Associations and Intervening Mechanisms Between Family Structure and Young Children’s Obesity

Abstract

This study examined the links between different family structures—capturing type and stability thereof—and preschool-aged children’s likelihood of being obese. We build on the limited number of studies that have pursued this topic by using a large, nationally representative sample of preschool-aged children from the Early Childhood Longitudinal Survey–Birth Cohort (n = 8,350) and exploring a wide range of mechanisms to explain these links. Results revealed that, compared with young children with stably married parents, children in cohabiting- and single-parent families that experienced a prior family structure change were more likely to be obese, except for children in single-parent families born to married parents. Children in step, stably single, and stably cohabiting families were at no greater risk of obesity. These patterns were largely driven by female children, for whom the effects of family structure were most robust. None of the 11 tested mechanisms explained such patterns.

Keywords

obesity family structure preschool children instability health

Historical changes in marital patterns and fertility behaviors have led to well-documented changes in the structure of U.S. families. These changes can be characterized a number of ways, although the most notable changes have been an increase in cohabiting-parent families, either with a stepfather or biological father, mother-only families, and married stepparent families, and a greater likelihood that children will experience instability in their family structures (Bzostek, McLanahan, & Carlson, 2012; Raley & Wildsmith, 2004). Following these trends, scholars have amassed a large literature examining the associations between different family structure states (i.e., married, single, cohabiting, step) and their stability for children’s academic and behavioral development. The basic conclusions are that children raised in mother-only families, followed by children with cohabiting parents, then stepparents, experience modest but consistent academic and behavioral disadvantages compared with children raised by married biological parents, which accumulate with each change in parents’ union status (Amato, 2005; Carlson & Corcoran, 2001).

One important question that has been overlooked within this family structure literature is whether there are implications of family structure for children’s weight status. We know of just a handful of studies on the topic, and even fewer which have pursued it in a way that captures the complexity of family structure today to reflect both new family forms (i.e., cohabitation) and the stability of these unions. This limitation is surprising given that increasing family complexity has mirrored the trends of increasing obesity among young children. Thus far, the research suggests that family structure matters for children’s weight status, but in ways that may differ from how family structure is associated with children’s academic and behavioral development. As one example, Schmeer (2012) finds that children with stably single parents were at higher risk of being overweight, but children in stably cohabiting families were not (also see Augustine & Kimbro, 2013; Bzostek & Beck, 2011).

While such work represents a key first step in assessing whether family structure matters for children’s weight development, more work must be done to asses their robustness to other methods, measures, and samples. Researchers must also take steps to clarify which dimensions of family structure matter most to children’s risk of obesity, differentiating among the relative salience of state, stabiliy, and family relatedness (i.e., biological, stepfather; Cragie, Brooks-Gunn, & Waldfogel, 2015; Magnuson & Berger, 2009), and why family structure matters for children’s weight. Our aim in this study is to contribute to each of these efforts by drawing on data from a large (n = 8,350), nationally representative sample of children participating in the Early Childhood Longitudinal Survey–Birth Cohort (ECLS-B).

Background

In this study, we turn our attention from whether and why family structure is linked to children’s academic and behavioral development to its significance for children’s weight status, and in particular, their risk of obesity. We focus on obesity over other dimensions of weight (i.e., overweight, body mass index [BMI]) because it carries greater risks to children’s health and well-being in both the short- and long-term. In the short-term, childhood obesity is associated with greater behavioral, social, and academic difficulties (Strauss, 2000). In the long-term, it is linked to overweight in adulthood and numerous morbidities (Dietz, 1998; Guo & Chumlea, 1999). Despite such risks, few studies on family structure have examined whether and why family structure is connected to young children’s risk of obesity (indeed, see Carr & Springer, 2010). This question has also been largely overlooked by health scholars, who have focused more on how other social factors, such as parents’ socioeconomic status, neighborhood context, and immigrant status contribute to children’s risk of obesity (Balistreri & Van Hook, 2009; Kimbro, Brooks-Gunn, & McLanahan, 2011; Ogden, Carroll, Curtin, Lamb, & Flegal, 2010).

Importantly, parents are the primary gatekeepers of young children’s health. They are the main source of the instrumental, material, and emotional resources which affect the development of children’s health and they control and monitor children’s health-related behaviors (Case & Paxson, 2002; Waldfogel, Craigie, & Brooks-Gunn, 2010). Given how parents’ ability to provide such supports is constrained and shaped by their family structures, it would follow that family structure has important implications for the development of children’s weight. Basic support for this conceptual idea comes from research connecting family structure to other domains of child health, including birth outcomes (e.g., low birth weight, preterm), mental health, general health, injury, and asthma (Bzostek & Beck, 2011; Kiernan & Pickett, 2006, Kimbro, 2008; Luo, Wilkins, & Kramer, 2004; Mathews & MacDorman, 2012; Schmeer, 2011; Wen, 2008). There is also preliminary support for an association from research linking family structure to young children’s obesity, although this literature has several limitations which we aim to build on.

Review of Research on Family Structure and Young Children’s Obesity

The small handful of studies linking family structure to young children’s obesity (or weight status in general) finds that children in two-parent or married parent families are less likely to be obese (or overweight) than children with single or unmarried parents (Anderson & Whitaker, 2010; A. Y. Chen & Escarce, 2010; Gable & Lutz, 2000; Gibson et al., 2007; Hesketh, Crawford, Salmon, Jackson, & Campbell, 2007; Huffman, Kanikireddy, & Patel, 2010; Kitsantas & Gaffney, 2010). The conclusions from these studies remain limited, however, as they fail to separate the effect of marital status from the presence of two parents or to distinguish among two-parent homes with a biological versus stepparent, do not consider whether the family structure was the same since the child’s birth, and do not control for many factors that may select women into different family structures, like the mother’s own BMI (The & Gordon-Larson, 2011).

We know of only three studies that take a more complex view of young children’s family structure and account for key sources of selection. Two of the studies (Bzostek & Beck, 2011; Schmeer, 2012) find that children in stably cohabiting and stably married families have similar odds of obesity or weight gain, while children in stably single families or in families that had experienced a family structure change are at increased risks, except for children born to married parents (Bzostek & Beck, 2011) or whose parents marry (Schmeer, 2012). The other (Augustine & Kimbro, 2013) finds fewer family structure differences in children’s obesity, except between stably married families and children in cohabiting homes with the biological father. Still, these studies are limited in other ways.

Only one uses nationally representative data (Augustine & Kimbro, [2013], the ECLS-B). The other two draw on a disadvantaged sample living in urban cities. Thus, the generalizability of the studies that report pronounced family structure differences to the U.S. population remains unknown. At the same time, Augustine and Kimbro (2013) measure family structure using 11 discrete categories which combine the state, relatedness, and stability—an approach which results in low statistical power due to small cell sizes among many categories and may explain why they found few family structure differences in children’s obesity. It also makes it difficult to discern if one dimension of family structure mattered more to young children’s obesity than the others. The other two studies share this latter limitation, as they did not capture children’s complete family structure histories since birth or disaggregate their current family structures in key ways (e.g., considering whether the coresident partner is a stepfather or biological father).

Thus, beyond pursuing an important and understudied research question, one innovation of this study is that we capture a more diverse and complete range of children’s family structure experiences up through preschool which we defined by the state (married, cohabiting, single), the family relatedness (stepfather, biological father), and stability since the child’s birth, taking care to assess the importance of each dimension to children’s obesity risk. We also acknowledge that even more complex conceptualizations could be used to allow for relations among siblings and extended kin, but limit the scope of the study to parent–child relationships.

Conceptual Model and Hypotheses

Our expectation of how child obesity might vary across different combinations of family structure, stability, and relatedness comes from the larger literature on family structure and child well-being. The findings from this literature are well-known among family scholars and have been summarized elsewhere (Brown, 2010; Crosnoe & Cavanagh, 2010; Waldfogel et al., 2010). In brief, married families typically have more resources beyond money; they also have more time (single mothers must take on all domestic duties; there is often less time investment by cohabiting partners) and emotional support. These social resources, in turn, lead to less conflict and maternal depression and better parenting. At the same time, the biological status of the partner often moderates these patterns. For example, time investment is less among partners, cohabiting or married, who are not the child’s biological father, and there is also less cooperation among the adult partners. Instability plays a role by disrupting continuities in household routines, although these changes may be more stressful for children transitioning to single or cohabiting families than to married ones, in which the aforementioned resources can buffer against the risks.

These general patterns suggest the following. Children in single and cohabiting families will experience greater risk of obesity than children in married parent families. These risks will be magnified in cohabiting stepfamilies or single-parent families in which the child experienced prior instability. They will also be comparatively greater than for children who had transitioned into a married stepparent family, although potentially less so for children in single and cohabiting families. At the same time, because researchers have yet to settle the matter of whether stability trumps state, or vice versa, we also acknowledge the possibility that children in stably single and stably cohabiting families may have similar rates of obesity to children in stably married families (see Cavanagh & Huston, 2006; Kamp Dush, 2009, for evidence of such).

Mechanisms

As the second step in analysis, we explore a wide range of child, mother, and household-level factors that may explain the expected associations. Our selection of such mechanisms is derived from two theoretical traditions, ecological and stress theory, and empirical research on young children’s family structure experiences and their families’ health-related behaviors. The connections between the mechanisms and different family structure configurations are informed by the ideas summarized above.

First, the ecological framework emphasizes the interactions between children and their home environment and thus, how organizational aspects of the environment shaped by family structure affect children’s risk of obesity (Bronfenbrenner, 1994). Broadly, these organizational features include household routines and time use. Operationally, they may be reflected in shared meals, children’s nutrition and physical activity, television watching, and bedtime, all of which have also been linked to child obesity. For example, families headed by single parents must often balance work, child care, meals, and housework, and thus, face serious time constraints. As such, their children are likely to spend more time engaged in sedentary behaviors like television watching, have fewer shared family meals, and eat less nutritious diets as the parent relies more on convenience foods (Anderson & Whitaker, 2010; Bowman & Harris, 2003; Kimbro et al., 2011). In families with stepparents or cohabiting parents in which only one parent is biologically related to the child, healthy parenting may be a challenge if the coresident partner spends some portion of his time in another household, is not in agreement with the mother on all parenting issues, or is less familiar with the household’s routines or rules. Parents in single and complex families may also be less efficacious about enforcing family routines, diet, television, and appropriate bedtimes (Anderson & Whitaker, 2010; Sanders & Woolley, 2005)—factors liked to young children’s risk of obesity (Must & Parisi, 2009).

Second, stress theory points to how the stresses associated with certain family structure circumstances, such as a recent family structure change or living with a single or stepparent, makes it more difficult for mothers to consistently invest time and energy into preparing healthy meals, monitor children’s activities, enforce rules, and generally organize the home environment in ways that reduce children’s obesity risk as explained above (Hetherington, Bridges, & Insabella, 1998). It also points to how stress conditions individual adaptation and thus family structure might matter for child obesity via more individual versus household-level indicators, such as the mothers’ parenting stress and depression and child’s internalizing and externalizing problems (McCubbin & Patterson, 1983). For example, mothers experience greater stress and depression and their children have more internalizing and externalizing problems following a family structure change (Cavanagh & Huston, 2008; Cooper, Osborne, Beck, & McLanahan, 2011), as do both mothers and children in single and cohabiting families (Beck, Cooper, McLanahan, & Brooks-Gunn, 2010; Gerald, Anderson, Johnson, Hoff, & Trimm, 1994). Both maternal stress and child behavior, in turn, have been posited as predictors of child obesity through the linked pathways in the brain for appetite and emotional regulation (Anderson, He, Schoppe-Sullivan, & Must, 2010; McEwen, 2008). Stress may be less in married stepparent families, however, in which there may be more cooperation among partners and greater effort at minimizing risks to children’s health.

Summary of Study

In sum, in this study, we examine the associations between young children’s family structures and their risk of obesity, taking care to differentiate among the importance of state, stability, and relatedness. In doing so, we account for an array of confounds that reflect selection into different family types (e.g., background characteristics, low child birth weight, mother BMI) and may proxy for mothers’ behaviors (Currie, 2008; Schmeer, 2012), and endogenous factors (high child birth weight and mother prepregnancy BMI), that reflect children’s genetic tendencies toward obesity (Schmeer, 2012). We also use nationally representative data. Finally, we examine which mechanisms may explain the family structure–related patterns in young children’s likelihood of obesity that we observe.

Method

Data and Sample

The ECLS-B is a nationally representative sample of children born in 2001 based on a clustered list frame sample of around 14,000 births registered in the National Center for Health Statistics vital statistics system. A total of 10,700 children whose parents participated in the first wave of the in-home interview around child age 9 months were enrolled (74%). Data were subsequently collected when children were 2 years, around age 4, and in kindergarten. At each wave, in-home interviews were conducted with the primary caregiver (almost always the mother). During interviews, the primary caregiver provided information on a variety of issues, including parenting practices, routines, and stress, the family structure and children’s behavior. Children’s height and weight were also assessed at each wave by trained data collectors.

Our analytic sample includes the around 8,350 children that completed the preschool assessment and thus had the longitudinal sampling weight, W31C0, created by National Center for Education Statistics to adjust for differential nonresponse at the baseline, over time attrition, and the initial sampling design in which children from select subgroups (e.g., Asian, American Indian, low birth weight) were oversampled. Univariate and bivariate analyses employ this weight. The multivariate analyses use Stata’s svy options to incorporate the survey weights, while also adjusting for the complex survey design in which selection was stratified along various factors (e.g., family income) and clustered within primary sampling units. While retaining this entire sample, we use the subpop feature in Stata to bracket out children whose mother was not the primary caregiver (n = 150) or were missing information on their height or weight (n = 150). Using subpop versus deleting cases yields correct standard errors, which should be based on the entire population (Cochran, 1977).

Measures

Obesity

Children’s heights and weights were measured by trained data collectors during the preschool (Wave III) in-home data collection. Data collectors used a stadiometer to measure height and a digital bathroom scale to measure weight. Each height and weight measurement was recorded twice, and then averaged together. We used the Center for Disease Control Growth Charts (see http://www.cdc.gov/growthcharts/cdc_charts.htm) appropriate for the child’s gender and age to calculate children’s BMI based on this height-for-weight measurement. Children with BMIs at or above the 95th percentile were classified as obese. All other children, including over- and underweight children (about 3%), were considered nonobese and sorted into the reference group.

Family Structure

At the 9-month, 2-year, and preschool data collection, mothers reported their marital status, relationship to the child, and relationship to other resident family members. The child’s birth certificate provides information on whether the mother was married at the time of the child’s birth. Using this information, we develop several measures of family structure, which we later combine to capture more complex family structure histories. Family structure at the preschool wave, when we examine obesity, was based on three mutually exclusive categories: married, unpartnered single mother, and cohabiting. Whether the biological father resided in the home during preschool was based on a dichotomous indicator (1 = yes, 0 = no). Family structure at the first wave of data collection, 9 months, was also dummy coded as single, married, cohabiting. The birth certificate allows us assess whether the mother was married at the child’s birth (1 = yes, 0 = no). Combining reports from the three waves and birth certificate, we account for whether there was ever instability in family structure between assessment periods. Children whose family structure ever changed at any point were assigned a value “1.” If it never changed, instability was coded as “0.”

Note, while the information collected by the ECLS-B allows us to capture the vast majority of family structure changes, we could not capture a change if (a) the mother reported cohabiting with a social father at all three data collections and was unmarried at the child’s birth, as we cannot be certain the social father was the same person at each wave (n < 50) or (b) the mother was single at every assessment because she could have had a short-term coresidence with a partner or a marriage that started and ended between waves (n = 850). In such instances, we may understate instability. Also, these measures represent our focal family structure variables, but we also conduct tests of sensitivity to account for the number of family structure transitions (dummy coded 0, 1, or 2 or more), the timing of them (coded as birth to 9-month, 9-month and 2-year, 2-year and preschool), and reason (e.g., divorce, dissolution of cohabiting union).

Mechanisms

The first set of mechanisms reflects household organization and time use. A measure of rules is based on the sum of maternal responses to four questions about whether there are rules for bedtime, chores, television watching, and what the child eats (1 = yes, 0 = no; range: 0-4). Children’s screen time is based on mothers’ reports on the total number of hours per day that child watches television or movies, which we dummy coded, 0 = no screen time, 1 = 1 hour per day, 2 = 2 hours, 3 = 3 or more hours, because it was heavily right skewed. Shared meals is based on mother responses to the question: How many days per week does the family eat dinner together (range: 0-7). Late bedtime is based on mother reports for when her child usually goes to bed (10 p.m. or later was coded as late bedtime = 1; bedtime at or before 9 p.m. = 0). Diet is based on the average of maternal responses to a series of questions on how often per week the child eats fast food, salty snacks, sweets, or drinks soda or juice (0 = never, 7 = three times per day; range: 0-7). Family outdoor play is based on mother’s responses to the question: “How many times per week does the family play outside together” (range: 0-7). Finally, children’s time in organized activities is based on mother reports of whether the child participated in an organized physical activity like soccer or dance (1 = yes, 0 = no).

The second set of mechanisms considers the potential for mothers’ or children’s adaptive skills to form a link between family structure to children’s risk of obesity. Child internalizing and externalizing problems were based on mother responses to 24 questions derived from several instruments (i.e., Preschool Behavior Scales–Second edition, Social Skills Rating System, Family and Child Experiences Study) about how often their child displayed certain behaviors. We created an externalizing subscale based on seven items (e.g., temper tantrums, impulsivity) and internalizing subscale based on two indicators (how often the child appeared worried or unhappy), following Roisman and Fraley (2012). Higher values reflect more problems. Mother depression is based on mothers’ responses to 12 questions (e.g., degree to which she feels anxious, how often she feels unusually bothered), with responses ranging from 1 (rarely or never or 1 day per week) to 4 (most of all of the time or 5-7 days a week), which we summed and averaged (range: 1-4). We also capture mothers’ parenting stress based on her responses to a series of questions adapted from the Parenting Stress Index–Short Form, which were summed and averaged. For both measures, higher values reflect poorer adaptation and psychosocial functioning.

Maternal Background Characteristics

To account for factors that may select women into different family structures and be confounded with their child’s weight status, we include several covariates taken from the 9-month data collection: whether the mother was foreign born (1 = yes, 0 = no), her household is primarily non-English speaking (1 = yes, 0 = no), her mother had a high school degree (1 = yes, 0 = no), her father had some college education (1 = yes, 0 = no), she grew up in a welfare receiving household (1 = yes, 0 = no), her prepregnancy BMI (based on self-reports of her height and prepregnancy weight), and the education of the child’s father (dummy coded as less than high school, high school degree, some college, and college degree). We also account for her age, which came from the birth certificate, and indicators of family background based on the preschool interviews, including mothers’ education (less than high school, high school degree, some college, college degree) and work status (not working, part-time, full-time), whether the coresident father/partner was currently employed (1 = yes, 0 = no); and household income, which in the ECLS-B is a continuous measure ranging from 1 to 13 (1 = $5,000 or less annual household income; 13 = $200,001 or more).

Child-Level Controls

Child-level covariates that come from the birth certificate include the child’s gender (0 = male, 1 = female), race (White, Black, Hispanic, Other), and birth weight (below 2,500 grams = low birth weight; above 4,500 grams = high birth weight). Using the 9-month report, we account for whether the child was ever breastfed (Byrne, Cook, Skouteris, & Do, 2011; Gable & Lutz, 2000). Information provided at the preschool wave accounts for the number of other children in the home and the child’s primary care arrangement (center, relative, group home, exclusive maternal care). We control for child’s age in months. We also considered covariates for child birth order and health at birth, if the delivery was paid for by Medicaid, the mother smoked during pregnancy, she experienced pregnancy or delivery complications, and the family received food stamps within 12 months of the preschool data collection, but excluded such factors from the analysis because they were not significantly associated with obesity.

Analysis Plan

Our analysis begins by using simple categories of children’s current family structure. We then take steps to capture more complete family structure histories by adding interactions which reflect whether a social/biological father is in the home, if the child experienced a prior family structure change, and what type (divorce, dissolution of a cohabiting family). By proceeding in this way, we aim to capture nearly every possible variation in family structure, identify the relative importance of different dimensions of family structure (e.g., state, stability), and avoid overly nuanced results that make comparisons between different family structures difficult. For example, if in testing for differences between married biological and married stepfamilies we find none, we retain the simpler measure of “married” and note any similarities in the discussion. We describe these main modeling steps here and any robustness checks in the results section.

We first estimate the association between the three family structure states at preschool (married, cohabiting, single) and children’s odds of obesity using binary logistic regression and controlling for the full set of covariates, with married as the reference group. Because predicted probabilities are a more intuitive way of interpreting the results than odds ratios (ORs), especially when using interaction terms (which most models include), we present only the predicted probabilities in the tables and the results of the postestimation significance tests which test for any significant group differences in children’s predicted probabilities of obesity. Next, we added in a measure for prior instability, which assessed its independent link to children’s obesity while accounting for its correlation with family structure states. We then interacted instability with state so as to compare different combinations of family structure and stability to evaluate whether (a) any effect of living in a single or cohabiting family was driven by the children who had experienced instability, (b) whether children in stably single or stably cohabiting families may have different probabilities of obesity compared with children in stably married families, and (c) whether the children currently in married families whose parents married a stepfather or the biological father after the child was born were more likely to be obese than children in stably married families.

We then turned to assessing whether the observed moderating effect of a family structure change for children in single or cohabiting families was driven by a divorce among those born to married parents (85% of all divorces were among families married at the child’s birth). We tested this by interacting family structure with marital status at the child’s birth, bracketing out the 150 remarried families (for whom the odds of obesity was no different than those in stably married families) so that the reference group was the same throughout the analysis: children in stably married two-parent families. Based on the results, which revealed that divorce was consequential to the obesity of children in cohabiting families but not single-parent families, we then assessed the significance of a previous cohabitation for children in single-parent families. Thus, we again interacted family structure with instability, but this time bracketed out children in single-parent families whose parents divorced (n = 600) so the predicted probability of obesity for children in single-parent families was for those who had experienced instability following a cohabitation, not divorce. As a final step, we tested a three-way interaction term for state × stability × biological father to assess whether there may be a moderating effect of prior family structure change for children currently in a cohabiting or married family based on the biological relationship of the father to the child.

After identifying the family structure circumstances associated with higher probabilities of obesity compared with being in stably married families, we sought to explain these patterns. We began by following the causal steps approach (Baron & Kenny, 1986) in which we tested for whether any of the 11 hypothesized mechanisms were significantly associated with both (a) the family structures in which we observed an elevated risk of obesity and (b) children’s odds of obesity. Continuous factors were predicted using ordinary least squares regression, bivariate logistic regression was used for dichotomous ones, and ordered logistic regression for ordered dependent variables. Should any of the mechanisms have met these minimum criteria, we would have then proceeded with more formal test of mediation (i.e., test of the indirect effect in structural equation modeling). None did, which we discuss more in Discussion section, so we do not present such tests.

As a final word, we used multiple imputation procedures in Stata using ice to produce 20 data sets to account for item-level missing data on all independent variables and the mi estimate suite of commands to analyze the multiply imputed data in the multivariate analyses.

Results

Descriptive Statistics

Table 1 presents bivariate associations between categories of children’s family structure at preschool (and weighted percentages thereof) and (a) other family structure related variables and (b) covariates. A total of 65% of children lived with married parents during preschool, 21% were in single-parent families, and 11% lived with cohabiting mothers. The remaining 3% were in other family types. Among children in married parent families, 96% lived with their biological father (10% of whose parents married after their birth) and 4% lived in stepparent families. For both children in single and cohabiting families, 50% experienced a family structure change by preschool. For children currently in single-parent families, the transition was linked to a divorce 63% of the time. For children in cohabiting families, they were linked to divorce one third of the time.

Table 1.

Descriptive Statistics for Study Variables and Covariates by Family Structure (n = 8,350).

	Mean (SD) or percentages
	Married	Single	Cohabiting	Full sample
Child obesity
Obese	16%^b	22%^a	26%^a	18%
Family structure factors
Married at birth	90%^a	27%^c	17%^b	68%
Live with biological father at pre-K	96%^a	—	74%^b	70%
Previous family structure transition	14%^b	50%^a	50%^a	26%
Maternal background factors
Maternal age at birth	29.02^a (5.65)	24.41^b (5.99)	23.97^b (5.70)	27.32 (6.17)
Non-English-speaking household	19%^b	12%^c	32%^a	18%
Number of children in household	1.49^a (1.09)	1.16^c (1.18)	1.35^b (1.10)	1.39 (1.12)
Grandmother high school degree	70%^a	60%^b	50%^c	64%
Grandfather some college education	71%^a	58%^b	53%^c	66%
Welfare receiving household of origin	8%^c	20%^a	16%^b	11%
Maternal prepregnancy body mass index	24.78^a (5.41)	25.15^a (6.07)	25.04^a (5.62)	24.87 (5.59)
Family socioeconomic factors
Mother no high school degree	11%^c	21%^b	33%^a	15%
Mother high school degree	22%^b	37%^a	37%^a	27%
Mother some college	32%^a	31%^a	26%^b	31%
Mother college degree	35%^a	10%^b	4%^c	26%
Mother part-time work (not working)	22%^a	17%^b	15%^b	20%
Mother full-time work	36%^b	50%^a	40%^b	40%
Father college degree	34%^a	7%^b	3%^c	24%
Father employed full-time	90%^a	—	80%^b	70%
Household income group (0-13)	9.02^a (2.89)	4.98^c (3.17)	5.72^b (2.71)	7.75 (3.42)
Child characteristics
Child gender (male)	50%^a	49%^a	46%^a	49%
Child race White	68%^a	36%^b	37%^b	57%
Child race Black	6%^c	37%^a	16%^b	14%
Child race Hispanic	20%^b	22%^b	43%^a	23%
Child other race/ethnicity	6%^a	5%^b	4%^b	6%
Child low birth weight	6%^b	10%^a	10%^a	8%
Child high birth weight	10%^a	6%^{a b}	9%^a	9%
Breastfed for 6 months	77%^a	54%^c	63%^b	70%
Center care	60%^a	60%^a	50%^b	58%
Group care	8%^a	8%^a	8%^a	8%
Relative care	11%^b	20%^a	16%^{a b}	14%
No nonparental care	21%^b	13%^c	26%^a	20%
Weighted percentage	65%	21%	11%	97%
Raw ns	5,600	1,750	850	8,350

Note. Data weighted using weight W13C0. Different superscripts across columns indicate significant difference in means (p < .05) determined by one-way analysis of variance and Duncan’s multiple range test. The ‘a’ superscript represents the highest mean level

Raw ns are rounded to nearest 50th.

Children in married parent families had the lowest rates of obesity (16%). Children in cohabiting families had the highest rates at 26%, followed by children in single-parent families (22%), but these differences were not significantly different. These links might be “controlled away” by socioeconomic factors—for example, only 4% of cohabiting mothers and 10% of single mothers had a college degree versus 35% of married mothers—or factors linked to selection into certain family structure trajectories. For example, single and cohabiting mothers were significantly more likely to have lived with families growing up that received public assistance (16%, 20%) than married mothers (8%). They could also be concentrated among the single and cohabiting families who had experienced instability, which these bivariate models cannot tease apart.

These patterns may also explained, in part, by the proposed mechanisms. The bivariate associations between them and family structure in Table 2 reveal that children in cohabiting and single-parent families lived in families with fewer rules than children in married families, had fewer shared meals, ate unhealthy foods more often, had more screen time, were more likely to go to bed late, and were less likely to be enrolled in a physical activity. Their mothers reported more depressive symptoms and externalizing problems in their children. With four exceptions (single mothers reported less parenting stress and less television watching for their children; cohabiting mothers had less depression and more shared meals), the differences between these two groups were not significant. A few bivariate patterns were contrary to our expectations; married mothers reported more parenting stress than other unmarried mothers and less family outdoor play and more child internalizing problems compared with single mothers.

Table 2.

Descriptive Statistics for Mechanisms (n = 8,350).

	Mean (SD) and percentages
	Married	Single	Cohabiting	Full sample
Household mechanisms
Rules	3.34^a (0.93)	3.10^b (1.03)	3.05^b (1.11)	3.41 (0.82)
Shared meals	5.63^a (1.77)	5.12^c (2.07)	5.41^b (1.94)	5.50 (1.86)
Child diet	2.51^b (0.88)	2.78^a (0.99)	2.75^a (0.99)	2.60 (0.92)
Organized physical activities	36%^a	22%^b	19%^b	31%
Family outdoor play	2.58^b (1.01)	2.71^a (1.11)	2.48^b (1.05)	2.60 (1.04)
Late bedtime	11%^b	16%^a	18%^a	13%
Screen time	3.52^c (3.18)	4.15^a (3.74)	4.50^b (3.80)	3.77 (3.41)
Child behavioral factors
Internalizing problems	2.05^a (0.65)	1.95^b (0.70)	2.03^{a b} (0.69)	2.03 (0.70)
Externalizing problems	2.32^b (0.57)	2.45^a (0.69)	2.44^a (0.65)	2.38 (0.62)
Mother factors
Parenting stress	3.05^a (0.64)	2.89^b (0.72)	2.99^a (0.73)	3.01 (0.67)
Maternal depression	1.37^c (0.42)	1.60^a (0.53)	1.46^b (0.51)	1.43 (0.46)

Note. Data are weighted using weight W13C0. Different superscripts across columns indicate significant difference in means (p < .05), as determined by one-way analysis of variance and Duncan’s multiple range test.

Represents the highest mean level. Coefficients with the same superscript do not significantly differ.

Multivariate Analyses

In Model 1, we estimate the association between three categories of children’s family structure during the preschool data collection and obesity based on assessments from the same wave, net of the covariates. Table 3 presents predicted probabilities of obesity by family structure where the first category in each panel is the reference category. Statistical significance reflects a significant difference between groups compared with the top group. We provide the ORs from Model 1 in the text. Results reveal that children in single and cohabiting families have a significantly higher odds of obesity than children in married families (OR = 1.54, standard error [SE] = 0.32; OR = 1.56, SE = 0.26). For both groups, the probability of obesity was 24%, compared with 18% for children with married parents. As a check of selection, we substituted the preschool measures of family structure with those from the first data collection at 9 months. If we found a significant difference in children’s probability of obesity at preschool by their family structure at 9 months, we might infer that unmeasured selection may be contributing to the Model 1 associations. We found no significant differences in children’s obesity by their family structure at 9 months.

Table 3.

Predicted Probabilities of Children’s Obesity by Family Structure (n = 8,350).

	Predicted probabilities (SE)
	Model 1	Model 2	Model 3	Model 4	Model 5
Panel A: Full sample
Married	.175 (.013)	.176 (.013)	—	—	—
Single	.242* (.031)	.240^† (.031)	—	—	—
Cohabiting	.244* (.032)	.241* (.032)	—	—	—
Panel B: Full sample
Stably married since birth	—	—	.175 (.015)	—	—
Married with prior transition	—	—	.171 (.021)	—	—
Stably single since birth	—	—	.238 (.037)	—	—
Single with prior transition	—	—	.249^† (.036)	—	—
Stably cohabiting since birth	—	—	.211 (.036)	—	—
Cohabiting with transition	—	—	.276** (.039)	—	—
Panel C: Restricted sample^a
Married, married at birth	—	—	—	.177 (.015)	—
Married, not married at birth^b	—	—	—	.188 (.024)	—
Single, unmarried at birth	—	—	—	.251^† (.035)	—
Single, married at birth	—	—	—	.214 (.036)	—
Cohabiting, unmarried at birth	—	—	—	.232^† (.033)	—
Cohabiting, married at birth	—	—	—	.311* (.053)	—
Panel D: Restricted sample^c
Stably married since birth	—	—	—	—	.178 (.014)
Married with prior transition	—	—	—	—	.172 (.021)
Stably single since birth	—	—	—	—	.237 (.038)
Single, previously transition	—	—	—	—	.284* (.046)
Stably cohabiting since birth	—	—	—	—	.210 (.036)
Cohabiting with transition	—	—	—	—	.275* (.039)

Note. SE = standard error. The reference category is first category listed under each panel.

Model 3 was estimated on a subsample that excludes remarried families using the subpop command. ^bIncludes families with a biological father present and stepparent families. ^cModel 4 used subpop to bracket out single-parent families with a prior divorce.

***

p < .001. **p < .01. *p < .05. ^†p < .10.

Next, we added in family structure instability to the model (Model 2). Doing so reduced the difference in the probability of obesity for children in single and married families to marginal significance, yet the difference for children in cohabiting families compared with children in married families remained significant at the p < .05 level. Here we conduct our next sensitivity check to assess whether there were any differences in the probability of obesity for children in cohabiting families with a biological father versus stepfather. We did this by limiting our sample to children in married and cohabiting families and interacting family structure with an indicator for whether the biological father was in the home. We did not find significant differences among children in cohabiting families with stepfathers compared with biological fathers. We also did not find any difference by whether the mother was married to the biological father versus stepparent. As such, at this point, we proceeded with our same categories for family structure.

In Model 3, we interacted family structure with instability. This approach allowed us to compare children in five categories with children in stably married families: married family with prior transition, stably single, single family with instability, stably cohabiting, and cohabiting with instability. Results revealed that compared with children living with a stably married family (Predicted probability [PP] = .175, SE = .015), children in cohabiting families (PP = .276, SE = .039) and children in single-parent families (PP = .249, SE = .036) who experienced some sort of family structure instability were significantly more likely to be obese, although the difference for the single group only reached significance at the p < .10 level. There were no significant differences between children in stably married families and children in unstably married families in their likelihood of obesity, or with children in stably single or stably cohabiting families. At this point, our results suggest that children in unmarried families who experienced instability have a higher risk of being obese.

This presents the question: What type of instability is driving this pattern? We begin by focusing on a prior divorce by interacting marital status at birth with family structure, excluding the remarried from the analysis (Model 4). As such, stably married is again the reference group, but our comparison groups are married who were unmarried at birth, single unmarried at birth, single married at birth, cohabiting and unmarried at birth, cohabiting married at birth. We find a marginally significant difference for children in single-parent families also born to unmarried parents (PP = .251, SE = .035, p < .055), but a nonsignifiant difference for children in single-parent families whose parents were married at their birth then divorced. This pattern suggests that the increased risk of obesity for children in unstable single-parent families may be driven by those who experienced the dissolution of a cohabiting family, not a divorce, but we must investigate further because this estimate includes stably single-parent families too. For children in cohabiting families, divorce may be associated with a heightened risk of obesity (PP = .311, SE = .053, p < .05), but the transition to a cohabiting family seems to matter too (PP = .232, SE = .033, p < .078).

In Model 5, we tease out the patterns in single- and cohabiting-parent families further by comparing children in stably married families with children in single- and cohabiting-parent families that were (a) unmarried at their birth and (b) had experienced a family structure change. We did so by interacting family structure with instability, removing children in single-parent families born to married families from the analytical sample. Here, we find significant differences in the probability of obesity between children in stably married families and children in single-parent families that experienced instability (PP = .284, SE = .046, p < .05), suggesting it is in fact the children in single-parent families who had once been in a short-term cohabiting or married family (parents married after they were born but since divorced) who were the subset of children in single-parent families more likely to be obese. Sample sizes were too small to adjudicate between these two family structures, but removing the previously married group did not change the results (PP = .275, SE = .039, p < .04).

Sensitivity Analyses

In models not shown, we test for differences in the married with prior transition group and cohabiting with prior transition according to whether the partner was a biological or stepparent. We find no significant differences by the relationship of the father to the child. We also tested for a cumulative and/or timing effect of instability. We tested for a cumulative effect by replacing the binary measure of instability (0, 1) with a continuous one (ranging from 0 to 3) that captured the total number of transitions. We tested for a timing effect two ways: by replacing the binary measure of instability with several binary ones for the timing of the instability (between birth and 9 months, 9 months and 2 years, 2 years and preschool), which allowed children to be in more than one category, but there was the potential for collinearity between variables. The other approach involved creating dummy codes for these categories, plus one for multiple transitions and one for no transition, which eliminated the collinearity problem but yielded smaller cell sizes. The results of each of these analyses revealed no evidence for a timing or cumulative effect. We also investigated whether the pattern of results would be different if we included overweight children with obese children (thus removing them from the reference category), but found a remarkably similar result with one exception. Divorce among children in single-parent families was significantly associated with increased risk of overweight/obesity.

Finally, we explored differences in the findings by the child’s gender. Estimating our models separately for males and females revealed that the results were largely driven by females, for whom the coefficients were both larger and stronger. The pattern of results for male children was similar to the pattern found for the full sample, but most coefficients did not reach statistical significance. We discuss the implications of this gender pattern in the Discussion. We did not have sufficient cell sizes to examine variation by race/ethnicity.

Testing for Mechanisms

As a final exploratory step, we sought to identify any mechanisms that explained why the children with certain family structure histories were more likely to be obese than children in stably married families. Thus, we created a new measure of family structure for children in cohabiting families formed after the child’s birth and one for single-parent families that experienced a prior short-term marriage or cohabitation. Next, we assessed whether any of the mechanisms, measured at the preschool wave, significantly varied between these groups and the stably married (see Table 4). We only observed two differences in the expected direction: in bedtime (OR = 1.44, SE = 0.27) and mothers’ depression (Β = 0.16, SE = 0.04) for the single group at the .05 level. We then regressed obesity on the significant mechanisms, but neither one was a significant predictor (not shown). Combining the two groups to increase statistical power also did not alter the pattern of results.

Table 4.

Logistic and Linear Regression Models of Family Structure Predicting Hypothesized Mechanisms (n = 8,350).

	Β (SE)
	Rules	Screen time	Shared meals	Diet	Outside time	Physical activity	Late bedtime	Internalizing	Externalzing	Parent stress	Depression
Reference = stably married
Single group^a	1.14 (0.19)	1.16 (0.18)	0.85 (0.15)	0.04 (0.08)	0.95 (0.16)	−0.13 (0.23)	1.44* (0.27)	0.08 (0.06)	0.04 (0.08)	−0.03* (0.05)	0.16*** (0.04)
Cohabiting group^b	1.16 (0.18)	0.82 (0.12)	1.04 (0.15)	−0.10 (0.06)	0.99 (0.13)	−0.01 (0.21)	1.41 (0.34)	0.01 (0.05)	0.08^† (0.05)	0.01 (0.06)	0.05 (0.03)

Note. SE = standard error.

Single group includes families with a prior family structure change related to cohabitation or the dissolution of a marriage that began after the child was born. ^bThe cohabiting group includes families with any prior family structure change.

***

p < .001. **p < .01. *p < .05. ^†p < .10.

As final efforts to explain our findings, we combined indicators of household routine and time use into a scale by taking the z score of each measure and averaging them to create greater variability. This new measure, while predicting obesity better than the individual indicators, did not vary by family structure. We also interacted family structure with the mechanisms that significantly varied by it in the expected direction (maternal depression, late bedtime) to test whether, in these families, such factors have a differential effect on children’s likelihood of obesity. Unfortunately, again, these tests also did not help explain our findings. Finally, we reran the analyses just on females, for whom the “effects” were most pronounced, with similar effect.

Discussion

As the proportion of children raised outside a traditional nuclear family continues to rise, scholars have amassed a large literature linking different family structures to various indicators of children’s academic and behavioral development. Yet few studies have sought to examine the implications of different family structures for children’s obesity. Thus, our aim was to examine how children’s likelihood of obesity varied across a range of family structure histories defined by their current status and family relatedness, prior instability, and type thereof. Our focus was on obesity among young (vs. older) children, for whom the long-term impact of obesity might be most severe (Field, Cook, & Gillman, 2005), doing so during the time before they have begun school, when the significance of family structure for children’s development might be most salient (Cavanagh & Huston, 2008). We also hoped to provide the first glimpse into what explained these links.

The results of our study revealed, first, that preschool-aged children in married parent families had the lowest rates of obesity, irrespective of whether their parents were married at the time of their birth or they lived with a stepfather. This finding is contrary to research linking family structure to other measures of child well-being, which find small but significant negative effects of living with a stepparent on young children’s behavioral and cognitive development (Hetherington et al., 1998). Yet it is consistent with both Schmeer (2012) and Bzostek and Beck (2011), who found no difference in children’s weight status between those with stably married families and those in married families who had experienced a prior transition. Based on these prior findings, it appears that having a stepparent may have less significance for children’s weight development than it does for other domains of child development.

The next key finding to emerge was that children in stably single and stably cohabiting families were no more likely to be obese than children living in married parent families. This finding echoes the growing concensus among family scholars that, in the situation of unmarried families, what matters for children’s well-being is not so much who is (or is not) in the home, but the stability of that family arrangement (Waldfogel et al., 2010). Ours is the first study to clarify the significance of family structure stability for the weight status of children with unmarried mothers. These results, however, are only partly consistent with Schmeer (2012) and Bzostek and Beck (2011), both of whom reported no elevated risk of obesity associated with living in stably cohabiting families, but higher risks for children living in stably single-parent families.

To explore whether this latter difference may be due to differences in the characteristics of our samples, we repeated our analysis with a subsample that better resembled the sample used in both studies, the Fragile Families and Child Wellbeing Study, by dropping the 25% of cases that had incomes at or above the 90th income percentile in the Fragile Families and Child Wellbeing Study (about $62,500). These results were consistent with our original results, except that in Model 2, we observed a higher probability of obesity for children in single-parent families versus married parent families even after accounting for instability; and in Model 3, we observed a marginally significant higher probability of obesity for children in stably single-parent families versus stably married parent families. Given that the sizes of these estimates were nearly identical, this analysis suggests that being stably single may be associated with a greater risk of obesity for children from more economically disadvantaged circumstances. Yet in our nationally representative sample, it is not.

The third key finding was that for children in both single and cohabiting families who had experienced instability, there was an elevated probability of obesity. For children in single-parent families, however, this finding was not generalizable to children whose mothers were married at their time of birth, divorced, and then remained single (although we did observe an association when predicting overweight). For this group, it is possible that certain resources remained in place that we could not account for, such as greater father involvement. It is also possible that these women are a select group whose decision not to recouple, at least while their children were young, reflected a degree of concern for their children’s well-being that was associated with their lower likelihood of obesity. The next noteworthy finding was that family relatedness (i.e., the biological relationship of the father to the child) did not matter for children’s weight status, once two other features of children’s family structures—the current state and stability of this state since birth—were accounted for. Finally, we found that girls, not boys, were driving much of the results. This finding is consistent with other studies (Hernandez, Pressler, Dorius, & Mitchell, 2014) that find that females’ weight is more sensitive to changes in family structure.

What explained these patterns, in which children in unmarried families who had experienced a prior family structure change (except for children in single-parent families born to married ones) had higher risks of obesity, but especially girls? Unfortunately, our analysis in which we explored 11 different mechanisms did not reveal a connection between these family structure experiences and children’s likelihood of obesity, even when focusing on the female subsample. This general result was not unprecedented. Bzostek and Beck (2011) also failed to identify a mechanism linking family structure to children’s health and obesity, and in response, suggested we need to think differently about how children’s family structure experiences matter for their weight development. We echo these sentiments and elaborate on them.

First, we argue for the need for greater contextual information that capture health-related behaviors on the part of children in their various environments, including formal child care and the care of relatives, friends, and nonresident fathers. Such environments are also likely to be more varied among children who experienced a prior family structure change, who experience more multiple nonparental care arrangements (J. Chen, 2013), on average, and time with nonresident fathers or other relatives, who may be more likely to break from healthy practices (e.g., induldging in snacks) than the mother (e.g., Stewart & Menning, 2009). Thus, the mother reported data in the ECLS-B may underestimate differences in the health behaviors between children in families with instability and children in stably married families. Moreover, given the multiplicity of contexts such children are likely to experience, it is possible that patterns of behavior exist too—which we also cannot observe from the ECLS-B, or other similar sources of data focused on the mother–child dyad—and it is the combination of factors, not singular ones, that increases the risk of obesity. It is also possible that other physiological processes are at play, which parents are not adept at observing and which result in weight gain (Epel, Lapidus, McEwan, & Brownell, 2001). They may explain why females, who cope with stresses associated with family contexts differently than males, had more negative weight outcomes associated with their family structure. Finally, we believe that we must pursue unexplored pathways linking family structure and children’s weight development. Qualitative data might be especially valuable in this pursuit.

As for limitations, we used an observational design and cannot infer causality given that unmeasured sources of selection associated with family structure and obesity may remain. Given that data collections were at birth, 9 months, 2 years, and preschool, we also could not account for some instances of instability that occurred between waves. Yet given our results, we believe that the instabilty “effects” might be even greater if we could capture all instances of it. Based on our analysis strategy, we also could not capture all family structure histories. While in most cases our approach linked children’s current family structure to one they were born into and all intervening ones, there were cases when some were glossed over—for example, if there were multiple transitions. In these cases, sample sizes would also be too small to detect significant patterns—another limitation. Fifth, our mechanisms were based on mother reports, which may be one reason why we did not find much variation in them by family strucure.

In sum, given the complex factors associated with the problem of child obesity, scholars must pay more attention to how family contexts are connected to children’s weight as they have done with other domains of child development. This study is a step in that direction. Based on nationally representative data, we identified the family structure circumstances associated with children’s increased likelihood of obesity in a way that provides additional support for, yet also expands, the findings reported elsewhere, although we failed to identify the mechanisms that explain this link. Thus, we hope this study can stimulate greater investigation into how children’s family structure circumstances shape family and child processes related to children’s weight.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Amato

(2005). The impact of family formation change on the cognitive, social, and emotional well-being of the next generation. Future of Children, 15, 75-96.

Anderson

S. E.

Schoppe-Sullivan

Must

(2010). Externalizing behavior in early childhood and body mass index from age 2 to 12 years: Longitudinal analysis of a prospective cohort study. BMC Pediatrics, 10, 49.

Anderson

S. E.

Whitaker

R. C.

(2010). Household routines and obesity in US preschool-aged children. Pediatrics, 125, 420-428.

Augustine

J. M.

Kimbro

R. T.

(2013). Family structure and obesity among U.S. children. Journal of Applied Research on Children, 4, 1-24.

Balistreri

Van Hook

(2009). Socioeconomic status and body mass index among Hispanic children of immigrants and children of natives. American Journal of Public Health, 9, 2238-2246.

Baron

R. M.

Kenny

(1986). The moderator-mediator variable distinction in social psychological research. Journal of Personality and Social Psychology, 51, 1173-1182.

Beck

Cooper

McLanahan

Brooks-Gunn

(2010). Partnership transitions and maternal parenting. Journal of Marriage and Family, 72, 219-233.

Bowman

S. A.

Harris

E. W.

(2003). Food security, dietary choices and television viewing: Status of preschool-aged children living in single-parent or two-parent households. Family Economics & Nutrition Review, 15, 29-34.

Bronfenbrenner

(1994). Ecological models of human development. In Peterson

Baker

McGaw

(Eds.), International encyclopedia of education (Vol. 3, 2nd ed., pp. 1643-1647). Oxford, England: Elsevier.

10.

Brown

S. L.

(2010). Marriage and child well-being: Research and policy perspectives. Journal of Marriage and Family, 72, 1059-1077.

11.

Byrne

L. K.

Cook

K. E.

Skouteris

(2011). Parental status and childhood obesity in Australia. International Journal of Pediatric Obesity, 6, 415-418.

12.

Bzostek

S. H

Beck

(2011). Familial instability and young children’s physical health. Social Science & Medicine, 73, 282-292.

13.

Bzostek

S. H.

McLanahan

Carlson

M. J.

(2012). Mothers’ repartnering after a nonmarital birth. Social Forces, 90, 817-841.

14.

Carlson

M. J.

Corcoran

M. E.

(2001). Family structure and children’s behavioral and cognitive outcomes. Journal of Marriage and Family, 63, 779-792.

15.

Carr

Springer

K. W.

(2010). Advances in families and health research in the 21st century. Journal of Marriage and Family, 72, 743-761.

16.

Case

Paxson

(2002). Parental behavior and child health. Health Affairs, 21, 164-178.

17.

Cavanagh

Huston

(2006). Family instability and children’s early problem behavior. Social Forces, 85, 575-605.

18.

Cavanagh

Huston

(2008). The timing of family instability and children’s social development. Journal of Marriage and Family, 70, 1258-1269.

19.

Chen

A. Y.

Escarce

J. J.

(2010). Family structure and childhood obesity, early childhood longitudinal study—Kindergarten cohort. Preventing Chronic Disease, 7, A50. Retrieved from http://www.cdc.gov/pcd/issues/2010/may/09_0156.htm

20.

Chen

(2013). Multiple childcare arrangements and health outcomes in early childhood. Maternal and Child Health Journal, 17, 448-455.

21.

Cochran

W. G.

(1977). Sampling techniques (3rd ed.). New York, NY: John Wiley & Sons.

22.

Cooper

C. E.

Osborne

C. A.

Beck

A. N.

McLanahan

S. S.

(2011). Partnership instability, school readiness, and gender disparities. Sociology of Education, 84, 246-259.

23.

Cragie

Brooks-Gunn

Waldfogel

(2015). Family structure, family stability, and early child wellbeing. Retrieved from http://crcw.princeton.edu/workingpapers/WP10-14-FF.pdf

24.

Crosnoe

Cavanagh

(2010). Families with children and adolescents. Journal of Marriage and Family, 72, 594-611.

25.

Currie

(2008, May). Healthy, wealthy, and wise: Socioeconomic status, poor health in childhood, and human capital development (Working Paper No. w13987). National Bureau of Economic Research, Cambridge, MA.

26.

Dietz

W. H.

(1998). Health consequences of obesity in youth: Childhood predictors of adult disease. Pediatrics, 101, 518-525.

27.

Epel

Lapidus

McEwan

Brownell

(2001). Stress may add bite to appetite in women: A laboratory study of stress induced cortisol and eating behavior. Psychoneuroendocrinology, 26, 37-49.

28.

Field

Cook

N. M.

Gillman

M. W.

(2005). Weight status in childhood as a predictor of becoming overweight or hypertensive in early adulthood. Obesity Research, 13, 163-169.

29.

Gable

Lutz

(2000). Household, parent, and child contributions to childhood obesity. Family Relations, 49, 293-300.

30.

Gerald

L. B.

Anderson

Johnson

G. D.

Hoff

Trimm

R. F.

(1994). Social class, social support and obesity risk in children. Child: Care, Health and Development, 20, 145-163.

31.

Gibson

L. Y.

Byrne

S. M.

Davis

E. A.

Blair

Jacoby

Zubrick

S. R.

(2007). The role of family and maternal factors in childhood obesity. Medical Journal of Australia, 186, 591-595.

32.

Guo

S. S.

Chumlea

(1999). Tracking of body mass index in children in relation to overweight in adulthood. American Journal of Clinical Nutrition, 70, 145-148.

33.

Hernandez

Pressler

Dorius

Mitchell

C. S.

(2014). Does family instability make girls fat? Gender differences between instability and weight. Journal of Marriage and Family, 76, 175-190.

34.

Hesketh

Crawford

Salmon

Jackson

Campbell

(2007). Associations between family circumstance and weight status of Australian children. International Journal of Pediatric Obesity, 2, 86-96.

35.

Hetherington

Bridges

Insabella

(1998). What matters? What does not? Five perspectives on the association between marital transitions and children’s adjustment. American Psychologist, 53, 167-184.

36.

Huffman

Kanikireddy

Patel

(2010). Parenthood—A contributing factor to childhood obesity. International Journal of Environmental Research and Public Health, 7, 2800-2810.

37.

Kamp Dush

C. M.

(2009). An examination of child well-being in stable single parent and married families. In Peters

H. E.

Kamp Dush

C. M.

(Eds.), Marriage and family: Perspectives and complexities (pp. 143-176). New York, NY: Columbia University Press.

38.

Kiernan

Pickett

K. E.

(2006). Marital status disparities in maternal smoking during pregnancy, breastfeeding and maternal depression. Social Science & Medicine, 63, 335-346.

39.

Kimbro

R. T.

(2008). Together forever? Romantic relationship characteristics and prenatal health behaviors. Journal of Marriage and Family, 70, 745-757.

40.

Kimbro

R. T.

Brooks-Gunn

McLanahan

(2011). Young children in urban areas: Links among neighborhood characteristics, weight status, outdoor play, and television-watching. Social Science & Medicine, 72, 668-676.

41.

Kitsantas

Gaffney

K. F.

(2010). Risk profiles for overweight/obesity among preschoolers. Early Human Development, 86, 563-568.

42.

Luo

Z. C.

Wilkins

Kramer

M. S.

(2004). Disparities in pregnancy outcomes according to marital and cohabitation status. Obstetrics and Gynecology, 103, 1300-1307.

43.

Magnuson

Berger

L. M.

(2009). Family structure states and transitions: Associations with children’s well-being during middle childhood. Journal of Marriage and Family, 71, 575-591.

44.

Mathews

T. J.

MacDorman

M. F.

(2012). Infant mortality statistics from the 2008 period linked birth/infant death data set. National Vital Statistics Reports, 60. Retrieved from http://www.cdc.gov/nchs/data/nvsr/nvsr60/nvsr60_05.pdf

45.

McCubbin

Patterson

(1983). The family stress process. Marriage & Family Review, 6, 7-37.

46.

McEwen

B. S.

(2008). Understanding the potency of stressful early life experiences on brain and body function. Metabolism: Clinical and Experimental, 57, 11-15.

47.

Must

Parisi

S. M.

(2009). Sedentary behavior and sleep: Paradoxical effects in association with childhood obesity. International Journal of Obesity, 33, 82-86.

48.

Ogden

C. L.

Carroll

M. D.

Curtin

L. R.

Lamb

M. M.

Flegal

K. M.

(2010). Prevalence of high body mass index in US children and adolescents, 2007-2008. Journal of the American Medical Association, 303, 242-249.

49.

Raley

R. K.

Wildsmith

(2004). Cohabitation and children’s family instability. Journal of Marriage and Family, 66, 210-219.

50.

Roisman

G. I.

Fraley

R. C.

(2012). A behavior-genetic study of the legacy of early caregiving experiences. Child Development, 83, 728-742.

51.

Sanders

M. R.

Woolley

M. L.

(2005). The relationship between maternal self-efficacy and parenting practices: Implications for parenting training. Child: Care, Health and Development, 31, 65-73.

52.

Schmeer

(2011). The child health disadvantage of parental cohabitation. Journal of Marriage and Family, 73, 181-193.

53.

Schmeer

(2012). Family structure and obesity in early childhood. Social Science Research, 41, 820-832.

54.

Stewart

S. D.

Menning

C. L.

(2009). Family structure, nonresident father involvement, and adolescent eating patterns. Journal of Adolescent Health, 45, 193-201.

55.

Strauss

R. S.

(2000). Childhood obesity and self-esteem. Pediatrics, 105, e15.

56.

The

N. S.

Gordon-Larson

(2009). Entry into romantic parternship is associated with obesity. Obesity, 17, 1441-1447.

57.

Waldfogel

Craigie

Brooks-Gunn

(2010). Fragile families and child wellbeing. Future of Children, 20, 87-112.

58.

Wen

(2008). Family structure and children’s health and behavior data from the 1999 national survey of America’s families. Journal of Family Issues, 29, 1492-1519.