Abstract

Living in a doubled-up, or shared, household is a common experience. Nearly one-half of children in the United States double up at some point during childhood, yet we know little about the cumulative effects of these households on children. This study estimates the effects on young adult health and educational attainment of childhood years spent in three doubled-up household types: (1) those formed with children’s grandparent(s), (2) those formed with children’s adult sibling(s), and (3) those formed with other extended family or non-kin adults. Using marginal structural models and inverse probability of treatment weighting—methods that account for the fact that household composition is both a cause and consequence of other family characteristics—I find that doubling up shapes children’s life chances, but the effects vary depending on children’s relationships with household members. Childhood years spent living with nongrandparent extended family or non-kin adults are associated with worse young adult outcomes, but coresidence with grandparents is not significantly associated with young adult outcomes after selection into these households is accounted for, and coresidence with adult siblings may be beneficial in some domains. By studying the effects of coresidence with adults beyond the nuclear family, this research contributes to a fuller understanding of the implications of family complexity for children.

Introduction

Children’s lives are profoundly affected by the adults with whom they live. Previous research has linked family structure to a variety of behavioral and cognitive childhood outcomes as well as young adult outcomes, such as family formation and employment (McLanahan and Percheski 2008; McLanahan et al. 2013). This research has mainly focused on the nuclear family, particularly the presence/absence of a father or other romantic partner of the mother. However, looking only at parents’ romantic partners and minor children fails to capture the full household experience of many children, particularly those from low-income families, who often spend at least part of childhood in households with more complicated arrays of residents. Doubling up—when a nuclear family coresides with other adults, such as grandparents, other extended family, or friends—is a common childhood experience. More than 45% of children in the United States double up at some point during childhood (author’s calculation). Despite extensive research on family structure, we know less about how shared household members affect children.

One reason doubled-up households are relatively neglected in the family literature may be that they are often considered a housing arrangement rather than a family form (e.g., Desmond 2012; Skobba and Goetz 2013). Yet, like family structure, household composition shapes children’s lives in myriad ways. Doubling up affects children’s access to resources and caregiving time and influences parents’ stress levels (Harvey 2020; Kalil et al. 2014; Mutchler and Baker 2009). In this study, I conceptualize doubling up as not just a residential outcome but also a social environment for children. I attend to the relational aspects of doubled-up households by considering how the effects of doubling up may vary based on children’s familial or nonfamilial relationships with other household members. By studying the effects of coresidence with adults beyond the nuclear family, this research contributes to a fuller understanding of the implications of household complexity for children. Moreover, rates of doubling up have increased in recent years (Eggers and Moumen 2013; Pilkauskas and Cross 2018), and documenting the effects of these households and variation by household type is important for considering potential repercussions of these changes.

The instability that characterizes many children’s households (Perkins 2017) complicates studies of the effects of household complexity on children’s long-term outcomes. Children’s outcomes are shaped not only by their immediate household environment but also by the sum of environments they have experienced in the past. Early environments put in motion processes of cumulative advantage and disadvantage that shape outcomes later in the life course (Elder 1998). Thus, a longitudinal approach, which accounts for household composition throughout childhood, is necessary for understanding long-term effects. Yet, as I discuss later, accurately modeling cumulative effects poses methodological challenges, and studies have generally estimated the impact of household composition at a single point in time (Astone and Washington 1994; DeLeire and Kalil 2002). These cross-sectional measures ignore the dynamic nature of household composition for many children. This limitation is particularly problematic for doubled-up households because these households are frequently unstable and many children transition in and out of doubled-up households multiple times (Mollborn et al. 2012; Pilkauskas et al. 2014).

In this article, I estimate the cumulative effects of childhood years spent in doubled-up households on young adult health and educational attainment. Recognizing that relationships between doubled-up household members are important for household functioning, I examine effects separately for three shared household types: (1) those formed with children’s grandparent(s), (2) those formed with children’s adult sibling(s), and (3) those formed with another extended family or non-kin adult(s). Rather than examining household composition at a single point, I operationalize household composition with a measure of cumulative exposure to each household type throughout childhood. I estimate the effects of additional years spent in each household type using marginal structural models and inverse probability of treatment weighting (IPTW). By identifying the long-term effects of doubling up in childhood, this research reveals that household members beyond parents and their romantic partners have enduring effects on children’s life chances and demonstrates the importance of expanding the study of family complexity to include household complexity involving adults outside the nuclear family.

Effects of Doubling Up in Childhood

Compared with more well-studied forms of family complexity, doubling up is a common childhood experience. More children in the United States live with a grandparent, other relative, and/or nonrelative than in either stepfamilies or cohabiting families (Kennedy and Fitch 2012). Although we know relatively little about the effects of doubling up and how these vary across household types, previous research has suggested that doubling up shapes children’s physical surroundings, access to material resources, and social environments—factors that are key to cognitive and socioemotional development and physical health. This analysis examines young adult outcomes in two domains: education and health. As outcomes of interest, I focus on (1) high school graduation and college attendance because of their role in labor market outcomes; and (2) depression, smoking, and obesity because of their importance for long-term health.

Although prior research has suggested many mechanisms through which doubling up may affect educational attainment and health, there is little consensus on the expected direction of this effect. Qualitative work has highlighted the importance of doubling up as a private housing safety net (Desmond 2012; Skobba and Goetz 2013). Doubled-up individuals can benefit from higher household income (Mutchler and Baker 2009; Mykyta and Macartney 2012) and economies of scale, including lower housing costs (Pilkauskas et al. 2014). Doubling up can also allow children to escape dangerous neighborhoods or attend better schools (Ahrentzen 2003; Goodman and Silverstein 2002). Thus, doubling up could benefit children by increasing material well-being and improving housing, neighborhood, and school environments.

Yet other research has suggested ways that doubling up may negatively affect children. Because doubled-up adults are more likely to have poverty-level personal incomes than non–doubled-up adults (Mykyta and Macartney 2012), increasing household size can strain limited resources (Clampet-Lundquist 2003). Doubling up may also expose children to overcrowded and unsafe conditions (Edin and Shaefer 2015; Seefeldt and Sandstrom 2015). Finally, doubled-up households are often stressful and conflictual environments (Domínguez and Watkins 2003; Harvey 2020). In sum, extant research has hypothesized mechanisms through which doubling up may both positively and negatively affect children, and the ultimate direction of the effect is unclear.

Different types of doubled-up households may have distinct dynamics that influence children in disparate ways. Most prior research focused on multigenerational households. Although findings from these studies are inconsistent (see Dunifon et al. 2014 for a review), multigenerational households are often considered supportive environments, especially for young mothers. Previous research has identified positive associations between multigenerational coresidence and mothers’ work and school activities (Gordon et al. 2004; Hao and Brinton 1997), and children receive substantial time investments from coresident grandparents (Kalil et al. 2014). However, other research has found negative associations between multigenerational coresidence and parenting quality (Black and Nitz 1996; Unger and Cooley 1992). Maternal stress and household conflict may reduce parenting quality in these households; additionally, there may be a “diffusion of parenting responsibility” if mothers and grandmothers assume that the other will take on more child-rearing responsibility (Chase-Lansdale et al. 1994:389).

In recent years, a growing number of young adults have “failed to launch” from their natal home or “boomeranged” back into it (Berlin et al. 2010). Although such arrangements are common, Americans hold ambivalent and negative views about whether adult children using coresidence with parents as a safety net is desirable (Seltzer et al. 2012), and these households are doubled up according to standard definitions (Eggers and Moumen 2013; Mykyta and Macartney 2012). Young adults living in parental homes drove much of the increase in doubling up during the Great Recession (Eggers and Moumen 2013). Such households appear consequential for both adult children and their parents; a growing literature documents associations with mental health and well-being, family formation, and financial security (Fingerman et al. 2012; Furstenberg 2010; Maroto 2017; Tosi and Grundy 2018; White 1994). Yet, to my knowledge, no study has examined how adult sibling coresidence affects minor children.

Coresidence with adult children may stretch parental resources and reduce parents’ capacities to invest in minor children, particularly if parents expected their adult child to leave home on a more traditional (earlier) schedule. Coresidence may lengthen the timeline of parenting obligations and increase the support parents provide to adult children (Fingerman et al. 2012; Swartz 2009). In these households, parents typically contribute most of the income and household work (White 1994), and coresident adult children are associated with declines in parents’ savings (Maroto 2017). Additionally, because parents and adult children generally perceive intensive parental support into adulthood as aberrant (Fingerman et al. 2012), doubling up may have psychological costs that affect household functioning.

Alternatively, coresidence with adult siblings might produce supportive childhood environments. Parents and adult children often have positive feelings about coresidence for young adults pursuing educational or occupational goals (Newman 2012; Sassler et al. 2008). Moreover, these households may have more established precedents for household functioning relative to other doubled-up household types (Harvey 2018). If adult siblings do not increase household stress, they may not negatively affect children’s home environments. Moreover, coresidence may be beneficial if the adult sibling is attentive to the younger child’s needs (Prime et al. 2014) or provides oversight or positive role modeling.

Although studies have often focused on intergenerational relationships, coresidence with other extended family members and/or nonrelatives is common as well. As of 2007–2009, about 6% of children under age 14 who lived with a parent(s) resided in a non-intergenerational household with extended family and/or non-kin (Kennedy and Fitch 2012). Adults sharing households with non-intergenerational extended kin are more likely to share household expenses than adults living in intergenerational households (Reyes 2018). Households shared with extended kin and non-kin may also have more disagreement over household economic arrangements and lower levels of household solidarity compared with multigenerational households (Harvey 2018). Likewise, within-household income inequality is associated with greater instability for doubled-up households formed with extended family or non-kin but not those formed with parents or adult children (Glick and Van Hook 2011). These findings suggest that compared with intergenerational coresidence, sharing a household with extended kin or non-kin may be relatively transactional and residents may be less invested in the well-being of other household members, which might make them less supportive environments for children. Additionally, although extended kin/non-kin household members might be less invested in children compared with coresident grandparents, parents who are doubled up in extended kin/non-kin households still experience interference in parenting decisions and diminished control over the home environment (Harvey 2020).

Despite these potential differences, studies of the effects of non-intergenerational households generally consider doubled-up households as a single category. Some studies have categorized households based on all coresident adults (kin and non-kin), and others have categorized based only on coresident kin (excluding non-kin), but extant research typically groups multigenerational and all non-multigenerational doubled-up households together (Ahrentzen 2003; Aquilino 1996; Entwisle and Alexander 1996; Kang and Cohen 2017; Park et al. 2011; Thompson et al. 1992). One exception is Mollborn et al. (2011), who examined the effects of doubled-up households formed with children’s grandparents versus those formed with other adults on cognitive scores and behavioral outcomes at age 2. Their results suggested that for most children, coresidence with grandparents is associated with better early childhood outcomes, particularly cognitive outcomes, than coresidence with nongrandparent adults.

Prior research on the effects of doubled-up households, regardless of type, has generally focused on cognitive and behavioral outcomes in childhood or adolescence (Augustine and Raley 2013; Dunifon and Kowaleski-Jones 2007; Leadbeater and Bishop 1994; Mollborn et al. 2011). Although the economic and social consequences of household composition suggest that it may influence children’s long-term outcomes, we know little about how enduring the effects of doubling up in childhood are. If the effects of childhood doubling up persist into adulthood, these households may play a role in the intergenerational transmission of disadvantage.

Dynamic Selection Into Household Types

As discussed, household instability makes it important to study doubling up longitudinally. Short-term measures of household composition have two primary limitations. First, static measures compare children who were doubled up during the survey, many of whom transition out of such households, with children who were not doubled up, although many double up at some point during childhood. Thus, these studies likely underestimate the impact of doubling up. Moreover, studies that use static measures estimate effects for children who were doubled up for a short duration with those doubled up for a long period. The mechanisms through which household composition may affect children—including changes in resources and availability of caregivers—likely have larger impacts over time. Thus, measuring duration in household types, rather than point-in-time residence, is key to understanding their cumulative effects.

To accurately estimate the effects of childhood years in different household types, I use marginal structural models and IPTW (Robins et al. 2000). Understanding cumulative effects of household composition requires capturing direct and indirect effects, but when time-varying characteristics both predict and are predicted by the independent variable—here, household composition—static models provide biased estimates of total effects. Many factors that predict household type also likely mediate the relationship between household composition and children’s outcomes. For example, individuals who become unemployed are more likely to double up (Wiemers 2014), and doubling up is positively associated with entering the workforce for mothers (Hao and Brinton 1997). Because maternal employment is one pathway through which doubling up may affect children, controlling for mother’s employment throughout childhood would “control away” this pathway and not capture the total effect of doubling up. Conversely, controlling only for maternal employment before the child was born would capture indirect effects of doubling up but also remain vulnerable to omitted variable bias. IPTW addresses the problem of time-varying confounders, such as maternal employment, by weighting each child by the inverse of the predicted probability that they would be in the series of household structures in which they were observed.

IPTW appropriately accounts for observed time-varying confounders, such as maternal employment, but does not solve issues resulting from unmeasured confounders. Thus, accurately modeling selection into doubled-up households is important, just as it would be with conventional regression methods. Factors that affect children’s residence in different household types and their young adult health and educational attainment can bias the estimates if not included in the IPTWs; for example, the upcoming section on limitations describes maternal depression as one potential unobserved confounder in this analysis. However, unlike other methods, IPTW provides unbiased estimates of the total effects of household type if selection is correctly modelled. Conventional regression methods, in contrast, require the additional assumption that household composition does not affect future values of time-varying confounders.

Data and Methods

I use data from the National Longitudinal Survey of Youth 1979 (NLSY79) and Child and Young Adult cohort (NLSY79-CYA). The NLSY79 surveyed more than 12,600 Americans, oversampling Hispanic and African American respondents, to create a nationally representative sample of men and women ages 14 to 21 in 1979. The NLSY79-CYA includes all children born to NLSY79 mothers and is representative of about 95% of all children born to this cohort of women (U.S. Bureau of Labor Statistics n.d.). In 2012, the young adult response rate was more than 80% (National Longitudinal Surveys n.d.). The NLSY is one of few surveys to follow children from birth through young adulthood, and the large sample size and extensive covariates available from maternal and child interviews make these data ideal for this analysis.

To study household structure throughout childhood and outcomes at age 20, I restrict my sample to children born between 1979 and 1995, about 80% of the original sample. Roughly one-half of the omitted births occurred before 1979 (to mothers under age 21), and one-half occurred after 1995 (to mothers over age 29). Of the children in my sample, 2,576 were lost to follow-up, and an additional 364 are missing measures on at least one outcome of interest, bringing the final sample size to 6,315. The NLSY79 was fielded annually from 1979 to 1994, and biennially since 1994. For years without a survey, I use values from the previous year; that is, rather than assuming children’s characteristics stay constant over a one-year period, I assume that they stay constant over two years.1 For other missing data, I use multiple imputation.

My treatment variables, created from maternal household roster data of all individuals who usually reside in the household, measure cumulative exposure to different household types from ages 1 to 17. Household type at birth is incorporated into my models as a baseline confounder and is not used to estimate the effects of household types (Wodtke et al. 2011). I consider children doubled up if they live with their mother in a household with at least one adult age 21 or older other than their mother and mother’s romantic partner (similar to Eggers and Moumen 2013).2 I classify shared households into three types, depending on whether the child and their mother are living with the mother’s or mother’s romantic partner’s: (1) parent/grandparent, (2) adult child/grandchild, or (3) other extended family or non-kin adult. Additionally, I include two nonshared household categories: (4) with mother in nonshared household, and (5) in any household without the mother. For doubled-up households with multiple additional adults, I assign children to the first household type for which they are eligible. For example, if a mother coresides with her mother and adult sister, I include her child in the multigenerational category (1). This ordering of doubled-up households is consistent with previous research (Glick and Van Hook 2011), and it reflects the additional adult that I expect to be most involved in the child’s life. Because this study focuses on the relationship of other adults to the mother, two nonsibling children in the same household may be classified into different household types. Qualitative data suggest that doubled-up children may receive differential treatment based on their familial relationships with household members (Harvey 2018).

Finally, although I include an indicator for how many years the child spent in a household without their mother, this estimate is not a focus of this research. I include this category to retain in my sample children who spent some of childhood in a household without their mother, such as while living with their father or other relatives, away at school, or on their own. However, my data, which come primarily from mothers, are poorly suited for studying the effects of nonmaternal households. Results for this group are presented in the tables but are not discussed in text.

My outcome variables measure young adult educational attainment and health. I consider whether the child, at age 20, (1) graduated high school; (2) attended college; (3) smoked in the past month; (4) has symptoms of depression, measured by a score of eight or higher on the CES-D-SF (CESD-R n.d.; Levine 2013); and (5) is obese, measured by a self-reported height and weight corresponding to a BMI greater than 30. If data at age 20 are unavailable, I accept measures from ages 19, 21, or 22. Because depressive symptoms data are not available for all interviewees in 2000 and 2002, I also accept measures from ages 18 and 23 for children born in 1979 or 1980.

Predictors

To estimate IPTWs, I predict household type from a multinomial logit model. Previous research has shown the importance of demographics, economic factors, and childcare needs in mothers’ likelihood of doubling up (Pilkauskas 2012; Sigle-Rushton and McLanahan 2002). Table 1 provides a summary of included covariates. Table 2 provides weighted descriptive statistics.

Demographic Factors

To capture demographic factors associated with doubling up, I include indicators for child’s race (Hispanic; non-Hispanic Black; non-Black, non-Hispanic) and whether the mother was born in the United States. I also include child’s sex, birth year, and age. Time-varying indicators measure whether the family lives in an urban area and region of residence (South, North Central, West, Northeast). To capture mothers’ social origins, I include indicators for the highest educational attainment of her parents (less than high school, high school, some college, 4+ years of college).

Economic Need

To reflect economic need, I include measures of total income of the mother and, if married, her spouse from wages and salary, business/farm income, and/or military income in the previous calendar year (in $10,000s). This measure is adjusted to 2014 dollars and is top-coded at the 95th percentile. I include an indicator for whether the mother received any welfare, including AFDC/TANF, food stamps, and/or SSI in the past calendar year and, if so, a measure of total welfare income (in $10,000s), adjusted to 2014 dollars and top-coded at the 95th percentile. Because doubling up can be a response to unemployment, I include an indicator for whether the mother reported that she or her spouse received unemployment income or that she was ever unemployed during the current calendar year. To further capture mothers’ earning potential, I include her 1979 Armed Forces Qualifying Test (AFQT) score percentile, a measure of cognitive achievement, and time-varying indicators for educational attainment (less than high school, high school, some college, four or more years of college).

Because members of the armed forces may receive housing, I include an indicator of whether the mother or her spouse received income from military service in the past year. Similarly, I include measures of whether the mother’s family lived in public housing or received a government rent subsidy in the past year and whether she or her spouse owns or is buying their home. These variables capture the availability of housing options. Additionally, military housing, public/subsidized housing, and landlords may impose occupancy rules, making it more difficult to double up. Finally, to capture residential instability that may be predictive of temporary housing arrangements, I include a variable for the total number of previous residential moves.

Childcare Needs

Mothers with greater childcare needs may be more likely to double up. To capture factors that could limit mothers’ abilities to care for children, I include variables for age at the birth of the child and her last measured Rosenberg self-esteem score before the child was born. As measures of child and maternal health, I include (1) a time-invariant indicator for whether the child or a sibling was low birth weight and (2) a time-varying measure of whether health limits the mother’s ability to work. An indicator for whether the mother reports having used cocaine/crack cocaine 10 or more times in her lifetime serves as a rough measure of drug use. I also include an indicator of whether the mother reports binge drinking (6+ drinks on a single occasion) in the past month. Because data on self-esteem, drug use, and binge drinking were gathered too inconsistently to be included as time-varying covariates, I use the last observation before the birth of the child. To account for time demands, which may affect childcare needs, I include time-varying indicators of mother’s employment status in the past calendar year (full-time, part-time, not employed) and whether she was enrolled in school.

Childcare needs are also influenced by the age and number of children. I include time-varying measures of how many biological, adopted, or stepchildren the mother has in the household and the age of her youngest child. Because romantic relationship status influences child-rearing assistance needs, I include a time-varying indicator for whether the mother is married and, for unmarried mothers, an indicator for the presence of a cohabiting partner. I measure relationship status changes with two indicators for whether the mother gained or lost a spouse/cohabiting partner between the previous and current survey wave. I account for household instability driven by adults other than romantic partners with a variable for the total number of previous transitions between household types. This measure excludes transitions into doubling up that are a result of an adult sibling aging into adulthood (rather than newly joining the household).

Inverse Probability of Treatment Weights

Following previous research (Sharkey and Elwert 2011; Wodtke et al. 2011), I use stabilized IPTWs. To construct IPTWs, I predict the child’s household type using multinomial logit models. For each child (i), the probability of treatment is the product of the year-specific probabilities of being in the household type in which they were observed from ages 1 to 17. The year-specific (k) predicted probabilities of a child being in the household in which they were observed (Aik) are based on household type (Ai(k − 1)) and time-varying covariates L¯k1 measured in the previous year, as well as time-invariant covariates and baseline values of time-varying covariates (L¯0). This product is the denominator of the stabilized weight. The numerator follows the same form but excludes time-varying predictors.
SWi=k=117PAk=akiAk1=ak1iL¯0=l0ik=117PAk=akiAk1=ak1iL¯k1=lk1iL¯0=l0i.
Following convention, I construct attrition weights to address possible nonrandom attrition from the sample (Lee and McLanahan 2015; Wodtke et al. 2011). These weights follow the same form as the stabilized IPTWs but adjust for children’s probability of remaining in the sample through age 19. I multiply the IPTWs and attrition weights to produce the final weights for the outcome models. To reduce the variance and lessen the influence of highly weighted observations, I top- and bottom-code the weights, respectively, at the 1st and 99th percentiles (Cole and Hernán 2008), producing a final weight with a mean of 1.06 and standard deviation of 1.38.

Marginal Structural Model Using IPTW

I estimate a series of logit models in which each outcome—high school graduation, college attendance, smoking, depression, and obesity—is a function of cumulative exposure to each household type from ages 1 through 17. In the following equation, the log odds ratios δ1– δ4 are the estimated impact of spending one additional childhood year in a given household type (multigenerational coresidence, adult sibling coresidence, extended kin/non-kin coresidence, or without mother) on the log odds of experiencing the outcome.
logitIPT-weightedPYi=1=θ0+δ1k=117a1ik+δ2k=117a2ik+δ3k=117a3ik+δ4k=117a4ik+γ2L¯i0.
Using stabilized IPTWs requires that the outcome models condition on time-invariant and baseline covariates (L¯i0) for doubling up to be unconfounded with these traits (Wodtke et al. 2011). For all models, I cluster standard errors at the mother level to account for nonindependence of observations from siblings. Because they do not account for the IPT weighting, the standard errors for the IPT-weighted outcome models are too large, resulting in conservative tests for statistical significance (Wodtke et al. 2011).

Results

Prevalence and Instability of Doubled-Up Households

Table 3 reports the proportion of children who experienced each household type, weighted to be representative of children born to NLSY79 mothers. These results show that living doubled up is a common childhood experience: 45.1% of children double up at some point from ages 1 through 17. The cumulative prevalence of doubling up underscores the importance of longitudinal measures of household composition. Although nearly one-half of children double up at some point, just slightly more than 10% of childhood years are spent doubled up, so point-in-time measures would miss many previous and future instances of doubling up.

Multigenerational households are the most common form of doubling up, with more than one-fifth of children experiencing this household type between ages 1 and 17. These households are more common when children are younger (cf. Pilkauskas 2012); 17.7% of children lived in multigenerational households at some point between ages 1 and 5, but only 7.6% did between ages 12 and 17. Approximately 19% of children lived in an adult sibling household, most commonly later in childhood. Finally, 15.3% of children lived in an extended family/non-kin household. As with multigenerational households, these arrangements were most common in early childhood; about 9% of children lived in an extended kin/non-kin household at some point from ages 1 to 5, whereas just 4% did from ages 12 to 17. Many children lived in multiple types of doubled-up households over childhood, with particular overlap in the children who experienced multigenerational and extended kin/non-kin households. More than 40% of children who ever lived in an extended kin/non-kin household also lived in a multigenerational household, a rate nearly twice that of children who never lived in an extended kin/non-kin household.

Table 4 shows the average number of years spent in each doubled-up household type, from ages 1 through 17, for children who ever experienced the household type. On average, children who double up spend a total of 3.9 years in these households.3 Children who double up in extended kin/non-kin households spend an average of 2.43 years in this household type. Similarly, children who live with adult siblings spend an average of 2.16 years in such households. Children tend to spend more time in multigenerational households than in other doubled-up household types. Children who live in a multigenerational household spend an average of 4.25 childhood years in these households.

These averages conceal considerable variation in the time children spend in each household type, especially for multigenerational homes. Although nearly 30% of children who live in multigenerational households spend a year or less in these households, more than one-fourth spend six or more childhood years in these households. Adult sibling and extended family/non-kin households are more consistently short-lived: more than 40% of children who experience these household types live in such households for a year or less. The variation in number of childhood years spent doubled up further demonstrates the importance of longitudinal measures of household composition.

Given the instability of doubled-up households, childhood years spent doubled up are not necessarily consecutive. Children who live in multigenerational and extended family/non-kin households are particularly likely to cycle in and out of different household types. Nearly one-fifth of children who ever live in multigenerational households and approximately one-fourth of children who ever live in extended family/non-kin households experience five or more transitions between household types during childhood. In contrast, approximately three-fourths of children who experience an adult sibling household experience two or fewer household type transitions (excluding transitions caused by coresident siblings aging into adulthood).

Household Type Prediction Model

Table 5 presents the prediction model used to produce the denominator of the IPTWs: a multinomial logit model predicting household type, with living in a nonshared household with a mother as the reference category. This model includes variables that may confound the relationship between household type and young adult outcomes (including both baseline and k – 1 values of time-varying characteristics) but is not designed to isolate the effect of any specific variable. Table A1 (online appendix) shows the numerator prediction model.

Net of family characteristics and household type at baseline, previous household type predicts current household type. Living in any doubled-up household in one wave is associated with higher odds of living in the same household type the following wave, relative to living non–doubled up. Moreover, residence in any doubled-up household type is associated with heightened odds of doubling up in another type in the following wave, relative to not being doubled up.

Income, housing options, and race are associated with living in a doubled-up household. For those who receive welfare income, the amount received is negatively associated with subsequent residence in multigenerational and adult sibling households, all else being equal. Additionally, earnings at baseline are negatively associated with residence in extended kin/non-kin households. When all other covariates and housing status at baseline are controlled for, residence in subsidized housing in one wave is negatively associated with living in a multigenerational or extended family/non-kin household in the subsequent wave, and homeownership is negatively associated with living in a multigenerational home. Finally, race remains significantly associated with doubling up, even when other characteristics are controlled for. Compared with being White, being Black or Hispanic is associated with higher odds of living with an adult sibling relative to living non–doubled up, and being Hispanic is associated with higher odds of living in an extended kin/non-kin household relative to living non–doubled up.

Compared with having an unmarried mother, having a married mother is associated with lower odds of subsequent residence in multigenerational and extended kin/non-kin households, with other family characteristics and marital status at baseline controlled for. Having a cohabiting mother is also associated with lower odds of living in multigenerational and extended kin/non-kin households, relative to non–doubled-up households. Additionally, all else being equal, having a mother marry or begin cohabiting is associated with lower odds of living in multigenerational or extended kin/non-kin households compared with living in a nonshared household, whereas having a mother end a marital or cohabiting relationship is associated with higher odds of multigenerational or extended family/non-kin coresidence. Household instability driven by nonromantic partners is also predictive: the number of previous transitions between household types is positively associated with residence in extended kin/non-kin households relative to non–doubled-up households.

Effects of Doubling Up on Young Adult Outcomes

The right-hand columns of Table 6 present results from IPT-weighted outcome models. For comparison, the left-hand columns show results for outcome models that are weighted only by attrition weights and do not account for selection into doubled-up households.

Multigenerational Households

The unadjusted models show significant negative associations between childhood years spent in multigenerational households and educational attainment. However, adjusting for selection attenuates these associations substantially, and the estimated effects from the IPT-weighted models are near 0 and are not statistically significant. Both the unadjusted and IPT-weighted models show relatively little association between multigenerational households and young adult health outcomes. Only the positive coefficient for obesity is significant in the unadjusted models, but the estimate is smaller and statistically insignificant in the IPT-weighted model. Together, these results do not suggest that childhood years in multigenerational households have substantial effects on young adult educational attainment or health after selection into these households is accounted for. These results do not necessarily indicate that grandparent coresidence is inconsequential for children. Indeed, previous research has suggested that multigenerational households may have little effect on children’s outcomes because they shape the home environment in both positive and negative ways. For instance, multigenerational households increase children’s access to caregiver time (Kalil et al. 2014) but may also introduce confusion or conflict between grandparents and parents over parenting rights and responsibilities (Chase-Lansdale et al. 1994; Harvey 2020).

Adult Sibling Households

In the unadjusted models, childhood years spent in adult sibling households are significantly negatively associated with high school graduation, college attendance, and smoking and significantly positively associated with obesity. Accounting for selection into these households changes these associations substantially. The adjusted model shows no significant harmful effects of these households, providing little support for the idea that coresident adult siblings divert parental resources in ways that are harmful for younger siblings. Moreover, an additional year in an adult sibling household is associated with 12% lower odds of smoking compared with an additional year in a non–doubled-up household. The negative association between years in adult sibling households and smoking, with no significant associations between these households and other young adult outcomes, may suggest that coresidence with adult siblings is most beneficial in reducing risky behavior, perhaps by increasing the oversight or positive role modelling that children receive.

Extended Kin/Non-Kin Households

The unadjusted models show negative associations between childhood years spent in extended kin/non-kin households and educational attainment and positive associations with the adverse young adult health outcomes of smoking and depression. Although accounting for selection into extended kin/non-kin households attenuates most of these associations, the coefficients for high school graduation, college attendance, and obesity are statistically significant in the IPT-weighted models. In the IPT-weighted models, an additional year in an extended kin/non-kin household is associated with 9% lower odds of high school graduation and 8% lower odds of college attendance, relative to an additional year in a nonshared household. An additional year in an extended kin/non-kin household is associated with 11% higher odds of obesity. The finding that years spent in extended kin/non-kin households negatively affect children’s young adult well-being may be driven by high levels of conflict within these households (Harvey 2018), which may increase parental and child stress and decrease household functioning. Additionally, these results may reflect how doubling up can intensify the negative effects of household resource constraints, particularly with extended kin/non-kin household members whose coresidence may be more transactional and who may not prioritize children’s well-being (see Glick and Van Hook 2011; Reyes 2018).

Limitations and Directions for Future Research

Doubled-up households represent a diverse group. This study improves on prior research by distinguishing between three doubled-up household types based on familial and nonfamilial relationships. However, doubled-up households can be classified in other theoretically relevant ways. I conducted supplemental analyses to consider some alternative definitions and categorizations. Section 4 of the online appendix shows that the estimates for adult sibling households are sensitive to the age cutoff used to identify doubled-up adults. Limiting adult siblings to those aged 24 or older suggests some positive effects of adult sibling coresidence, similar to the main analysis (which uses an age cutoff of 21), while including adult siblings aged 18 or older produces no significant associations. Data limitations prevent me from comparing coefficients across models, but future research should consider whether the effects of coresidence with adult siblings may vary by the age of the sibling. Perhaps more mature adult siblings offer greater oversight or emotional support for minor children compared with younger adult siblings. Additionally, because more than one-fourth of multigenerational households include an extended family/non-kin household member as well, I also reran the analysis including two multigenerational household categories: one for multigenerational households that also include a nongrandparent extended family member/nonrelative, and one for those that do not include any nongrandparent extended family members/nonrelatives. Section 5 of the online appendix shows that when analyzed as a separate category, households with both a grandparent and another extended kin/non-kin adult generally have estimated effects that fall between those for multigenerational-only and extended kin/non-kin households. Finally, Table A3 in the online appendix shows that stratifying by mother’s marital status at baseline does not suggest substantial differences between children born to married and unmarried mothers. However, limited power prevents me from drawing strong conclusions from these results. Data limitations also prevent me from identifying differences between households shared with extended kin and those shared with non-kin and from identifying how the effects of household types may vary by child’s race or age when doubled up. Future research should ask how the effects of doubling up may differ along these dimensions.

Additionally, my data come from children born between 1979 and 1995 to a sample of mothers who were 14–22 years old and living in the United States in 1979; they may not reflect the experiences of more recent cohorts. Doubling up, particularly in multigenerational households, has become more common in recent decades, driven in part by changes in the racial/ethnic composition of children and mothers’ marital status; moreover, multigenerational doubling up has seen particular growth among Hispanic children and children with married, older, and more highly educated mothers (Pilkauskas and Cross 2018). Future research, drawing on multiple birth cohorts, should consider whether doubled-up households’ effects may change over time as the doubled-up population grows and the composition of this population changes.

This analysis remains subject to other important limitations. IPTW does not address bias from unobserved confounders, and although the NLSY provides an extensive list of covariates, it does not contain all factors relevant to doubling up. For example, depression might lead mothers to seek coresidential support, but maternal depression data are not consistently available. However, depression is associated with self-esteem (Baumeister et al. 2003), which is measured at baseline, and should also be partially captured by the time-varying indicator of whether health limits the mother’s ability to work. The differences between coefficients in the unadjusted and IPT-weighted models provide some evidence that the IPTWs are accounting for selection into doubled-up household types. Additionally, following Lee and Jackson (2017), I reran the prediction model used to produce the denominator of the weights with the sample weighted by the IPT weights. If IPT weighting successfully achieves balance, household type should have little association with the measured time-varying confounders except by chance. (Recall that the stabilized IPTWs do not adjust for time-invariant confounders, which is why I include these variables in the outcome models presented in Table 6.) In the IPT-weighted model, only slightly more of the associations are significant than we would expect by chance, and nearly all are smaller in magnitude compared with the unweighted model (see Table A2, online appendix). These results provide some evidence that IPT weighting is largely successful in achieving balance. However, omitted variables remain an important consideration.

Additionally, the NLSY does not identify which household member holds the lease/mortgage to the home. Previous research has shown that when adults live with their parents, the older generation tends to be the householder, suggesting that mothers are generally householders in adult sibling households, and grandparents are generally householders in multigenerational households (Cohen and Casper 2002; Maroto 2017; White 1994). Whether their mother is the householder or lives in someone else’s home shapes children’s experiences (Harvey 2020), so examining how the effects of doubling up differ by mother’s householder status is an important question for future research.

Finally, as previously described, instability makes studying doubled-up households from a dynamic perspective vital. Instability is also likely a mechanism through which doubled-up households may affect children; like family instability, household instability has negative effects on child well-being (Perkins 2019). This study focuses on estimating the total effect of doubling up, rather than the effects of instability in these households, and I include previous transitions between household types as a predictor of subsequent household type. Although children rarely remain doubled up throughout childhood, there is substantial variation in how long children spend doubled up. Examining the relative stability of different doubled-up household types and how instability shapes their effects on children is another important direction for future research.

Conclusion

Prior studies of children’s household composition have typically focused on estimating the relationship between doubling up—often measured at a single point in time—and childhood outcomes. By examining the cumulative effects of doubling up throughout childhood, this study extends this line of research and shows that childhood household composition can have enduring impacts on young adult well-being. To examine the long-term effects of doubling up, I draw on longitudinal data that include household composition throughout childhood. The results demonstrate the importance of a longitudinal approach: I find that nearly one-half of children double up at some point from ages 1 to 17. By accounting for household composition throughout childhood, rather than at one point in time, this study reflects an understanding that households are dynamic and that children’s lives are shaped by the sum of their childhood environments.

I use IPTW and marginal structural models to estimate the cumulative impact of years spent in different doubled-up household types. By employing methods that capture both direct and mediated effects, this study takes seriously the life course theory premise that early environments affect later outcomes, both directly and indirectly through their effects on later environments (Elder 1998). Yet unlike other methods that capture full effects, these methods also account for dynamic selection into doubled-up household types, allowing for the possibility that the same characteristics that predict household composition are also affected by household composition. I find that selection accounts for the associations between childhood residence in multigenerational and adult sibling households and worse young adult outcomes. However, extended kin/non-kin households’ negative associations with educational attainment and positive association with obesity are significant after adjustment for selection into these households. These results underscore the importance of rigorous methods for distinguishing between selection and causal effects while still capturing both direct and indirect impacts of childhood environments.

These findings have implications for our conceptualization of family complexity. That coresidence with adults other than parents and parents’ romantic partners influences children’s long-term outcomes suggests that focusing exclusively on the nuclear family—defined by parents, romantic partners, and minor children—is too limited. For children living with at least one parent, coresidence with grandparents, extended family, and non-kin is more common than residence in either cohabiting or stepfamily households (Kennedy and Fitch 2012), and I find that some doubled-up household types have enduring effects on children’s life chances.

Although qualitative work has suggested that doubled-up household types vary in the environments they create for children, quantitative research has generally grouped all doubled-up households together or examined only multigenerational households. In this study, I estimate the impact of three doubled-up household types: those formed with children’s grandparents, adult siblings, and other extended family or non-kin. I find that childhood years spent in extended family/non-kin households are detrimental for young adult outcomes, and years spent in adult sibling households may be beneficial, but I find no indication that that multigenerational coresidence has a lasting effect. These results provide evidence against conceptualizing doubled-up households as a uniform category. Future research should adopt a relational understanding of doubled-up households and continue disentangling how coresidence with different adults shapes children’s lives. Because of this study’s limitations in distinguishing between households formed with extended family and non-kin, future research should work to identify the role of (non) familial ties in shaping doubled-up households’ effects on children.

Given evidence of a link between doubling up and children’s long-term outcomes, future research should investigate potential mechanisms. I find that adult sibling households have significant negative associations with smoking, suggesting that adult siblings may contribute oversight or positive role modeling to younger siblings. For extended kin/non-kin households, I find evidence of detrimental effects on educational attainment and obesity but not depression or smoking. These results are interesting in light of the literature on family structure; studies on father absence have consistently found negative effects for adult mental health and substance use, including smoking, but there is less evidence of effects on cognitive development (McLanahan et al. 2013). These divergent results may reflect differences in how or the degree to which family structure and household composition affect children, underscoring the need for future research that examines household composition in concert with family structure.

Regardless of the mechanisms at work, the negative effects of extended family/non-kin households identified in this study are troubling given large recent increases in multifamily households. The number of households with unrelated subfamilies more than tripled between 2003 and 2009 (Eggers and Moumen 2013), and this study raises concerns about how this household type may influence children’s lives. However, young adults living in their natal home also increased substantially in recent years (Eggers and Moumen 2013). My findings suggest that adult children living with their parents tend not to create harmful environments for minor children. Although more research is needed on how children experience coresidence with adult siblings, these results are reassuring given the increasingly extended transition to adulthood.

Acknowledgments

I received helpful feedback on this project from Alexandra Killewald, Devah Pager, Kathryn Edin, Mario Small, Brielle Bryan, Alexandra Feldberg, Barbara Kiviat, Katherine Morris, Margot Moinester, Kelly Musick, Kristin Perkins, Adriana Reyes, Alix Winter, Xiaolin Zhou, and Jonathan Spader. This research was supported in part by fellowships from the Harvard Joint Center for Housing Studies, the Multidisciplinary Program in Inequality and Social Policy, and the Radcliffe Institute for Advanced Study at Harvard University.

Notes

1

The annual/biennial survey design is a limitation of this analysis. Assuming that children are not doubled up in years when there is no survey unless their household type is the same is the previous and subsequent surveys produces substantively similar results. Using only biennial data produces similar results for multigenerational and extended kin/non-kin households but no significant associations between adult sibling households and young adult outcomes.

2

The limitations section and section 4 of the online appendix discuss how the results change with age cutoffs of 24 and 18, respectively.

3

The annual/biennial surveys likely miss some short-term doubled-up households. By excluding shorter-duration households, I may underestimate the average total number of years spent doubled up. Alternatively, by assuming that each spell lasts at least a full year, I may overestimate these averages.

References

References
Ahrentzen, S. (
2003
).
Double indemnity or double delight? The health consequences of shared housing and “doubling up.”
Journal of Social Issues
,
59
,
547
568
.
Aquilino, W. S. (
1996
).
The life course of children born to unmarried mothers: Childhood living arrangements and young adult outcomes
.
Journal of Marriage and the Family
,
58
,
293
310
.
Astone, N. M., & Washington, M. L. (
1994
).
The association between grandparental coresidence and adolescent childbearing
.
Journal of Family Issues
,
15
,
574
589
.
Augustine, J. M., & Raley, R. K. (
2013
).
Multigenerational households and the school readiness of children born to unmarried mothers
.
Journal of Family Issues
,
34
,
431
459
.
Baumeister, R. F., Campbell, J. D., Krueger, J. I., & Vohs, K. D. (
2003
).
Does high self-esteem cause better performance, interpersonal success, happiness, or healthier lifestyles?
.
Psychological Science in the Public Interest
,
4
(
1
),
1
44
.
Berlin, G., Furstenberg, F. F.Jr, & Waters, M. C. (
2010
).
Introducing the issue: Transition to adulthood
.
The Future of Children
,
20
(
1
),
3
18
.
Black, M. M., & Nitz, K. (
1996
).
Grandmother co-residence, parenting, and child development among low income, urban teen mothers
.
Journal of Adolescent Health
,
18
,
218
226
.
Center for Epidemiologic Studies Depression Scale Revised (CESD-R)
. (n.d.).
CESD-R explanation
. Retrieved from http://cesd-r.com/cesdr/
Chase-Lansdale, P. L., Brooks-Gunn, J., & Zamsky, E. S. (
1994
).
Young African-American multigenerational families in poverty: Quality of mothering and grandmothering
.
Child Development
,
65
,
373
393
.
Clampet-Lundquist, S. (
2003
).
Finding and keeping affordable housing: Analyzing the experiences of single-mother families in north Philadelphia
.
Journal of Sociology and Social Welfare
,
30
(
4
),
123
140
.
Cohen, P. N., & Casper, L. M. (
2002
).
In whose home? Multigenerational families in the United States, 1998–2000
.
Sociological Perspectives
,
45
,
1
20
.
Cole, S. R., & Hernán, M. A. (
2008
).
Constructing inverse probability weights for marginal structural models
.
American Journal of Epidemiology
,
168
,
656
664
.
DeLeire, T., & Kalil, A. (
2002
).
Good things come in threes: Single-parent multigenerational family structure and adolescent adjustment
.
Demography
,
39
,
393
413
.
Desmond, M. (
2012
).
Disposable ties and the urban poor
.
American Journal of Sociology
,
117
,
1295
1335
.
Domínguez, S., & Watkins, C. (
2003
).
Creating networks for survival and mobility: Social capital among African-American and Latin-American low-income mothers
.
Social Problems
,
50
,
111
135
.
Dunifon, R., & Kowaleski-Jones, L. (
2007
).
The influence of grandparents in single-mother families
.
Journal of Marriage and Family
,
69
,
465
481
.
Dunifon, R. E., Ziol-Guest, K. M., & Kopko, K. (
2014
).
Grandparent coresidence and family well-being: Implications for research and policy
.
Annals of the American Academy of Political and Social Science
,
654
,
110
126
.
Edin, K., & Shaefer, H. L. (
2015
).
$2.00 a day: Living on almost nothing in America
.
Boston, MA
:
Houghton Mifflin Harcourt
.
Eggers, F. J., & Moumen, F. (
2013
).
Analysis of trends in household composition using American housing survey data (Report)
.
Washington, DC
:
U.S. Department of Housing & Urban Development
.
Elder, G. H.Jr (
1998
).
The life course as developmental theory
.
Child Development
,
69
,
1
12
.
Entwisle, D. R., & Alexander, K. L. (
1996
).
Family type and children’s growth in reading and math over the primary grades
.
Journal of Marriage and the Family
,
58
,
341
355
.
Fingerman, K. L., Cheng, Y-P, Wesselmann, E. D., Zarit, S., Furstenberg, F., & Birditt, K. S. (
2012
).
Helicopter parents and landing pad kids: Intense parental support of grown children
.
Journal of Marriage and Family
,
74
,
880
896
.
Furstenberg, F. F.Jr (
2010
).
On a new schedule: Transitions to adulthood and family change
.
Future of Children
,
20
(
1
),
67
87
.
Glick, J. E., & Van Hook, J. (
2011
).
Does a house divided stand? Kinship and the continuity of shared living arrangements
.
Journal of Marriage and Family
,
73
,
1149
1164
.
Goodman, C., & Silverstein, M. (
2002
).
Grandmothers raising grandchildren: Family structure and well-being in culturally diverse families
.
The Gerontologist
,
42
,
676
689
.
Gordon, R. A., Chase-Lansdale, L. P., & Brooks-Gunn, J. (
2004
).
Extended households and the life course of young mothers: Understanding the associations using a sample of mothers with premature, low birth weight babies
.
Child Development
,
75
,
1013
1038
.
Hao, L., & Brinton, M. C. (
1997
).
Productive activities and support systems of single mothers
.
American Journal of Sociology
,
102
,
1305
1344
.
Harvey, H. (
2018
,
August
).
Economic exchange and relational work within doubled-up households
.
Paper presented at the annual meeting of the American Sociological Association
,
Philadelphia, PA
.
Harvey, H. (
2020
).
When mothers can’t “pay the cost to be the boss”: Roles and identity within doubled-up households
.
Social Problems
(in press).
Kalil, A., Ryan, R., & Chor, E. (
2014
).
Time investments in children across family structures
.
Annals of the American Academy of Political and Social Science
,
654
,
150
168
.
Kang, J., & Cohen, P. N. (
2017
).
Extended kin and children’s behavioral functioning: Family structure and parental immigrant status
.
Social Science & Medicine
,
186
,
61
69
.
Kennedy, S., & Fitch, C. A. (
2012
).
Measuring cohabitation and family structure in the United States: Assessing the impact of new data from the Current Population Survey
.
Demography
,
49
,
1479
1498
.
Leadbeater, B. J., & Bishop, S. J. (
1994
).
Predictors of behavior problems in preschool children of inner-city Afro-American and Puerto Rican adolescent mothers
.
Child Development
,
65
,
638
648
.
Lee, D., & Jackson, M. (
2017
).
The simultaneous effects of socioeconomic disadvantage and child health on children’s cognitive development
.
Demography
,
54
,
1845
1871
.
Lee, D., & McLanahan, S. (
2015
).
Family structure transitions and child development: Instability, selection, and population heterogeneity
.
American Sociological Review
,
80
,
738
763
.
Levine, S. Z. (
2013
).
Evaluating the seven-item Center for Epidemiologic Studies depression scale short-form: A longitudinal U.S. community study
.
Social Psychiatry and Psychiatric Epidemiology
,
48
,
1519
1526
.
Maroto, M. (
2017
).
When the kids live at home: Coresidence, parental assets, and economic insecurity
.
Journal of Marriage and Family
,
79
,
1041
1059
.
McLanahan, S., & Percheski, C. (
2008
).
Family structure and the reproduction of inequalities
.
Annual Review of Sociology
,
34
,
257
276
.
McLanahan, S., Tach, L., & Schneider, D. (
2013
).
The causal effects of father absence
.
Annual Review of Sociology
,
39
,
399
427
.
Mollborn, S., Fomby, P., & Dennis, J. A. (
2011
).
Who matters for children’s early development? Race/ethnicity and extended household structures in the United States
.
Child Indicators Research
,
4
,
389
411
.
Mollborn, S., Fomby, P., & Dennis, J. A. (
2012
).
Extended household transitions, race/ethnicity, and early childhood cognitive outcomes
.
Social Science Research
,
41
,
1152
1165
.
Mutchler, J. E., & Baker, L. A. (
2009
).
The implications of grandparent coresidence for economic hardship among children in mother-only families
.
Journal of Family Issues
,
30
,
1576
1597
.
Mykyta, L., & Macartney, S. (
2012
).
Sharing a household: Household composition and economic well-being: 2007–2010
(Current Population Report No. P60-24).
Washington, DC
:
U.S. Census Bureau
.
National Longitudinal Surveys
. (n.d.).
Retention
. Retrieved from https://www.nlsinfo.org/content/cohorts/nlsy79-children/intro-to-the-sample/retention/page/0/1
Newman, K. S. (
2012
).
The accordion family: Boomerang kids, anxious parents, and the private toll of global competition
.
Boston, MA
:
Beacon Press
.
Park, J. M., Fertig, A. R., & Allison, P. D. (
2011
).
Physical and mental health, cognitive development, and health care use by housing status of low-income young children in 20 American cities: A prospective cohort study
.
American Journal of Public Health
,
101
(
Suppl. 1
),
S255
S261
.
Perkins, K. L. (
2017
).
Household complexity and change among children in the United States, 1984 to 2010
.
Sociological Science
,
4
,
701
724
. 10.15195/v4.a29
Perkins, K. L. (
2019
).
Changes in household composition and children’s educational attainment
.
Demography
,
56
,
525
548
.
Pilkauskas, N. V. (
2012
).
Three-generation family households: Differences by family structure at birth
.
Journal of Marriage and Family
,
74
,
931
943
.
Pilkauskas, N. V., & Cross, C. (
2018
).
Beyond the nuclear family: Trends in children living in shared households
.
Demography
,
55
,
2283
2297
.
Pilkauskas, N. V., Garfinkel, I., & McLanahan, S. S. (
2014
).
The prevalence and economic value of doubling up
.
Demography
,
51
,
1667
1676
.
Prime, H., Pauker, S., Plamondon, A., Perlman, M., & Jenkins, J. (
2014
).
Sibship size, sibling cognitive sensitivity, and children’s receptive vocabulary
.
Pediatrics
,
133
,
e394
e401
. 10.1542/peds.2012-2874
Reyes, A. M. (
2018
).
The economic organization of extended family households by race or ethnicity and socioeconomic status
.
Journal of Marriage and Family
,
80
,
119
133
.
Robins, J. M., Hernán, M.Á., & Brumback, B. (
2000
).
Marginal structural models and causal inference in epidemiology
.
Epidemiology
,
11
,
550
560
.
Sassler, S., Ciambrone, D., & Benway, G. (
2008
).
Are they really mama’s boys/daddy’s girls? The negotiation of adulthood upon returning to the parental home
.
Sociological Forum
,
23
,
670
698
.
Seefeldt, K. S., & Sandstrom, H. (
2015
).
When there is no welfare: The income packaging strategies of mothers without earnings or cash assistance following an economic downturn
.
Russell Sage Foundation Journal of the Social Sciences
,
1
(
1
),
139
158
. 10.7758/RSF.2015.1.1.08
Seltzer, J. A., Lau, C. Q., & Bianchi, S. M. (
2012
).
Doubling up when times are tough: A study of obligations to share a home in response to economic hardship
.
Social Science Research
,
41
,
1307
1319
.
Sharkey, P., & Elwert, F. (
2011
).
The legacy of disadvantage: Multigenerational neighborhood effects on cognitive ability
.
American Journal of Sociology
,
116
,
1934
1981
.
Sigle-Rushton, W., & McLanahan, S. (
2002
).
The living arrangements of new unmarried mothers
.
Demography
,
39
,
415
433
.
Skobba, K., & Goetz, E. G. (
2013
).
Mobility decisions of very low-income households
.
Cityscape
,
15
(
2
),
155
171
.
Swartz, T. T. (
2009
).
Intergenerational family relations in adulthood: Patterns, variations, and implications in the contemporary United States
.
Annual Review of Sociology
,
35
,
191
212
.
Thompson, M. S., Entwisle, D. R., Alexander, K. L., & Sundius, M. J. (
1992
).
The influence of family composition on children’s conformity to the student role
.
American Educational Research Journal
,
29
,
405
424
.
Tosi, M., & Grundy, E. (
2018
).
Returns home by children and changes in parents’ well-being in Europe
.
Social Science & Medicine
,
200
,
99
106
.
Unger, D. G., & Cooley, M. (
1992
).
Partner and grandmother contact in Black and White teen parent families
.
Journal of Adolescent Health
,
13
,
546
552
.
U.S. Bureau of Labor Statistics
. (n.d.).
NLSY79 Child/Young Adult sample design
. Retrieved from https://www.nlsinfo.org/content/cohorts/nlsy79-children/intro-to-the-sample/sample-design
White, L. (
1994
).
Coresidence and leaving home: Young adults and their parents
.
Annual Review of Sociology
,
20
,
81
102
.
Wiemers, E. E. (
2014
).
The effect of unemployment on household composition and doubling up
.
Demography
,
51
,
2155
2178
.
Wodtke, G. T., Harding, D. J., & Elwert, F. (
2011
).
Neighborhood effects in temporal perspective: The impact of long-term exposure to concentrated disadvantage on high school graduation
.
American Sociological Review
,
76
,
713
736
.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary data