Annotatated Bibliography

Tanya Rouleau Whitworth University of Massachusetts Amherst

Teen Childbearing and Depression: Do Pregnancy

Attitudes Matter?

The relationship between teen childbearing and depression has been extensively studied; how- ever, little is known about how young women’s own attitudes toward becoming pregnant shape this association. This study used data from the National Longitudinal Study of Adolescent Health to investigate whether the relationship between teen childbearing and adult depression is moderated by adolescent attitudes toward becoming pregnant. The results showed that although, on average, women who had first births between ages 16 and 19 experienced no more depressive symptoms in adulthood than women who had first births at age 20 or older, the relationship between teen childbearing and adult depression varied significantly based on adolescent pregnancy attitudes. When they had negative adolescent attitudes toward getting pregnant, teen mothers had similar levels of depression as adult mothers, but when they had positive adolescent pregnancy attitudes, teen mothers actually had fewer depressive symptoms than women with adult first births.

Teenage childbearing is a social problem that came to the forefront of public consciousness in the 1970s and continues to hold the atten- tion of academics, policy makers, and the public.

Department of Sociology, 7th Floor Thompson Hall, University of Massachusetts, Amherst, MA 01003 ([email protected]).

This article was edited by Kelly Raley.

Key Words: adolescent childbearing, depression, fertility, mental health, stress.

Although the teen birth rate has declined in the United States for the past 2 decades, the United States continues to have the highest teen birth rate among industrialized countries (Kearney & Levine, 2012). In the United States in 2014, an estimated 27 of 1,000 women aged 15 to 19 years experienced a teen birth (Child Trends Databank, 2015b). Adolescent births are con- centrated among the most disadvantaged girls in our society. For example, birth rates are higher among poor adolescents than among their more advantaged peers, and birth rates are higher among Black and Hispanic adolescents than among White adolescents (Child Trends Data- bank, 2015b; Kearney & Levine, 2012; Mollborn & Morningstar, 2009; Trent & Crowder, 1997). Moreover, 89% of births to girls aged 15 to 19 years that occurred in the United States in 2014 were to unmarried mothers (Child Trends Data- bank, 2015b), and survey data from 2006 to 2010 showed that only 46% of teens who had nonmar- ital births were cohabiting with a partner at the time (Child Trends Databank, 2015a).

The general consensus in public opinion holds that teen childbearing is bad for the individ- ual and for society, especially when it takes place outside of marriage (Furstenberg, 2003, 2007; Luker, 1996; Nathanson, 1991). Support for this position has come from research find- ing that adolescent childbearing and nonmarital childbearing are both associated with maternal depression (e.g., Kalil & Kunz, 2002; Mirowsky & Ross, 2002). Maternal depression is a seri- ous public health problem, not only harming the women themselves but also increasing the risk that their children will develop poor health or

390 Journal of Marriage and Family 79 (April 2017): 390–404 DOI:10.1111/jomf.12380

Teen Childbearing and Depression 391

behavior problems (Meadows, McLanahan, & Brooks-Gunn, 2007; Turney, 2011a, 2011b).

A major limitation of past studies of the relationship between teen childbearing and depression is their failure to take into account the heterogeneity of teenage women’s experiences with childbearing. Specifically, past research has neglected the role of adolescents’ attitudes in shaping their reactions to having teenage births. Overlooking the role of attitudes seems particularly problematic in light of the fact that both quantitative and qualitative research shows some adolescents report positive or ambivalent attitudes toward teenage pregnancy (Barber, Yarger, & Gatny, 2015; Edin & Kefalas, 2005; Geronimus, 2003; Jaccard, Dodge, & Dittus, 2003; SmithBattle, 1995). In the current study, I examined whether teen childbearing remains associated with depression when adolescent pregnancy attitudes are taken into account. I analyze this question using the frameworks of the life course perspective, stress process theory, self-discrepancy theory, and expectancy-value theory. To my knowledge, no study has inves- tigated the relationship between pregnancy attitudes and depression among women who had teen births. This article will make important con- tributions to the literature on teen childbearing and to scholarship on the importance of fertility attitudes and expectations for mental health.

Literature Review

Despite decades of research on the relationship between teen childbearing and mental health, there is still debate in the literature about the nature of this relationship. Scholars in the life course perspective have argued that women who have teen births risk poorer mental health than women who follow the “normative” life course order (Elder, 1975; Wickrama, Conger, Lorenz, & Jung, 2008). The life course perspective focuses on role entries and exits over the life course, assumes that a normative life course order exists and people are aware of it, and suggests that deviations from the normative life course order have negative consequences (Elder, 1975).

Stress process theory similarly predicts that teen childbearing increases women’s risk of depression when compared with having a child in adulthood at a more normative age (Pearlin, Schieman, Fazio, & Meersman, 2005). According to this theory, individual variation

in exposure to stressors—both life events and chronic strains—is said to predict levels of psychological distress, although the mental health impact of stressors can be buffered by access to social, emotional, and financial resources (Pearlin, 1999). Importantly, stress process theory holds that becoming a parent at any age is a major life event stressor, and research demonstrates that parents in general are more depressed on average than nonparents (Evenson & Simon, 2005). Yet, the theory also suggests that teen mothers are more likely to be depressed because they experience more stres- sors, including financial strain, work–family conflict, and barriers to pursuing education (Mirowsky & Ross, 2002; Pearlin et al., 2005; SmithBattle, 2007), and have access to fewer resources—especially financial resources—to buffer against stressors than women who have children later (Pearlin, 1989). For example, research has convincingly shown that women who experience teen births end up with lower educational attainment and lower income than women who do not experience teen births, although this relationship is not definitively causal (e.g., Fletcher & Wolfe, 2009; Hoffman, Foster, & Furstenberg, 1993).

Consistent with the prediction of stress pro- cess theory, research has shown that adolescent childbearing is associated with poorer maternal mental and physical health. Teenage childbear- ing has been associated with higher depressive symptoms in adulthood (Mirowsky & Ross, 2002), poorer midlife physical health (Taylor, 2009), and an increased risk of death at older ages (Henretta, 2007). Considering that the vast majority of adolescent births are nonmarital, it is also relevant that research has found a negative association between nonmarital childbearing and maternal mental and physical health (Avi- son, Ali, & Walters, 2007; Williams, Sassler, Frech, Addo, & Cooksey, 2011). Avison et al. (2007) found that single mothers experienced more stressors than married mothers and that these stressors predicted the higher levels of psychological distress observed among single mothers.

Importantly, other researchers who have examined the relationship between adolescent childbearing and mental health have come to a different conclusion. According to these stud- ies, any negative association between teenage childbearing and mental health can be explained by the selection of disadvantaged women into

392 Journal of Marriage and Family

teen childbearing. For example, Mollborn and Morningstar (2009) showed that, when com- pared with their peers who delayed childbearing, young women who had teen births were more depressed both before and after the transition to parenthood. Examining mental health at midlife, Taylor (2009) found no difference in depres- sive symptoms or positive well-being between teen child bearers and adult child bearers, after controlling for family of origin social class.

To complicate matters further, other researchers have theorized positive or neutral impacts of teen childbearing on mental health for some young women. Edin and Kefalas (2005) interviewed poor mothers and found that in the context of diminished opportunities for achieving conventional “middle-class” mark- ers of adulthood, many poor young women saw motherhood as a source of adult status, meaning, and purpose in life. From a theoretical perspec- tive, parenthood could increase one’s sense of mattering to others as well as one’s sense of meaning and purpose (Thoits, 2011; Umberson & Montez, 2010). In turn, mattering has been shown to predict depressive symptoms (Taylor & Turner, 2001), and feelings of purpose and meaning in life may promote health and healthy behaviors (Umberson & Montez, 2010). Inter- views with women who had teen births revealed that many saw the transition to parenthood as a turning point where they committed to improv- ing their lives for their children (Barcelos & Gubrium, 2014; SmithBattle, 1995, 2007). All this is not to say that poor women valorize teen childbearing or find it unproblematic; in fact, stigma against teen childbearing coexists with acceptance in many communities, and even teen mothers with positive experiences acknowledge the challenges they have faced (Barber et al., 2015; Barcelos & Gubrium, 2014; Geronimus, 2003; Kaplan, 1997; SmithBattle, 1995). What this body of research does show, said best by SmithBattle (1995), is the following: “These young mothers’ sense of gaining or losing, of doing better or worse, of becoming a better per- son or remaining adrift, was interpreted against the background of what preceded the pregnancy and what they imagined for themselves in the future” (p. 32).

The lack of consensus in the literature about the consequences of teen childbearing for depression suggests a need for additional research. Neither approach (i.e., those that directly address selection or those that ignore

it) has considered the likely situation of hetero- geneity in the causal effect of teen childbearing on depression. If it is the case that teen childbear- ing undermines mental health for some women but not for others, then differences in sample selection and analytical approach—including the omission of key variables—could be respon- sible for the competing findings. Therefore, research is needed that explores important sources of heterogeneity in the causal effect. This article focuses on one possible source of heterogeneity: individual fertility desires and attitudes toward teen childbearing.

Fertility Attitudes, Parenthood, and Depression

Several central tenets of stress process theory lead to predictions of substantial heterogeneity in the average association of teen childbearing with depression as a function of women’s prior attitudes toward teen pregnancy. Proponents of stress process theory believe that health disparities can be explained by differential exposure to stressors (Pearlin et al., 2005); how- ever, the extent to which an event is stressful depends on individual subjective appraisals (Lazarus & Folkman, 1984). For example, Martinez-Torteya, Bogat, von Eye, Levendosky, and Davidson (2009) found that among women who experienced intimate partner violence, subjective stressfulness appraisal was a better predictor of depressive symptoms than intimate partner violence frequency or severity. In addi- tion, stress researchers maintain that life events are especially stressful if they are undesirable or unexpected and that unwillingly occupied roles produce a specific type of chronic strain called role captivity (Pearlin, 1989, 1999). The impli- cation is that having a child as an adolescent will only lead to distress if it is experienced as a stressor, which requires girls to appraise it as a negative event or experience role captivity.

Self-discrepancy theory also leads to the pre- diction that attitudes toward becoming pregnant will affect depression levels after the transition to parenthood. According to self-discrepancy theory, individuals compare their actual selves to their ideal selves, and it is potentially prob- lematic if there are discrepancies between the two (Higgins, 1987). Higgins (1987) wrote that “if a person possesses this discrepancy, the current state of his or her actual attributes, from the person’s own standpoint, does not match the ideal state that he or she personally hopes

Teen Childbearing and Depression 393

or wishes to attain … and thus the person is predicted to be vulnerable to dejection-related emotions” (p. 322). From this perspective, an individual who has a positive attitude toward becoming pregnant as a teenager will have little discrepancy between her actual and ideal selves if she does have a teen birth. Consequently, teen childbearing should be much less detrimental to depression levels in this situation.

There is a small body of literature that exam- ines the mental health effects of meeting or deviating from one’s parenthood timing expec- tations (Carlson, 2011; Carlson & Williams, 2011; Mossakowski, 2011). These three studies all used data from the 1979 National Longitudi- nal Survey of Youth (NLSY79; http://www.bls. gov/nls/nlsy79.htm) and took advantage of the fact that NLSY79 respondents were asked about their fertility expectations during ado- lescence, before the transition to parenthood. Mossakowski (2011) found that individuals who became parents unexpectedly as teenagers or early adults were significantly more depressed in adulthood than individuals who met their parenthood timing expectations, even when controlling for baseline mental health. Simi- larly, Carlson and Williams (2011) found that, on average, first births that occurred earlier than expected were associated with higher depressive symptoms than first births that occurred at the age they were expected. Carlson (2011), how- ever, demonstrated that deviation from expected birth timing did not explain much of the differ- ence in depressive symptoms between women who had teen or early births and women who had their first births after the age of 21. This suggests that other factors may better explain mental health differences between women who have teen first births and women who have their first children in adulthood.

A limitation of the studies by Carlson (2011), Carlson and Williams (2011), and Mossakowski (2011) is that they relied exclusively on NLSY79 data. The NLSY79 is an excellent longitudinal, nationally representative data set; however, the cohort it represents is in later midlife today. It is important to study issues of family for- mation and mental health in younger cohorts as well. Furthermore, although fertility timing expectations have been empirically useful, there are other theoretically important measures of ideation about fertility and parenthood that were not measured in the NLSY79 (Carlson, 2015). According to expectancy-value theory, both

expectations and values (including specific atti- tudes, preferences, and assessments of cost) are important predictors of behavior (Carlson, 2015; Feather & Newton, 1982; Wigfield & Eccles, 2000). Indeed, previous research has identi- fied fertility attitudes and fertility expectations as distinct concepts. For example, one could expect to become pregnant without necessarily having a positive attitude or preference about becoming pregnant (see Barber et al., 2015). In concluding a recent article, Carlson (2015) wrote, “preferences may be more predictive of future behavior than expectations” (p. 13). Thus, I argue that research is needed that assesses the role of specific fertility attitudes and preferences in moderating the relationship between age at first birth and depression. The preceding review leads to the following hypothesis:

(a) Depressive symptoms associated with teen childbearing will be greatest among women who had negative attitudes toward becoming pregnant in adolescence, and (b) differences in depressive symptoms between those who had teen first births and those who delayed childbearing until adult- hood will diminish as adolescent pregnancy atti- tudes become more positive.

Method

Sample

Data were from Wave 1 and Wave 4 of the National Longitudinal Study of Adolescent Health (Add Health; Harris, 2009). Add Health is a longitudinal survey of adolescents based on a nationally representative sample of U.S. schools. The first wave was conducted in 1995 when par- ticipants were in Grades 7 to 12. The second wave was conducted in 1996 when participants were in Grades 8 to 12, and the third wave was conducted in 2002 when participants were aged 18 to 28 years. The fourth and most recent wave was conducted in 2008 when participants were aged 24 to 34 years. Of the original 20,745 Wave 1 respondents, 15,701 participated in Wave 4 and 14,800 had nonmissing sample weights.

The sample for this study consisted of female respondents who were aged 15 to 19 years at Wave 1, participated in both Waves 1 and 4, and reported a first live birth between Waves 1 and 4. For the purpose of the current study, it was necessary for the adolescent pregnancy attitude measure to temporally precede the first birth;

394 Journal of Marriage and Family

therefore, 328 respondents were excluded from the sample because they reported that their first live birth occurred before Wave 1 or within 9 months after Wave 1. Bivariate statistical tests revealed that, when compared with women who had teen births that were included in the analy- sis, these women were significantly more likely to be Black, more likely to be from families receiving public assistance at Wave 1, less likely to know the education level of either parent, less likely to be from two-parent families, more likely to be from “other” family structures, and less likely to be married or cohabiting at their first birth. They also had their first births at age 16.6 on average, almost two years earlier than the mean age at first birth of the teen mothers in the analytic sample (18.2 years). Compared to the teen mothers in the analytic sample, the women who had births before Wave 1 reported significantly more positive attitudes toward adolescent pregnancy at Wave 1, which makes sense in light of the fact that they had already experienced a teen birth. Importantly, the women who had births before Wave 1 were no more depressed at Waves 1 or 4 than the teen mothers in the analytic sample. Nonetheless, because of the exclusion of women who had births before Wave 1, it is possible that the findings presented in this article underestimate the effect of teen childbirth on adult depression.

With the exception of measures of Wave 1 family socioeconomic status, for which “un- known” categories were created, missing data were dealt with using listwise deletion. The majority of missing cases were a result of miss- ing data for adolescent pregnancy attitudes. Of the women who had first births between Waves 1 and 4, 27% were missing responses to this ques- tion. All but 13 of these missing responses were attributable to a survey skip pattern that omitted the pregnancy attitudes questions for respon- dents who were younger than 15 years old. This is the reason that the analytic sample was restricted to those who were aged 15 to 19 years at Wave 1. It would have been inappropriate to impute missing data for pregnancy attitudes because the data for this measure were not miss- ing at random, and it was one of the primary independent variables. In addition, 1% of the sample-eligible cases were dropped because of missing data on other analytic variables. In sup- plementary analyses (not shown) using multiple imputation to handle this small amount of

missing data, results did not differ meaningfully from those presented.

In total, 430 women who had teen first births between Waves 1 and 4 were excluded from the analysis because of missing data, primarily because they were too young at Wave 1 to be asked about pregnancy attitudes. Bivariate sta- tistical tests revealed few significant differences between these women and the teen mothers who were included in the analysis. The women who had teen first births but were excluded from the analysis because of missing data were younger on average at Wave 1 than the teen mothers in the analytic sample, which can be explained by the fact that younger girls were not asked about pregnancy attitudes (the most common reason for missing data). They also had their first birth at age 17.5 on average, whereas the teen mothers in the analytic sample had their first birth at age 18.2 on average.

After listwise deletion of respondents with missing data for pregnancy attitudes or any other of the analytic variables, there were 2,898 women in the sample, including 592 women who had teen first births and 2,306 women who had adult first births. Age at first birth in the analytic sample ranged from 16 to 19 years old for the women who had teen first births and from 20 to 32 years old for the women who had adult first births. Research demonstrates that parents are more depressed on average than nonparents (Evenson & Simon, 2005). Thus, women who had adult first births were the most appropriate comparison group. Supplementary analyses (not shown) using women who remained childless at Wave 4 as the comparison group produced the same pattern of results.

Dependent Variable

Adult depressive symptoms were measured at Wave 4 using a 10-item version of the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977). The CES-D asks respondents to rate how often they experienced certain feelings from 0 = never or rarely to 3 = most of the time or all of the time. For example, one item asks how often “you felt that you were too tired to do things.” Responses to the 10 items were summed to create a scale with possible values from 0 to 30. The scale reliability coefficient was .85. For ease of interpretation, the CES-D was analyzed in untransformed form; supplementary analyses

Teen Childbearing and Depression 395

(not shown) showed that using the natural log of depressive symptoms did not change the pattern of results.

Explanatory Variables

Adolescent pregnancy attitudes were measured at Wave 1 by a question that asked respondents how much they agreed with the following state- ment: “It wouldn’t be all that bad if you got pregnant at this time in your life.” Responses to this question were reverse-coded so that 0 = strongly disagree and 4 = strongly agree. Thus, a higher numerical value for this item represented more positive attitudes toward teen pregnancy. This item was chosen over several other items measuring pregnancy attitudes and costs because it had the highest face validity. One other item I considered measured agree- ment with the following statement: “Getting pregnant at this time in your life is one of the worst things that could happen to you.” Supple- mentary analyses (not shown) using this item produced a similar pattern of results.

A respondent was classified as having experi- enced a teen birth if she was younger than age 20 at the date of her first live birth. Each respondent provided a pregnancy history at Wave 4, which was used to identify the date of her first live birth. The child’s birth date was then compared with the respondent’s birth date to calculate the respondent’s age at first birth.

Control Variables

Adolescent depressive symptoms were mea- sured at Wave 1 using a 19-item version of the CES-D. For consistency, I used only the 10 items that were administered at Wave 4 when constructing the Wave 1 CES-D scale (including all 19 Wave 1 items did not affect the results). Responses to the 10 items were summed to create a scale with possible values from 0 to 30. The scale reliability coefficient was .83.

Race was measured with dummy variables constructed from two questions: the first asked if the respondent was of Hispanic ethnicity and the second asked the respondent to identify with one or more racial categories. As suggested by the Add Health documentation, the following order of preference was used to assign each respon- dent to a single racial or ethnic category: His- panic, Black, Asian, Native American, other, White. Because of the relatively small number of Asian and Native American respondents, these

racial categories were combined with the exist- ing “other” category. The reference category in all analyses was non-Hispanic White.

The following two Wave 1 measures of family socioeconomic status were included in the models: family receipt of public aid and parental education. An indicator of whether the adolescent’s family received public aid was constructed from parents’ responses to one item asking “Are you receiving public assistance, such as welfare?” and five items asking whether they received (a) Supplemental Security Income, (b) Aid to Families with Dependent Children, (c) food stamps, (d) unemployment or worker’s compensations, or (e) a housing subsidy or public housing in the past month. If a parent answered “yes” to receiving any of these six types of aid, the adolescent was coded as receiv- ing public aid. Because of the high number of missing responses to this question, I also included a dummy variable for unknown family receipt of public aid. The reference category in all analyses was no family receipt of public aid. The parental education variable was constructed from the adolescents’ reports of educational attainment for their resident mother and resident father. The highest nonmissing value for res- ident mother’s and resident father’s education was used to create a single parental education variable. Parental education was included in the analyses as a series of dummy variables, including a dummy variable for unknown parental education. The reference category in all analyses was high school parental education.

Attitudes toward teen pregnancy and the like- lihood of teen birth vary according to adolescent family structure and religiosity (Hayford & Mor- gan, 2008; Trent & Crowder, 1997); therefore, Wave 1 measures of family structure and reli- giosity were included in the models as controls. The measure of adolescent family structure was constructed from adolescent reports on the Wave 1 household roster. All adolescents were classi- fied as living in one of the following four family structures: (a) two biological or adoptive parents, (b) stepfamily (including cohabiting stepfami- lies), (c) single parent, or (d) other family struc- ture. The reference category in all analyses was two biological or adoptive parents. The religios- ity item asked respondents to report how impor- tant religion was to them. Responses to the item were reverse coded so that 1 = not important at all and 4 = very important. Respondents who were missing responses to this item because they

396 Journal of Marriage and Family

reported that they had no religion were coded as “not important at all.”

Age at Wave 1 was also included in the mod- els as a control variable, as women who had teen births were slightly younger at Wave 1 than those who had adult births and age may also influence attitudes toward teen pregnancy and depressive symptoms. Finally, two indicators of relationship status at first birth were constructed from responses on the Wave 4 pregnancy histo- ries. For each pregnancy they reported, respon- dents were asked whether they were married to the child’s father at the time of the pregnancy or birth. Unmarried respondents were asked whether they were cohabiting with the child’s father at the time of the pregnancy or birth. These two dichotomous variables were included in all of the models to control for marital and cohabi- tation status at first birth.

I chose not to include any additional Wave 4 controls in this analysis. I could have included controls for Wave 4 measures such as marital status, marital history, and socioeconomic sta- tus; however, both research and theory suggest that these are mechanisms through which teen birth affects adult depression. Women who have teen births often have more unstable marital tra- jectories and lower socioeconomic status attain- ment than women who delay childbearing until adulthood (Furstenberg, 2007; Taylor, 2009), and both marital status and socioeconomic sta- tus are known to affect mental health (Turner, Wheaton, & Lloyd, 1995). Thus, including Wave 4 characteristics in the models would underes- timate the true gross association of adolescent childbearing with depression. For example, in a supplementary analysis (not shown), including Wave 4 educational attainment as a control in the final model reduced the coefficient for the effect of adolescent pregnancy attitudes on Wave 4 depression among women who had teen births by approximately 5%.

Analytic Strategy

There is much debate among researchers and statisticians about the most appropriate way to measure the causal effect of teen childbearing. I found it useful to conceptualize having a teen birth as a transition: At Wave 1, no members of the analytic sample were teen parents; at Wave 4, some had transitioned to being teen parents and others had not (they had become adult parents instead). According to Johnson (2005), there

are two primary approaches to modeling the effects of transitions using two-wave panel data. First, there is what he calls a “lagged depen- dent variable” (LDV) model, an ordinary least squares (OLS) regression model with a control for a Time 1 measure of the dependent variable. Second, there is what he calls a “change score” (CS) model, which can be estimated using a time-series regression model with individual fixed effects. Johnson (2005) points out that in some cases, the two modeling strategies produce contradictory answers to the same questions. He recommends that family researchers studying transitions adopt the change score (individual fixed effects) approach, but also acknowledges that both approaches have advantages and disad- vantages. Problematically for the current study, the time-series regression model with individual fixed effects does not allow for the inclusion of time-invariant control variables. According to Johnson (2005),

If the researcher proceeds to use the LDV approach estimated with OLS regression, however, then it would be strongly recommended that they also fit a CS model to the same data. If the findings are consistent, then a methodological artifact because of measurement error or omitted variables is not likely to be biasing the results (p. 1074).

For this article, it was useful to estimate OLS regression models to include time-invariant control variables. Following Johnson’s (2005) advice, individual fixed effects time-series regression models were also estimated, and results from the two types of models were compared.

The first set of models employed OLS regression with a lagged dependent variable to investigate the relationship between Wave 1 attitudes toward teen pregnancy and Wave 4 depressive symptoms among women who had teen and adult first births between Waves 1 and 4. In Model 1, Wave 4 depression was regressed on Wave 1 depression and the teen first birth dummy variable. In Model 2, all control variables were added to Model 1. Finally, in Model 3, the interaction term between the teen birth dummy variable and adolescent pregnancy attitudes was introduced. Survey estimation with sampling weights in STATA 13 (StataCorp, 2013) was used to generate descriptive statistics for all variables and to estimate the OLS models.

The second set of models employed time-series regression with individual fixed

Teen Childbearing and Depression 397

effects to predict change in depressive symp- toms from change in teen parent status and adolescent pregnancy attitudes. The individ- ual fixed effects models controlled for all exogenous, time-invariant characteristics of individuals, so even though there was only one time-varying control (age), the models still controlled for many things about the individual respondents. It is akin to adding a dummy vari- able for each individual respondent to an OLS model. In Model 1, change in depressive symp- toms from Wave 1 to Wave 4 was regressed on age and the teen first birth dummy variable. In Model 2, the interaction term between the teen birth dummy variable and adolescent pregnancy attitudes was added to Model 1. Although it is not possible to include time-invariant variables in individual fixed effects regression mod- els, they can be included if interacted with a time-varying variable (Johnson, 2005). Thus it was appropriate to include the interaction term of the teen birth dummy variable and adolescent pregnancy attitudes even though adolescent pregnancy attitudes were only assessed at one time point. I used the “xtreg” command in STATA 13 (StataCorp, 2013) to estimate the weighted fixed effects models with standard errors adjusted for the clustered sample design.

Results

Table 1 presents the weighted descriptive statis- tics for all of the variables in this analysis for the full sample (N = 2,898) and separately for the subsamples of women who had teen (n = 592) and adult (n = 2,306) first births. An exami- nation of these descriptive statistics revealed some mean differences between the women who had teen births and the women who had adult births. Although average depressive symptoms were lower at Wave 4 than at Wave 1 for both groups, the women who had teen first births were more depressed at both Waves 1 and 4 when compared with the women who had adult first births. This finding was consistent with the results of Mollborn and Morningstar (2009). Women who had teen first births also had sig- nificantly more positive attitudes toward teenage pregnancy at Wave 1. On the pregnancy attitude measure, which ranged from 0 to 4 (higher val- ues indicated more positive attitudes toward ado- lescent pregnancy), the mean for women who had teen first births was 1.05 (standard devi- ation = 1.11), and the mean for women who

had adult first births was 0.80 (standard devi- ation = 0.98). Women who had teen first births were less likely to be married and more likely to be cohabiting at the time of the first birth when compared with women who had adult first births. A higher proportion of the women who had teen first births were Black and a lower pro- portion were White when compared with the women who had adult first births. The women who had teen first births were also more likely to be from families who were receiving pub- lic aid at Wave 1, and their parents generally had less education. When compared with women who had adult first births, a smaller proportion of the women who had teen first births were from two biological or adoptive parent families, and the proportions from the other family structures were larger. The women who had teen first births had a slightly lower mean age at Wave 1 than the women who had adult first births (15.91 years vs. 16.61 years).

Table 2 presents the results of a series of OLS models that investigated the relationship between teen childbirth, adolescent attitudes toward becoming pregnant, and adult depressive symptoms among women who had first births between Waves 1 and 4. In Model 1, Wave 4 depressive symptoms were regressed on the teen birth dummy variable and Wave 1 depressive symptoms. Although the bivariate analysis indicated more Wave 4 depressive symptoms among women who had a teen first birth when compared with an adult first birth, this asso- ciation became nonsignificant when a control for prebirth depressive symptoms was included. This is consistent with the results of Mollborn and Morningstar (2009).

In Model 2, adolescent pregnancy attitudes and all other controls were added to Model 1. This resulted in a further reduction in size for the teen birth coefficient, which remained nonsignif- icant. Wave 1 depression was a strong predictor of Wave 4 depressive symptoms (B = 0.26, stan- dard error [SE] = 0.03, p < .001), as was mem- bership in the “other” race and ethnicity category and residing in a stepfamily at Wave 1.

In Model 3, an interaction term between the teen birth and pregnancy attitudes variables was added to Model 2. The coefficient for the teen birth dummy variable remained nonsignificant. The coefficient for adolescent attitude toward becoming pregnant was significant (B = 0.34, SE = 0.13, p < .05). This means that on average, among women who had first births at age 20

398 Journal of Marriage and Family

Table 1. Weighted Descriptive Statistics

Full sample Teen moms Adult moms Mean/% Mean/% Mean/%

(SD) (SD) (SD)

Wave 4 depressive symptoms (0–30) 6.40 7.15** 6.21 (4.76) (5.10) (4.66)

Adolescent attitude toward becoming pregnant (0 = most negative, 4 = most positive) 0.85 1.05*** 0.80 (1.01) (1.11) (0.98)

Proportion teenager at first birth (0 = 20+ at first birth) 19.51% Wave 1 depressive symptoms (0–30) 7.85 8.82*** 7.62

(5.20) (5.37) (5.13) Race/ethnicity

White 66.45% 56.94%*** 68.75% Black 16.79% 23.50%*** 15.17% Hispanic 11.40% 14.92% 10.54% Other 5.37% 4.64% 5.54%

Family receipt of public aid No public aid 65.03% 57.82%** 66.78% Public aid 19.07% 24.54%* 17.74% Unknown 15.91% 17.64% 15.49%

Education of resident parent Less than high school 14.72% 20.27%*** 13.38% High school 32.86% 37.83% 31.65% Some college 22.17% 19.15% 22.90% Four-year college or higher 24.68% 16.28%*** 26.72% Unknown 5.57% 6.47% 5.35%

Family structure Two biological or adoptive parents 49.81% 35.35%*** 53.31% Stepfamily 18.40% 23.71%** 17.11% Single parent 24.91% 30.86%* 23.47% Other 6.88% 10.07%* 6.10%

Religiosity (1 = not important at all, 4 = very important) 3.07 3.09 3.07 (1.06) (1.06) (1.05)

Age 16.47 15.91*** 16.61 (1.14) (0.96) (1.14)

Proportion married at first birth 49.66% 21.23%*** 56.55% Proportion cohabiting at first birth 24.41% 29.40%* 23.20% n (unweighted) 2,898 592 2,306

Note. Asterisks indicate statistically significant differences in variable means between the two subgroups. *p < .05. **p < .01. ***p < .001.

years or older, each one unit increase in (the positivity of) adolescent pregnancy attitudes was associated with an increase of 0.34 depres- sive symptoms, holding control variables con- stant. The interaction term was also significant (B = −0.68, SE = 0.31, p < .05). Thus, each one unit increase in (the positivity of) pregnancy atti- tudes reduced the difference between women with teen births and women with adult births in Wave 4 depression by 0.68 symptoms, on aver- age and holding control variables constant.

Table 3 presents the results of the individual fixed effects transition models. Model 1 showed that there was no significant difference between women who had teen and adult first births in change in depressive symptoms from Wave 1 to Wave 4 after controlling for age and unob- served heterogeneity. In Model 2, the coeffi- cient for the interaction term between teen birth status and adolescent pregnancy attitude was significant. This indicated that any change in depressive symptoms associated with having a

Teen Childbearing and Depression 399

Table 2. Weighted Ordinary Least Squares Regressions of Adult Depressive Symptoms on Adolescent Attitudes Toward Becoming Pregnant and Controls

Model 1 Model 2 Model 3

B SE B B SE B B SE B

Wave 1 depressive symptoms 0.28*** 0.03 0.26*** 0.03 0.26*** 0.03 Teenager at first birth (ref. = 20+ at first birth) 0.61 0.31 0.25 0.37 0.92 0.47 Adolescent attitude toward becoming pregnant (0–4) 0.18 0.12 0.34* 0.13 Pregnancy Attitude × Teenager at Birth −0.68* 0.31 Race/ethnicity (ref. = White)

Black 0.50 0.38 0.45 0.37 Hispanic 0.08 0.33 0.08 0.32 Other 0.98** 0.36 0.96** 0.36

Wave 1 family receipt of public aid (ref. = no aid) Public aid 0.59 0.38 0.57 0.38 Unknown −0.36 0.32 −0.33 0.32

Wave 1 parental education (ref. = high school) Less than high school −0.07 0.36 −0.09 0.36 Some college −0.20 0.33 −0.20 0.33 Four-year college or higher −0.46 0.35 −0.41 0.35 Unknown −0.08 0.66 −0.18 0.66

Wave 1 family structure (ref. = two parents) Stepfamily 0.68* 0.30 0.68* 0.30 Single parent −0.02 0.29 −0.01 0.29 Other 0.55 0.62 0.61 0.63

Wave 1 religiosity (1–4) −0.03 0.11 −0.03 0.11 Wave 1 age 0.06 0.11 0.06 0.11 Married at first birth (ref. = unmarried) −0.57 0.32 −0.59 0.32 Cohabiting at birth (ref. = not cohabiting) −0.03 0.38 −0.04 0.38 Constant 4.08*** 0.22 3.33 1.91 3.18 1.91 n 2,898 2,898 2,898 R2 0.10 0.12 0.12

Note. ref. = reference. *p < .05. **p < .01. ***p < .001.

teen first birth depended on attitudes toward pregnancy prior to the birth. These findings were entirely consistent with the findings from the OLS regression analysis. The interaction term coefficient of −0.98 was similar to the coef- ficient from Model 3 in Table 2, which was −0.68. The fact that the OLS and individual fixed effects models produced such similar findings suggested that the results were not biased (John- son, 2005).

Figure 1 provides a graphical representation of the interaction term from Model 2. This graph shows the predicted depressive symptoms for women based on the interaction between their adolescent attitudes toward teen pregnancy and whether they had their first births as teenagers, holding age at the Wave 4 mean (29.3 years old). Pairwise comparisons of predicted depressive

symptoms for teen and adult child bearers revealed the values of pregnancy attitudes at which teen mothers significantly differed from adult mothers. As shown in Figure 1, teen moth- ers with the most negative adolescent attitudes toward teen pregnancy (strongly disagree) were more depressed than women who had their first births at age 20 or older, although this difference was not significant. On the other hand, teen mothers who had more positive attitudes toward teen pregnancy were significantly less depressed than women who had their first births as adults.

Discussion

This study has two main findings. First, the results showed that teen childbearing was not associated with more depressive symptoms in

400 Journal of Marriage and Family

Table 3. Weighted Fixed Effects Models Predicting Adult Depressive Symptoms

Model 1 Model 2

B SE B B SE B

Teenager at first birth (ref. = 20+ at first birth) −0.26 0.35 0.77 0.49 Pregnancy Attitude (0–4) × Teenager at Birth −0.98** 0.34 Age −0.11*** 0.02 −0.11*** 0.02 Constant 9.66*** 0.35 9.66*** 0.35 Sigma_u 4.02 4.04 Sigma_e 4.14 4.13 Rho 0.49 0.49 Number of observations 5,796 5,796 Number of groups 2,898 2,898

Note. The standard errors were adjusted for the clustered sample design. ref. = reference. **p < .01. ***p < .001.

Figure 1. Predicted Values for Wave 4 Depressive Symptoms among Women Who Had Teen and Adult First Births, By Adolescent Attitude toward Becoming Pregnant, Holding Age at the Wave 4 Mean (Individual

Fixed Effects Estimates).

0

1

2

3

4

5

6

7

8

Strongly disagree (0)

Disagree (1) Neither agree nor disagree

(2)*

Agree (3)** Strongly agree (4)**

P re

d ic

te d

W av

e 4

C E

S -D

S co

re (

0- 30

)

Wave 1 agreement with "it wouldn't be all that bad if you got pregnant at this time in your life"

Teen first birth

Adult first birth

Note. Asterisks indicate significant differences between women with teen and adult first births. CES-D = Center for Epidemiologic Studies Depression Scale. *p < .05. **p < .01.

adulthood when compared with adult child- bearing, after accounting for adolescent depres- sive symptoms and background characteristics. This result contradicts both conventional wis- dom and some previous empirical evidence that teen childbearing is detrimental to mental health (e.g., Mirowsky & Ross, 2002). It fits well, however, within an emerging body of scholar- ship suggesting that any negative association between teen childbearing and mental health can be explained by the selection of depressed

or disadvantaged women into teen childbearing (e.g., Mollborn & Morningstar, 2009; Taylor, 2009).

Second, the results support the hypothesis that the association of teen childbearing with more depressive symptoms when compared with adult childbearing would be weaker among those with more positive adolescent attitudes toward becoming pregnant. In fact, the results showed that among women who had more posi- tive adolescent pregnancy attitudes, women who

Teen Childbearing and Depression 401

had teen first births were actually less depressed in adulthood than women who had their first births as adults. This pattern is consistent with the idea that life course expectations profoundly influence mental health adjustment to life events (Carlson, 2011; Carlson & Williams, 2011; Mossakowski, 2011) as well as with the predictions from stress process theory and self-discrepancy theory that cognitive appraisals of events and role occupancies matter for men- tal health adjustment (Higgins, 1987; Pearlin, 1999).

This article has posited that adolescent pregnancy attitudes shape depression after teen childbirth through primarily psycholog- ical mechanisms. It is important to note that, although the study was motivated by these theories, the positive findings cannot identify which mechanism produces the association. Adolescents’ responses to the pregnancy atti- tude question could simply reflect their accurate observations about their social and material resources and ability to care for a child. In this case, the observed differences in depressive symptoms according to adolescent pregnancy attitudes would be caused by expected dif- ferences in the ability to marshal social and economic resources to care for a child. Those who did not have enough resources would encounter problems, which would be damaging to mental health. Although this is a reason- able alternative explanation, I am inclined to place more emphasis on the psychological mechanisms suggested by stress process and self-discrepancy theories. The fact that the coef- ficient for the interaction term between teen birth status and pregnancy attitudes was statistically significant after accounting for sociodemo- graphic control variables suggests that the pregnancy attitude question was not simply a measure of whether the adolescent had suffi- cient resources to care for a child. Nonetheless, future research should specifically investigate the mechanisms through which adolescent pregnancy attitudes moderate the relationship between teen childbearing and adult depression. For example, the mechanisms implied here (e.g., stressors, discrepancy between actual and ideal selves) could be explicitly measured and their explanatory power compared to that of more external mechanisms such as educational attainment or income.

There are some limitations to this study. First, because the adolescents were in Grades 7 to

12 at Wave 1, some of the sample participants were already approaching the end of their teen years when the pregnancy attitudes measure was administered. This means that some women who had teen births had to be excluded from the sample because they had their first births before Wave 1. Second, the Add Health survey only asked the pregnancy attitudes question of ado- lescents who were at least 15 years old. Thus, this study does not address whether the conse- quences for adult depression of teen childbear- ing vary according to pregnancy attitudes for the subset of adolescents who had very young births. Notably, however, very few pregnancies and teen births occur to girls who are younger than age 15 (Guttmacher Institute, 2014). Furthermore, it is not clear that valid answers would be obtained by asking girls younger than age 15 about their attitudes toward becoming pregnant.

Another limitation is the wording of the preg- nancy attitudes question. The question asked adolescents to rate their agreement with the fol- lowing statement: “It wouldn’t be all that bad if you got pregnant at this time in your life.” Note that it is not possible to give an answer indicating that becoming pregnant would be good; it is only possible to strongly agree that it would not be bad. Although the question word- ing may have biased responses toward more negative pregnancy attitudes, I do not believe it would have changed the relative positions of individual respondents on the scale of preg- nancy attitudes. In this case, the biased ques- tion wording should not threaten the validity of the results. Still, it will be important for future research to replicate these findings using other measures of adolescent pregnancy atti- tudes. Fortunately, researchers are beginning to develop more nuanced prospective measures of adolescent pregnancy attitudes (see Barber et al., 2015), making this a promising avenue for future research.

In conclusion, it is extremely important to understand what shapes the mental health out- comes of women who experience teen births. I do not wish to downplay the importance of educating teens about avoiding pregnancy, but I also hope to highlight the heterogeneity of women’s experiences with teen childbearing and point out that teen childbearing may not be detrimental (in terms of depression) to all women. In fact, what this article clearly shows is that individual desires, attitudes, and prefer- ences do matter for mental health adjustment to

402 Journal of Marriage and Family

major life events. In turn, adults who are less depressed are less likely to be unemployed and their children are less likely to have physical and mental health problems (Dooley, Prause, & Ham-Rowbottom, 2000; Meadows et al., 2007; Turney, 2011a, 2011b). Of course, there are many other important outcomes of teen child- bearing beyond maternal depression, including maternal educational attainment, employment, and parenting quality. Indeed, researchers of the relationship between teen childbearing and these other outcomes also debate the degree to which observed associations are a result of selection versus causation. Future research should investi- gate whether individual pregnancy attitudes and preferences moderate the associations of teen childbearing with a variety of other outcomes.

Note

I would like to thank Kristi Williams, Reanne Frank, John Casterline, Elizabeth Cooksey, Anthony Paik, Jennifer Lundquist, and Michelle Budig, as well as Janice Irvine and my classmates in her graduate writing class for reading drafts of this article and providing valuable feedback. I would also like to thank the Institute for Population Research at The Ohio State University for providing office space and funding to pursue this research. Previous versions of this article were presented at the Ohio State University/Bowling Green State University 2013 Annual Graduate Student Conference on Population and the Population Association of America 2014 Annual Meeting.

References

Avison, W. R., Ali, J., & Walters, D. (2007). Fam- ily structure, stress and psychological distress: A demonstration of the impact of differential expo- sure. Journal of Health and Social Behavior, 48, 301–317. doi:10.1177/002214650704800307

Barber, J. S., Yarger, J. E., & Gatny, H. H. (2015). Black-White differences in attitudes related to pregnancy among young women. Demography, 52, 751–786. doi:10.1007/s13524-015-0391-4

Barcelos, C. A., & Gubrium, A. C. (2014). Reproduc- ing stories: Strategic narratives of teen pregnancy and motherhood. Social Problems, 61, 466–481. doi:10.1525/sp.2014.12241

Carlson, D. L. (2011). Explaining the curvilinear rela- tionship between age at first birth and depression among women. Social Science & Medicine, 72, 494–503. doi:10.1016/j.socscimed.2010.12.001

Carlson, D. L. (2015). Do differences in expecta- tions and preferences explain racial/ethnic varia- tion in family formation outcomes? Advances in Life Course Research, 25, 1–15. doi:10.1016/j.alcr. 2015.05.002

Carlson, D. L., & Williams, K. (2011). Parent- hood, life course expectations, and mental health. Society and Mental Health, 1, 20–40. doi:10.1177/2156869310394541

Child Trends Databank. (2015a). Births to unmarried women. Retrieved from http://www.childtrends. org/?indicators=births-to-unmarried-women

Child Trends Databank. (2015b). Teen births. Retrieved from http://www.childtrends.org/? indicators=teen-births

Dooley, D., Prause, J., & Ham-Rowbottom, K. A. (2000). Underemployment and depression: Longi- tudinal relationships. Journal of Health and Social Behavior, 41, 421–436. doi:10.2307/2676295

Edin, K., & Kefalas, M. (2005). Promises I can keep: Why poor women put motherhood before marriage. Berkeley, CA: University of California Press.

Elder, G. H., Jr. (1975). Age differentiation and the life course. Annual Review of Sociology, 1, 165–190. doi:10.1146/annurev.so.01.080175. 001121

Evenson, R. J., & Simon, R. W. (2005). Clarifying the relationship between parenthood and depres- sion. Journal of Health and Social Behavior, 46, 341–358. doi:10.1177/002214650504600403

Feather, N. T., & Newton, J. W. (1982). Values, expec- tations, and the prediction of social action: An expectancy-valence analysis. Motivation and Emo- tion, 6, 217–244. doi:10.1007/bf00992246

Fletcher, J. M., & Wolfe, B. L. (2009). Educa- tion and labor market consequences of teenage childbearing: Evidence using the timing of preg- nancy outcomes and community fixed effects. Journal of Human Resources, 44, 303–325. doi:10.1353/jhr.2009.0026

Furstenberg, F. F. (2003). Teenage childbearing as a public issue and private concern. Annual Review of Sociology, 29, 23–39. doi:10.1146/annurev. soc.29.010202.100205

Furstenberg, F. F. (2007). Destinies of the disadvan- taged: The politics of teen childbearing. New York: Russell Sage.

Geronimus, A. T. (2003). Damned if you do: Culture, identity, privilege, and teenage childbearing in the United States. Social Science & Medicine, 57, 881–893. doi:10.1016/s0277-9536(02)00456-2

Guttmacher Institute. (2014). American teens’ sexual and reproductive health. Retrieved from https:// www.guttmacher.org/pubs/FB-ATSRH.html

Harris, K. M. (2009). The National Longitudinal Study of Adolescent Health (Add Health), Waves I & II, 1994–1996; Wave III, 2001–2002; Wave IV, 2007–2009 [Machine-readable data file and doc- umentation]. Chapel Hill, NC: Carolina Popula- tion Center, University of North Carolina at Chapel Hill.

Hayford, S. R., & Morgan, S. P. (2008). Religios- ity and fertility in the United States: The role of

Teen Childbearing and Depression 403

fertility intentions. Social Forces, 86, 1163–1188. doi:10.1353/sof.0.0000

Henretta, J. C. (2007). Early childbearing, marital status, and women’s health and mortality after age 50. Journal of Health and Social Behavior, 48, 254–266. doi:10.1177/002214650704800304

Higgins, T. E. (1987). Self discrepancy: A theory relating self and affect. Psychological Review, 94, 319–340. doi:10.1037/0033-295x.94.3.319

Hoffman, S. D., Foster, E. M., & Furstenberg, F. F. (1993). Reevaluating the costs of teenage child- bearing. Demography, 30, 1–13. doi:10.2307/ 2061859

Jaccard, J., Dodge, T., & Dittus, P. (2003). Do adoles- cents want to avoid pregnancy? Attitudes toward pregnancy as predictors of pregnancy. Journal of Adolescent Health, 33, 79–83. doi:10.1016/s1054- 139x(03)00134-4

Johnson, D. (2005). Two-wave panel analysis: Comparing statistical methods for studying the effects of transitions. Journal of Marriage and Family, 67, 1061–1075. doi:10.1111/j.1741-3737. 2005.00194.x

Kalil, A., & Kunz, J. (2002). Teenage childbear- ing, marital status, and depressive symptoms in later life. Child Development, 73, 1748–1760. doi:10.1111/1467-8624.00503

Kaplan, E. B. (1997). Not our kind of girl: Unraveling the myths of Black teenage motherhood. Berkeley and Los Angeles, CA: University of California Press.

Kearney, M. S., & Levine, P. B. (2012). Why is the teen birth rate in the United States so high and why does it matter? Journal of Economic Perspectives, 26, 141–166. doi:10.1257/jep.26.2.141

Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer.

Luker, K. (1996). Dubious conceptions: The politics of teenage pregnancy. Cambridge, MA: Harvard University Press.

Martinez-Torteya, C., Bogat, G. A., von Eye, A., Levendosky, A. A., & Davidson, W. S., II. (2009). Women’s appraisals of intimate partner violence stressfulness and their relationship to depressive and posttraumatic stress disorder symptoms. Violence and Victims, 24, 707–722. doi:10.1891/ 0886-6708.24.6.707

Meadows, S. O., McLanahan, S. S., & Brooks-Gunn, J. (2007). Parental depression and anxiety and early childhood behavior problems across fam- ily types. Journal of Marriage and Family, 69, 1162–1177. doi:10.1111/j.1741-3737.2007. 00439.x

Mirowsky, J., & Ross, C. E. (2002). Depression, parenthood, and age at first birth. Social Science & Medicine, 54, 1281–1298. doi:10.1016/s0277- 9536(01)00096-x

Mollborn, S., & Morningstar, E. (2009). Investigat- ing the relationship between teenage childbearing

and psychological distress using longitudinal evi- dence. Journal of Health and Social Behavior, 50, 310–326. doi:10.1177/002214650905000305

Mossakowski, K. N. (2011). Unfulfilled expecta- tions and symptoms of depression among young adults. Social Science & Medicine, 73, 729–736. doi:10.1016/j.socscimed.2011.06.021

Nathanson, C. A. (1991). Dangerous passage: The social control of sexuality in women’s adolescence. Philadelphia, PA: Temple University Press.

Pearlin, L. I. (1989). The sociological study of stress. Journal of Health and Social Behavior, 30, 241–256. doi:10.2307/2136956

Pearlin, L. I. (1999). Stress and mental health: A conceptual overview. In A. V. Horwitz & T. L. Scheid (Eds.), A handbook for the study of mental health (pp. 161–175). Cambridge, UK: Cambridge University Press.

Pearlin, L. I., Schieman, S., Fazio, E., & Meers- man, S. (2005). Stress, health, and the life course: Some conceptual perspectives. Journal of Health and Social Behavior, 46, 205–219. doi:10.1177/ 002214650504600206

Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general pop- ulation. Applied Psychological Measurement, 1, 385–401. doi:10.1177/014662167700100306

SmithBattle, L. (1995). Teenage mothers’ narra- tives of self: An examination of risking the future. Advances in Nursing Science, 17, 22–36. doi:10.1097/00012272-199506000-00004

SmithBattle, L. (2007). “I wanna have a good future”: Teen mothers’ rise in educational aspirations, com- peting demands, and limited school support. Youth and Society, 38, 348–371. doi:10.1177/0044118 x06287962

StataCorp. (2013). Stata Statistical Software: Release 13 [Computer software]. College Station, TX: StataCorp LP.

Taylor, J., & Turner, R. J. (2001). A longitudinal study of the role and significance of mattering to others for depressive symptoms. Journal of Health and Social Behavior, 42, 310–325. doi:10.2307/ 3090217

Taylor, J. L. (2009). Midlife impacts of adolescent parenthood. Journal of Family Issues, 30, 484–510. doi:10.1177/0192513x08329601

Thoits, P. A. (2011). Mechanisms linking social ties and support to physical and mental health. Jour- nal of Health and Social Behavior, 52, 145–161. doi:10.1177/0022146510395592

Trent, K., & Crowder, K. (1997). Adolescent birth intentions, social disadvantage, and behavioral out- comes. Journal of Marriage and the Family, 59, 523–535. doi:10.2307/353943

Turner, R. J., Wheaton, B., & Lloyd, D. A. (1995). The epidemiology of social stress. American Sociolog- ical Review, 60, 104–125. doi:10.2307/2096348

404 Journal of Marriage and Family

Turney, K. (2011a). Chronic and proximate depres- sion among mothers: Implications for child well-being. Journal of Marriage and Family, 73, 149–163. doi:10.1111/j.1741-3737.2010.00795.x

Turney, K. (2011b). Maternal depression and child- hood health inequalities. Journal of Health and Social Behavior, 52, 314–332. doi:10.1177/002 2146511408096

Umberson, D., & Montez, J. K. (2010). Social rela- tionships and health: A flashpoint for health pol- icy. Journal of Health and Social Behavior, 51, S54–S66. doi:10.1177/0022146510383501

Wickrama, K. A. S., Conger, R. D., Lorenz, F. O., & Jung, T. (2008). Family antecedents and

consequences of trajectories of depressive symp- toms from adolescence to young adulthood: A life course investigation. Journal of Health and Social Behavior, 49, 468–483. doi:10.1177/002 214650804900407

Wigfield, A., & Eccles, J. S. (2000). Expectancy-value theory of achievement motivation. Contem- porary Educational Psychology, 25, 68–81. doi:10.1006/ceps.1999.1015

Williams, K., Sassler, S., Frech, A., Addo, F., & Cook- sey, E. (2011). Nonmarital childbearing, union his- tory, and women’s health at midlife. American Sociological Review, 76, 465–486. doi:10.1177/ 0003122411409705

Copyright of Journal of Marriage & Family is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.