJ: Lawrence Erlbaum. couple and blended families. (New Federalism Na- Manning, W. (2002). The implications of cohabitation tional Survey of America’s Families B-31). Washing- for children’s well-being. In A. Booth & A. Crouter ton, DC: Urban Institute. (Eds.), Just living together: Implications of cohabi- O’Connor, T. G., Hetherington, E. M., & Clingempeel, tation on families, children, and social policy (pp. W. G. (1997). Systems and bi-directional influences 121–152). Mahwah, NJ: Lawrence Erlbaum. in families. Journal of Social and Personal Relation- Manning, W., & Lichter, D. (1996). Parental cohabita- ships, 14, 491–504. tion and children’s economic well-being. Journal of Stewart, S. (2001). Contemporary American steppar- Marriage and the Family, 58, 998–1010. enthood: Integrating cohabiting and nonresident step- Manning, W., Smock, P., & Majumdar, D. (2003). The parents. Population Research and Policy Review, 20, relative stability of cohabiting and marital unions for 345–364. children. Population Research and Policy Review. Thomson, E., Hanson, T., & McLanahan, S. (1994). McLanahan, S. (1997). Parent absence or poverty: Family structure and child well-being: Economic re- Which matters more? In G. Duncan & J. Brooks- sources vs. parental behavior. Social Forces, 73, 221– Gunn (Eds.), Consequences of growing up poor (pp. 242. 35–48). New York: Russell Sage Foundation. Thomson, E., McLanahan, S., & Curtin, R. B. (1992). McLanahan, S., & Sandefur, G. (1994). Growing up Family structure, gender, and parental socialization. with a single parent: What hurts, what helps. Cam- Journal of Marriage and the Family, 54, 368–378. Udry, J. R. (1998). The National Longitudinal Study of bridge, MA: Harvard University Press. Adolescent Health (Add Health), Waves I & II, 1994– Morrison, D. R. (1998,March). Child well-being in step 1996 [Data file]. Chapel Hill, NC: Carolina Popula- families and cohabiting unions following divorce: A tion Center, University of North Carolina. dynamic appraisal. Paper presented at the Annual Waite, L. J., & Gallagher, M. (2000). The case for mar- Meeting of the Population Association of America, riage. New York: Doubleday. Chicago, IL. White, L., & Gilbreth, J. (2001). When children have Morrison, D. R. (2000,March). The costs of economic two fathers: Effects of relationships with stepfathers uncertainty: Child well-being in cohabitating and re- and noncustodial fathers on adolescent outcomes. married unions following parental divorce. Paper Journal of Marriage and Family, 63, 155–167. presented at the Annual Meeting of the Population Wu, L., & Martinson, B. (1993). Family structure and Association of America, Los Angeles, CA. the risk of a premarital birth. American Sociological Nelson, S., Clark, R., & Acs, G. (2001). Beyond the Review, 58, 210–232.