A growing literature in family demography examines children's residence in doubled-up (shared) households with extended family members and nonkin. This research has largely overlooked the role of doubling up as a housing strategy, with “hosts” (householders) providing housing support for “guests” living in their home. Yet, understanding children's experiences in doubled-up households requires attention to host/guest status. Using the American Community Survey and Survey of Income and Program Participation, we identify the prevalence of children doubling up as hosts and guests in different household compositions (multigenerational, extended family, nonkin), show how this varies by demographic characteristics, and examine children's patterns of residence across these household types. We find large variation by demographic characteristics. More disadvantaged children have higher rates of doubling up as guests than hosts, whereas more advantaged children have higher rates of doubling up as hosts than guests. Additionally, compared with hosts, guests more often use doubling up as a longer-term strategy; a greater share of guests live consistently doubled up over a three-year period, but those who do transition between household types experience more transitions on average than do hosts. Our findings show the importance of attending to both housing status and household composition when studying children living in doubled-up households.
Scholars have grown increasingly attentive to family complexity, showing how variation from the simple nuclear family has changed the nature of family life (Carlson and Meyer 2014). This research has largely focused on family complexity introduced by parents and their romantic partners and children. Yet as of 2018, more than 15% of U.S. children lived in doubled-up (shared) households with additional adults beyond their siblings, parents, and parents' romantic partners (authors' calculations using data from the 2018 American Community Survey). Although such household complexity remains understudied, a growing body of research has made a compelling case for extending the study of family complexity beyond the nuclear family. This research has found that coresidence with extended family and nonkin is common and increasing in prevalence, contributes substantial instability to children's households, and is associated with child well-being (Cross 2018; Harvey 2020a; Mollborn et al. 2011; Perkins 2019; Pilkauskas and Cross 2018; Raley et al. 2019).
Children's residence in doubled-up households remains only partly understood, however. Prior studies have largely focused on who is living with the child—the presence of extended family or nonkin—without attending to whether the child lives in their own home (i.e., their parent or parent's romantic partner is the mortgage/lease-holder) or someone else's. A child may be doubled up because their family “hosts” another adult in their home or because they are “guests” in someone else's home.
Although the family demography literature rarely distinguishes between hosts and guests, research on doubling up as a housing strategy suggests that children's experiences differ dramatically by host/guest status. Doubled-up households act as private housing safety nets by providing guests an alternative to unaffordable market-rate housing (Edin and Shaefer 2015; Skobba and Goetz 2015). Thus, doubling up often serves different functions for hosts, who provide housing support, and guests, who receive it. Moreover, host/guest status shapes how families experience living doubled-up. Hosts typically maintain authority within the home, whereas guests largely hold subordinate roles (Burton and Clark 2005; Harvey 2020b). Although host/guest status is central to understanding what it means to be doubled up, we know little about how the prevalence and stability of children's doubled-up household residence varies by this dimension, nor do we know whether or how the characteristics of children who host and guest differ.
We use data from the American Community Survey and Survey of Income and Program Participation to distinguish between doubled-up households based on both household composition and host/guest status. We examine six doubled-up household types: those in which the child and their parent are guests in the home of a grandparent, another extended family member, or a nonkin adult; and those in which the child and their parent are hosting a grandparent, another extended family member, or a nonkin adult. We contribute to the literature on household complexity by addressing three sets of research questions. First, what share of children live doubled up in each of these household types, and how has the prevalence of these arrangements changed over time? Our analysis identifies the share of children whose families provide housing to individuals outside the nuclear family and the share of children who live doubled up with no home of their own. Second, how does the prevalence of each doubled-up household type vary by socioeconomic status (measured by maternal education level), childcare needs (measured by child age and maternal marital status), and race/ethnicity? Building on prior research showing that each of these factors is associated with children's rates of doubling up, we identify how these characteristics are differentially associated with hosting and guesting. Finally, what are children's patterns and duration of residence in different types of doubled-up households over time? Our analysis shows that the stability of doubling up varies by both host/guest status and household composition.
Background and Contribution
Importance of Host/Guest Status
Identifying whether children are doubled up as hosts or guests is important for two reasons: (1) it provides insight into their level of housing insecurity, and (2) it sheds light on daily life within the household. First, the function of doubling up may differ by host/guest status; prior research has highlighted how doubling up provides housing assistance to guests, with hosts acting as housing support providers. Families often double up as guests because they cannot afford private market rent (Seefeldt and Sandstrom 2015; Stack 1974) and are unable to access the limited supply of public housing assistance (Leopold 2012). Doubling up as a guest is a common response to forced moves, such as eviction (Desmond 2016), and to income shocks, such as unemployment (Wiemers 2014). Barriers like eviction records and financial constraints can hinder doubled-up guest families' efforts to move into independent housing (Desmond 2016). Moreover, because they can be asked to leave at any time, guests are often precariously housed and at risk of homelessness (Skobba and Goetz 2015). Thus, doubling up often reflects—and likely contributes to—housing insecurity and instability, which can be harmful to child well-being (Desmond 2016; Haynie and South 2005). The role of doubling up as a housing strategy underscores the importance of attending to host/guest status, which often reflects whether the child's family is housing insecure themselves or is providing housing support to others.
Second, host/guest status shapes interpersonal dynamics within the home. Because many lower-income families face difficulty securing housing, host status—and the accompanying ability to evict household members—conveys power in intrahousehold relationships (Harvey 2020b; Welsh and Burton 2016). In doubled-up households, guests typically occupy a subordinate role, with little ability to challenge hosts' rules for the household (Burton and Clark 2005; Clampet-Lundquist 2003; Harvey 2020b). With limited authority within the home, guest parents struggle to set rules and routines for their children and, in some cases, cannot regulate the safety of the home environment (Edin and Shaefer 2015; Harvey 2020b). Furthermore, this arrangement can take a psychological toll; mothers describe doubling up as a guest, and the associated loss of authority, as incompatible with their ideals of adulthood and family life (Harvey 2020b). Intrahousehold power dynamics may help explain why individuals seem to prefer doubling up as a host over doubling up as a guest (Cohen and Casper 2002; Harvey 2020b; Pilkauskas and Michelmore 2019). If guest parents have less authority in the household and are more dissatisfied with their living arrangement relative to hosts, doubling up may subject guest children to more family stress.
Of course, although hosts typically provide housing support and guests receive it, both hosts and guests can benefit from doubling up. Because guests often contribute toward housing costs, doubling up can lower hosts' housing expenses (Harvey 2018; Pilkauskas et al. 2014). Additionally, not all doubled-up households are formed in response to housing needs. Coresidence may also be driven by generational needs (e.g., childcare or eldercare needs), cultural preferences, or for economic benefits beyond housing (Aquilino 1990; Kamo 2000). However, host/guest status appears meaningful in these situations as well. Parents associate household headship with authority in the home, even when the guest has other housing options or is providing financial assistance or care work for the host (Clark et al. 2011; Harvey 2020b). Likewise, although high rates of doubling up in certain racial/ethnic groups may reflect a cultural affinity toward doubling up, individuals across groups seem to prefer hosting over guesting (Cohen and Casper 2002).
Importance of Household Composition
Although differentiating between host/guest status is our main contribution, we also distinguish between three household compositions, based on whether the child and their parent(s) live with a grandparent(s), nongrandparent extended family member(s), or nonkin. Most prior research on doubled-up children has focused on multigenerational households (children living with grandparents). However, the formation and implications of doubled-up households likely vary by household composition (Harvey 2020a; Mollborn et al. 2011). For example, compared with multigenerational households, households formed with other extended family or nonrelatives may share household expenses to a greater extent and have more disagreements over resource sharing (Harvey 2018; Reyes 2018). Likewise, the role household members play in children's lives may vary based on their relationship to the child. Coresident grandparents often provide substantial child-rearing assistance (Dunifon et al. 2014; Dunifon et al. 2018), but the role of other relatives and nonkin is not clear. Thus, prior research provides compelling reasons to differentiate between household compositions, as well as host/guest status, when considering doubled-up household dynamics.
Children's Residence in Doubled-Up Households
Prevalence of Doubled-Up Households
Alongside research on doubling up as a housing arrangement, a growing literature in demography has documented the prevalence of doubled-up households (Mykyta and Macartney 2012; Pilkauskas et al. 2014). One recent study found that the share of children living with extended family or nonkin increased by over 18% from 1996 to 2009, driven primarily by multigenerational households (Pilkauskas and Cross 2018). Several other studies have likewise documented recent increases in children living in multigenerational households (Dunifon et al. 2014; Kreider and Ellis 2011; Pilkauskas 2012; Pilkauskas et al. 2020). Previous research has thus highlighted the large and growing number of doubled-up children; however, hosts and guests have typically been grouped together, despite evidence that families double up as hosts and guests for different reasons and that host/guest status shapes how they experience these arrangements.
Differences by Socioeconomic Status, Care Needs, and Race/Ethnicity
Prior studies have identified three key factors associated with doubling up: socioeconomic status, generational care needs, and cultural preferences, often proxied by race/ethnicity. First, families who are relatively disadvantaged in terms of education level and economic need are more likely to live doubled up than their more advantaged counterparts (Pilkauskas et al. 2014). Second, mothers with greater child-rearing support needs are more likely to live doubled up. Doubling up is more common among younger mothers and mothers with young children (Amorim et al. 2017; Pilkauskas et al. 2014). Likewise, unmarried mothers are more likely to be doubled up than married mothers (Dunifon et al. 2014; Pilkauskas 2012; Pilkauskas et al. 2014). Finally, rates of doubling up differ by race/ethnicity: compared with White families, Black, Hispanic, Asian, and Native American families are more likely to be doubled up (Amorim et al. 2017; Cross 2018; Kreider and Ellis 2011; Pilkauskas 2014; Pilkauskas et al. 2014). Whereas variation in rates of doubling up by socioeconomic status, care needs, and race/ethnicity are well documented, it is not clear whether these patterns are consistent across host/guest status.
Although little is known about how the share of children in doubled-up households varies by host/guest status, a small number of studies have documented differences between U.S. adults living doubled up as hosts and guests. These studies have found that compared with hosts, guests are, on average, younger, poorer, less highly educated, less likely to be employed, and less likely to be married (Beck and Beck 1989; Cohen and Casper 2002; Mykyta and Macartney 2012). Additionally, Kamo (2000) found that in White and Black adults' extended households, hosts are generally older than guests; by contrast, in Hispanic adults' extended households, hosts are often younger or similar in age to guests, and in Asian adults' extended family households, hosts are more evenly split between being older, younger, and similar in age to guests.
These studies provide some evidence of variation by host/guest status among U.S. adults. We extend this line of research by focusing on families with children and using recent data from the American Community Survey. Understanding children's experiences in doubled-up households is particularly important given the growing number of children in these arrangements (Pilkauskas and Cross 2018) and evidence that doubled-up households and changes in doubled-up household composition are associated with child well-being (Harvey 2020a; Mollborn et al. 2011; Perkins 2019). Because the factors associated with host/guest status may vary by household composition, we examine three distinct household types (households formed with grandparents, other extended family, and nonkin). We show how the share of children doubled up as hosts and guests in each household type varies by socioeconomic status, care needs, and race/ethnicity. Our descriptive analysis shows that doubling up as a host and guest are differentially linked with family characteristics. These findings build on the previously described qualitative research, which suggests that families may have different reasons for hosting and guesting.
Duration, Stability, and Transition Patterns
Finally, understanding children's living arrangements requires attention to their patterns of residence in different household types. Individuals beyond the nuclear family contribute substantial instability to children's households (Perkins 2017; Raley et al. 2019), and such instability is associated with child well-being (Mollborn et al. 2012; Perkins 2019). We build on studies of household composition instability, measured by transitions of individuals in and out of children's households (Perkins 2017; Raley et al. 2019), by examining children's transition patterns between household types. We also build on earlier work by documenting children's duration of residence in doubled-up household types.
First, we consider how long children typically reside in different doubled-up household types. To our knowledge, we provide the first estimates of children's duration of residence in different doubled-up household types, distinguishing by household composition and host/guest status. This analysis shed light on whether different doubled-up household types typically serve as temporary sources of support or long-term arrangements.
Next, we examine children's transition patterns between household types. A few studies have documented transitions in and out of doubled-up status, as well as transitions in and out of multigenerational coresidence. Using a sample of disproportionately lower-income families, Pilkauskas et al. (2014) and Pilkauskas (2012) examined children's households at five time points from birth to age 9 and found that families rarely doubled up over a long period; only 8% of mothers who doubled up did so at every survey wave (Pilkauskas et al. 2014), and only 2% of mothers who lived in multigenerational households did so at every survey wave (Pilkauskas 2012; see also Pilkauskas and Martinson 2014). Although these studies documented the prevalence of transitions in and out of doubled-up (or multigenerational household) status, the data they used have important limitations: they are based on a lower-income sample, which may not have representative patterns of doubling up; they have long intervals (up to four years) between survey waves, which may miss many transitions; and they focus only on early childhood. Our study documents transition patterns in a nationally representative sample of children for whom household type was assessed in four-month intervals. We also estimate differences by host/guest status, providing new evidence that hosting and guesting are distinct experiences.
Further, we show how children move between household types over time. This innovation is important because doubling up with different household compositions, and particularly as a host or a guest, may be experienced differently by children. For example, a child whose family lives continuously in a grandparent's home and a child whose family moves from a grandparent's home into their own home where they host an aunt are both continuously doubled up, but they experience very different living arrangements. Our analysis distinguishes between these trajectories. By examining transitions into, out of, and between doubled-up household types, we show how different household types fit into families' broader housing trajectories.
Data and Method
Data for this study come from two sources: the American Community Survey (ACS) and Survey of Income and Program Participation (SIPP). We use these two data sets in conjunction because they allow us to understand different aspects of children's doubled-up households and each has advantages and disadvantages. The ACS provides the most up-to-date national data on the share of children doubled up as hosts and guests. However, because the ACS is a repeated cross-section, it cannot describe household duration or patterns over time. Although slightly older than the ACS data, the longitudinal SIPP data allow us to examine the duration of doubled-up household types and to consider patterns in living arrangements over time. Together, they provide a comprehensive picture of children's residence in doubled-up households by host/guest status.
American Community Survey
To examine the prevalence of doubling up and variation by demographic characteristics, we use the ACS. The ACS is a nationally representative survey of the U.S. population that samples approximately 3 million households annually and is collected by the Census Bureau. The ACS data for this study were drawn from extracts made by the Integrated Public Use Microdata Sample (Ruggles et al. 2020). Our analyses of prevalence and differences by demographic characteristics use the 2018 ACS, and we also use data from 2005–2017 to examine trends over time. We restrict our sample to children under the age of 18 (N = 651,948 in 2018; sample sizes range from 704,608 for 2005 to 653,886 for 2017). For the analyses by parental demographic characteristics, we further restrict the sample to children under age 18 with at least one parent present (N = 620,031, 2018 only).1
Survey of Income and Program Participation
To examine household duration and transition patterns, we use longitudinal data from the 1996, 2001, 2004, and 2008 panels of the SIPP.2 We also use cross-sectional data from Wave 2 of the 1996, 2001, and 2004 panels, paired with ACS data from 2005–2018, to examine trends over time.3 SIPP panels are nationally representative samples of U.S. households collected by the Census Bureau. The SIPP interviews all household members age 15 years and older who are present at the time of the survey and gathers proxy responses for other household members. Because the SIPP follows all individuals over age 15 from the originally sampled household, even if they move to a new household, we are able to follow individuals over time as they dissolve households and form new ones. However, because the SIPP follows only individuals age 15 and over, younger children who change households and no longer live with a followable sample member (e.g., moving from a mother's household to live with a grandmother in a skipped-generation household) are lost from the sample.
The SIPP collects data at four-month intervals for 9 to 16 waves. We limit our longitudinal analyses to the first 9 waves (three years) of each panel for consistency. Although household composition data are collected for each month, we restrict our analyses to the SIPP reporting month, when reports are most accurate (Moore 2008). Children in all household types, including children not living with a parent, are included in all SIPP analyses. Our cross-sectional analyses include all children under age 18 at Wave 2 of each SIPP panel, a sample of 25,843 children for 1996, 19,973 for 2001, and 27,944 for 2004. Supplemental analyses shown in Table A3 and Figure A1 of the online appendix also use Wave 2 of the 2008 SIPP panel (N = 25,197).
For the longitudinal analyses, we limit our sample to children under age 15 in Wave 1 so that they can be observed for three childhood years (through age 17). The SIPP panels include 90,765 children under age 15 at Wave 1. The Census Bureau computes longitudinal panel weights for individuals who are in the sample at the beginning of the panel and for whom data are available for every month for which they are eligible for the survey (U.S. Census Bureau 2019). We first restrict our sample to the 40,857 children who have SIPP-generated longitudinal panel weights that run through Wave 9 so that we can accurately weight our analyses to be nationally representative (adjusting for attrition and oversamples). We also exclude 689 children who have longitudinal weights available but exited the sample before Wave 9 and 34 children who have mid-sequence missing spells longer than two waves (children who became ineligible, or temporarily ineligible, for the sample). These restrictions produce a final sample of 40,134 children. Comparing the demographic characteristics of our analytic sample with those of all SIPP children under age 15 at Wave 1, we find that children omitted from our sample are slightly more disadvantaged, a slightly smaller share are White, and a larger share have doubled up (see section 1 of the online appendix). However, these differences are small, and the longitudinal weights adjust for attrition.
The SIPP imputes data for item nonresponse (U.S. Census Bureau 2016). In our data, 151 of 361,206 child-waves are missing household type; for our longitudinal analyses, we assign household type for up to two consecutive missing waves using the child's prior and subsequent household type observations (following Raley et al. 2019). For children who were in the same household type before and after the missing spell, we treat them as if their household had not changed. For those who experienced a household type change, we treat them as if the change occurred at the wave in which they were first unobserved. Section 1 of the online appendix provides further information on attrition from the sample and missing household type.
In both the ACS and SIPP, we identify household relationships using parent pointers and household rosters. Parent pointers identify the child's mother and father. Household rosters describe the relationship of each person in the household to one household member, known as the household reference person. The household reference person is the person whose name is on the lease/mortgage, and we refer to this person as the householder. We consider children to be doubled up if they live in a household with their parent(s) as well as any adult other than their parent's romantic partner and their parent's adult children (the child's siblings). Doubled-up children are considered hosts if their parent or parent's romantic partner is the householder and are considered guests if an extended family member or nonrelative is the householder.
We categorize doubled-up children into six mutually exclusive household types based on household composition and host/guest status: (1) guest multigenerational if they live with a parent(s) and their grandparent is the householder; (2) guest extended family if they live with a parent(s) and a nongrandparent extended family member is the householder; (3) guest nonkin if they live with a parent(s) and a nonrelative is the householder; (4) host multigenerational if their parent or parent's romantic partner is the householder and the household also includes their grandparent(s); (5) host extended family if their parent or parent's romantic partner is the householder and the household also includes a nongrandparent adult extended family member(s); and (6) host nonkin if their parent or parent's romantic partner is the householder and the household also includes an adult nonrelative.
For guests, household type is determined by the child's relationship to the householder (the person with whom the child is living as a guest). However, other adults may live in the household as well. For example, if the child's family are guests in an aunt's home, the child is categorized as living in a guest extended family household, even if the household includes grandparents and/or nonrelatives as well.4 Host households (households in which the child's parent or parent's romantic partner is the householder) may also have multiple nonparent, nonsibling adults; we construct mutually exclusive categories by privileging grandparents, followed by other extended family, and finally nonkin. For example, if the child's parent is the householder and the household includes both an aunt and a nonrelative, the child is categorized as living in a host extended family household. The ordering of household types reflects the additional adult we expect to be most involved in the child's life.5
We also construct categories for two non-doubled-up household types: (7) non-doubled-up household (“nuclear family household”) if the child lives with a parent(s) in a household that includes no adults other than their siblings, parents, and parent's romantic partner; and (8) without a parent if they live in a household with no coresident parent (e.g., with a custodial grandparent). The online appendix (section 2) provides further detail on the ACS and SIPP, how they vary, and how household types are coded.
Our first set of analyses examines differences in the prevalence of doubled-up households for children by four key demographic characteristics from the ACS. First, we measure socioeconomic differences using maternal education (less than high school, high school, some college, or bachelor's degree or higher). Second, we include two measures of care needs: child's age (0–17) and maternal marital status (married, previously married, or never married). Finally, we examine differences by maternal race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or non-Hispanic Asian). For the analysis of prevalence by race/ethnicity, we omit children whose mothers are categorized as “other race.” We also present additional information on maternal earnings (in $1,000s, in 2018 dollars), labor force participation (full-time if working 35 hours or more per week, part-time if working less than 35 hours per week, unemployed, or not in the labor force), age at child's birth, whether the mother was born outside the United States, and whether she has a disability (cognitive difficulties, mobility problems, difficulty caring for personal needs, or vision/hearing impairment). For these demographic characteristics, if a mother is not in the household, we use information on the father; we refer to these as maternal characteristics throughout the paper because 95% of children are categorized based on their mother's characteristics. We also present data on the child's number of coresident siblings.
Table 1 describes the characteristics of children in our ACS sample who live with at least one parent. Only 10% of children have a mother with less than a high school diploma, and about one-third have a mother with a bachelor's degree or higher. A majority of children have a mother who is employed full- or part-time (68% total), and children's mothers earn an average of about $35,000 annually. Most children live with a married mother (67%), although 18% have a never-married mother. A majority of children in the sample have a non-Hispanic White mother (55%); respective percentages of children with non-Hispanic Black, Hispanic, and Asian mothers are 13%, 23%, and 6%. Nearly one-quarter of children have a mother who was not born in the United States. About 5% have a mother with a disability. Table A3 (online appendix) provides parallel descriptive statistics for the cross-sectional SIPP sample.
This study addresses three sets of research questions. First, what share of children live in each doubled-up household type, and how has the prevalence of these arrangements changed over time? Second, how does the prevalence of each household type vary by socioeconomic status, childcare needs, and race/ethnicity? To address these first two research questions, we present weighted percentages and cross-tabulations using ACS data.6
Our third research question is, what are children's patterns and duration of residence in different types of doubled-up households over time? To address this question, we use longitudinal data from the SIPP. To visualize transition patterns, we use Stata's SQ-Ados package to produce sequence index plots (Brzinsky-Fay et al. 2006). To measure children's duration in different doubled-up household types, we estimate the average length of all spells of residence in a given household type during our three-year observation window. Note that stable residence in a given doubled-up household type does not necessarily imply stability in household composition. Other research has focused on household composition changes (Perkins 2017; Raley et al. 2019), but our measure identifies patterns of residence across different household types (building on Pilkauskas et al. 2014) and shows how doubling up in a given household type fits into families' overall household trajectories.
What Share of Children Live in Doubled-Up Households as Guests and Hosts, and How Has This Changed Over Time?
We first use the 2018 ACS to describe the percentage of children who live with their parent(s) in a doubled-up household as guests (in someone else's home) and as hosts (in a home headed by the child's parent or parent's romantic partner). Figure 1 shows that 15.4% of children live doubled up, and these children are almost equally split between hosts (7.9%) and guests (7.5%).7 The composition of children's doubled-up households varies by host/guest status. Most guests live in a grandparent's home (6.1%), and very few live in a nonrelative's home (0.3%). In contrast, just under one-half of hosts live with a grandparent (3.8%), and the rest are split between households with other extended family (2.2%) and households with nonkin (1.8%).
To examine whether the share of children in each household type has remained consistent over time, we use data from the SIPP for 1996, 2001, and 2004 and from the ACS for 2005–2018. Although these two data sets have some differences in sampling techniques (see the online appendix, section 2), together they allow us to examine the prevalence of each household type over the last two decades. The results, presented in Figure 2, show that increases in children's residence in doubled-up households have been driven primarily by multigenerational households. The share of children in host multigenerational households increased from 2.1% of children in 1996 to 3.8% in 2018. Similarly, guest multigenerational households increased substantially, from 3.7% to 6.1% over the same period. No other doubled-up household type experienced stable growth over this period.
How Does the Prevalence of Doubling Up as a Guest and Host Vary by Socioeconomic Status, Childcare Needs, and Race/Ethnicity?
First, we consider how the prevalence of doubling up as a guest and host varies by socioeconomic status. Because doubling up is a common response to economic and housing needs, we expect that children whose mothers have lower education levels would have higher rates of doubling up, especially as guests. Figure 3, which presents the prevalence of each household type by mother's education level, is partly consistent with this premise; children whose mothers have higher levels of education have lower rates of doubling up than children whose mothers have lower levels of education. Differences by host/guest status follow a less consistent pattern. Children whose mothers have a bachelor's degree or higher have higher rates of doubling up as hosts (6.4%) than guests (3.0%), and children whose mothers completed high school only have higher rates of doubling up as guests (12.6%) than hosts (9.5%). However, for children whose mothers did not complete high school or who completed some college, rates of doubling up as guests and hosts are similar (about 11% and 8%, respectively).
Nearly all doubled-up household types are less prevalent at higher maternal education levels, suggesting that economic need may influence doubling up for both hosts and guests. Only host multigenerational households lack this gradient. Children have roughly similar rates of residence in host multigenerational households (3.6% to 4.3%) across maternal education levels. The lower rates of overall hosting among children who have more highly educated mothers, compared with those who have less-educated mothers, is driven by low rates of hosting nongrandparent extended family and nonkin.
The descriptive statistics in Table 1 provide further evidence that guests tend to be more economically disadvantaged than hosts. Compared with the full sample ($34,770), guest children had mothers with lower earnings ($20,500), whereas host children had mothers with earnings that were similar to the overall average ($35,200). Moreover, the economic disadvantage of guests is likely understated because a greater share of host children's mothers are married (60%, compared with 26% for guests) and thus have another potential earner in the family.
Next, we consider how the prevalence of doubling up varies by childcare needs. Prior research suggests that children are more likely to live doubled up during early childhood (Amorim et al. 2017; Pilkauskas et al. 2014), when families may need additional support. We examine whether this finding holds across household types and how rates of living doubled up vary throughout childhood. Figure 4 shows the percentage of children in each doubled-up household type at each age. We find that the decline in the prevalence of doubled-up residence with age is driven by guest multigenerational households. Although 10.3% of children who are under age 1 live in guest multigenerational households, just 3.1% of 17-year-old children do, a decline of nearly 70%.8 Likewise, the prevalence of guest extended family households declines by 50% between ages 0 and 17 (from 1.4% to 0.7%), although these households are less common overall. In contrast, the prevalence of other doubled-up household types does not vary substantially by age.
Next, we examine how the share of children who are doubled up varies along another indicator of care needs: mother's marital status. Previous research found that unmarried mothers, who may have additional childcare and financial needs, have higher rates of doubling up (Pilkauskas et al. 2014). Additional household members, especially grandparents, can provide child-rearing support for lone mothers who would otherwise have no coresidential assistance (Kalil et al. 2014). Consistent with prior research, Figure 5 shows that children with unmarried mothers have higher rates of doubling up than children with married mothers. Moreover, these patterns vary by host/guest status. The prevalence of doubling up as a host is only modestly higher for children of never-married (10.3%) and previously married mothers (10.1%) than for children of married mothers (7.3%). The slightly higher rate of hosting among children of never-married and previously married mothers compared with married mothers is driven by higher rates of hosting extended family and nonkin. In contrast, doubling up as a guest is nearly twice as common for children of never-married mothers than for children of previously married mothers (22.4% vs. 12.1%) and more than five times as common than for children of married mothers (22.4% vs. 3.9%). Across all marital status groups, the majority of guests live in multigenerational households. Finally, children of married mothers have higher rates of hosting than guesting, whereas the opposite is true for children of previously married and, especially, never-married mothers.
Finally, we build on previous findings of racial/ethnic variation in rates of doubling up by examining differences by host/guest status, shown in Figure 6. Our findings confirm that a greater share of children of Black, Hispanic, and Asian mothers live doubled up than children of White mothers. However, these patterns mask variation by host/guest status. Children of Black mothers have fairly high rates of hosting (8.6%) and even higher rates of guesting (11.6%), especially in guest multigenerational households (9.7%). A smaller share of children of White mothers are doubled up as either guests (5.7%) or hosts (5.7%); however, when guests, they are almost always in multigenerational households (5.0%). Children of Hispanic mothers have high rates of doubling up as both guests (10.7%) and hosts (11.9%), and compared with other groups, they have higher rates of hosting extended family (4.2%) or nonkin (2.4%). Finally, compared with other groups, children of Asian mothers have the highest rates of hosting (16.6%) and have particularly high rates of hosting multigenerational households (12.2%). In contrast, they have among the lowest levels of doubling up as guests (5.8%).
What Are Children's Duration and Patterns of Residence in Doubled-Up Households, and How Do They Vary by Household Type?
We use longitudinal SIPP data to examine the duration of different doubled-up household types over a three-year period. As described in the Method section, our analysis focuses on how long children lived in each household type rather than how stable their overall household composition is. For example, we consider how long a child lived doubled up in a guest multigenerational household, regardless of whether the composition of that household type changed because individual household members moved in and out. The first row of Table 2 shows the average length of time (in years) that children spent in each household type during the three years of SIPP data. On average, spells of residence in multigenerational households were longest (1.9 years for guests, 1.8 for hosts), nonkin households were shortest (1.3 years for guests, 1.0 for hosts), and extended family households were in between (1.5 years for guests, 1.2 for hosts). Within each household composition (multigenerational, extended family, and nonkin), spells in which children were guests lasted slightly longer, on average, than spells as hosts. Additionally, children spent longer, on average, doubled up as guests (across household types) than as hosts (1.9 years vs. 1.4 years).9
The third row of Table 2 shows the proportion of children who remained in a specific household type throughout the three-year observation period (i.e., had no household type transitions). Children in multigenerational households had the highest rates of household type stability over the three years (47% of guests, 39% of hosts), whereas children in nonkin households had the lowest rates of household type stability (18% of guests, 9% of hosts). For each household composition (multigenerational, extended family, and nonkin), children in guest households had higher rates of household type stability than children in host households.
Last, we examine the transition patterns of children who doubled up, again using SIPP data. Figure 7 presents visual representations of the sequences of household types over three years. In these plots, each row visualizes a single child's sequence of households. The x-axis shows time (survey wave number), and the y-axis shows the number of children. Each plot includes all children who ever experienced a particular household type, so children who lived in multiple doubled-up household types during the three-year period are included in multiple plots. These plots reveal common patterns and variation among children who experienced each household type. The solid rows at the top of each plot depict children who resided consistently in a specific doubled-up household type. The rows below depict children who experienced at least one transition between household types over the three-year period.
Figure 7 shows that although guests had higher rates of remaining in one household type over the three years than hosts, guests who do not consistently live in a single household type throughout the three-year period experienced a variety of other household types. In contrast, hosts who did not live consistently in a single household type throughout the three-year period more often spent much of the rest of the observation period in a household that was not doubled up (dark gray).
To examine movement between household types, Table 3 focuses on children who doubled up at some point during the three years and experienced at least one household type transition (children who were consistently doubled-up in one household type are not included in this table).10 Among children with at least one transition, guests had more transitions over the three-year period on average (1.75) than hosts (1.65). Thus, although guests had higher rates of remaining in one household type over the three years than hosts (see Table 2), among children who did not remain in a single household type, guests experienced more movement between types on average than hosts (Table 3). Among children with at least one transition, those in host multigenerational households experienced the fewest transitions between household types on average (1.60), whereas those in nonkin guest households experienced the most (1.87).
Furthermore, the results show differences by household type in the nature of transitions children typically experienced. The second row of Table 3 shows that multiple spells of residence in a specific household type (i.e., exiting a household type and reentering it again within the three-year period) was most common for children who lived in guest multigenerational households; 14% of children who lived in guest multigenerational households for part of their observation period experienced multiple spells of residence in such households over the three years. Additionally, multiple spells of residence were fairly common for children who lived in host extended kin households (8%) and host nonkin households (10%).
The final two rows of Table 3 show the proportion of children who lived in each household type for part of their sequence and also lived in one or more (or two or more) other doubled-up household types. Of children with at least one transition, guests more often lived in multiple doubled-up household types over the three years than hosts. About one-third of children who lived in guest multigenerational households for part of their sequence also lived in at least one other doubled-up household type, as did 40% of children who lived in guest extended family households and 46% of children who lived in guest nonkin households. Children who lived as guests in one household type had higher rates of living as guests in other household types (see Table A4 in the online appendix for full information on children's residence in multiple household types). This finding suggests that a subset of guests use doubling up as a longer-term housing strategy, cycling between households with grandparents, nongrandparent extended family, and nonkin. In contrast, less than three-tenths of children who lived in each host household type for part of their sequence lived in multiple doubled-up household types.
This study contributes to the growing family demography literature on children's coresidence with adults beyond the nuclear family by attending to both housing status and household composition. We show that the 15% of American children who live doubled up are approximately equally split between hosts and guests: about half live in a home that is rented/owned by their parent or parent's romantic partner, and the other half are guests in someone else's home. Host/guest status often reflects whether the family receives housing support or provides it to others. Relying on others for housing can leave families precariously housed and limit parents' control over the home environment (Edin and Shaefer 2015; Harvey 2020b); thus, it is notable that nearly 8% of U.S. children are doubled-up guests, with no home of their own. However, assuming that most doubled-up children live in someone else's home would be a mistake; half are doubled up with their nuclear family providing housing support to others.
Children's household composition varies dramatically by host/guest status. Guests primarily live with grandparents, but hosts are more evenly divided between hosting grandparents, other extended family, and nonkin. This variation may have implications for understanding the results of prior studies. For example, previous research has found that coresidence with grandparents may have different associations with child outcomes than coresidence with other extended family or nonkin (Harvey 2020a; Mollborn et al. 2011, 2012). Future research should explore how differences in whether the parent is giving or receiving housing support contribute to these patterns.
Building on earlier research (Pilkauskas and Cross 2018), we find that both guest and host multigenerational households have become more common in the past two decades. This growth may reflect need on the part of both grandparents and families with children. Scholars have argued that intergenerational ties are increasingly important, and research has often focused on flows of support from older to younger generations (Swartz 2009). Consistent with this focus, we find that in multigenerational households, grandparents typically provide housing to the grandchild and parent(s). However, because hosting a grandparent is increasingly common as well, this arrangement deserves greater research attention.
Socioeconomic status and childcare needs are associated with whether doubled-up families are hosts or guests. Compared with younger children, older children have lower rates of doubling up as guests in multigenerational and extended family households, but the prevalence of doubling up as a host is fairly consistent across child age. This trend may be driven by the needs of families with young children: as parents and their children age, they rely less on support from other family members. Similarly, children generally have higher rates of guesting if their families exhibit greater need in terms of maternal education and family structure. Children whose mothers are highly educated and children with married mothers have higher rates of hosting than guesting. Qualitative research suggests that relative to hosting, doubling up as a guest often reflects a greater need for support, and our findings are largely consistent with this idea.
However, hosts are not necessarily advantaged. Compared with guests, host children more often have highly educated and married mothers, but both hosting and guesting are more common for children of unmarried mothers and mothers with lower education levels. Disadvantaged families are often enmeshed in disadvantaged social networks (Pilkauskas et al. 2017), which may lead them to provide housing support to others when they can. Additionally, guests often contribute financially to hosts (Harvey 2018), so hosting may also be an income strategy for disadvantaged families. These factors may help explain why children whose mothers have less than a high school diploma have similar, and high, rates of doubling up as both hosts and guests. Future research should explore the circumstances and motivations that prompt families to double up as hosts and guests and consider how hosts' and guests' needs might interact in decisions to double up. Our analysis incorporates only measures of the child's family's needs, but future studies should examine how the size and content of children's social network (e.g., whether their grandparents are living) and network members' needs (e.g., whether their grandparents are in poor health) might influence whether children double up.
We also find variation in rates of hosting and guesting by race/ethnicity. Black children have higher rates of guesting than hosting; by contrast, Asian children have far higher rates of hosting than guesting, and Hispanic children have high rates of both hosting and guesting. Our findings replicate and build on previous research showing that Asian, Hispanic, and Black children double up at higher rates than White children (Cross 2018), and we show the importance of distinguishing between hosts and guests for understanding these racial/ethnic differences. Future research should consider whether these patterns reflect variation in norms surrounding the provision of support and should attempt to disentangle the independent effects of needs and preferences. Additionally, future research should ask whether host/guest status contributes to differences by race/ethnicity in the associations between doubling up and child well-being (Mollborn et al. 2011; Pilkauskas 2014).
To our knowledge, our study is the first to estimate children's length of residence in different shared household types by host/guest status. We find that children's spells of residence in guest households tend to be somewhat longer than in host households, and that spells in multigenerational households (both guest and host) are the longest of any doubled-up household composition. We also find that transitions between household types are common; among children who experience at least one transition between household types, guests experience more transitions on average and have higher rates of living in multiple doubled-up household types, compared with hosts. These findings may reflect differences in guest and host families' available housing options. Families with children who double up as guests may be likely to find a host (often a grandparent) who is willing to host them for as long as they need. Additionally, even if they change households, guests who do not have their own housing may be forced to remain doubled up until they are able to obtain an independent home of their own. In contrast, children's families may be less willing to bring additional adults into the household for long periods, and because they already have their own housing, hosts can more easily transition to a nuclear family household. Although evaluating the effects of doubling up is beyond the scope of this analysis, our findings suggest that guest children may be particularly vulnerable. Continued coresidence with others may reflect a lack of options, especially given evidence that families prefer residential independence (Harvey 2020b; Pilkauskas and Michelmore 2019). Prior research on the link between doubling up and child well-being, like much of the literature on doubling up, has typically not distinguished between host and guest status, but our findings suggest that future research should account for this important dimension of shared household residence.
Our study is not without limitations. First, children live in many household types. We focus on host/guest status in three shared household compositions (multigenerational, extended family, and nonkin), but our categories do not account for full household composition; for example, we classify guests according to the child's relationship to the householder, even if other adults live with them. Future research should explore the full household composition of doubled-up households and consider how household composition instability might vary by host/guest status. Second, our analyses of transitions only cover a three-year period, which is a limitation of the SIPP data. The SIPP's frequent data collection (every four months) is well-suited for capturing short-term changes, which is particularly important because doubled-up households tend to be highly unstable. However, future research with data that cover a longer time frame would be valuable for understanding children's transition patterns throughout childhood. Finally, although SIPP weights adjust for attrition, we may underestimate transitions and doubling up over time as those who attrite are somewhat more disadvantaged.
Nonetheless, our study makes important contributions to knowledge about doubling up among children. Our findings show that treating hosts and guests as a single group masks important differences in children's experiences in doubled-up households. We highlight the uniquely precarious situation faced by guest children: compared with hosts, guests more often are socioeconomically disadvantaged and seem to use doubling up—whether in a single household type or across multiple household types—as a longer-term strategy. Overall, these results underscore the importance of attending to housing status, in addition to household composition, when studying children's doubled-up households.
We exclude children who do not live with parents (4.3% of children; N = 31,917, unweighted).
The most recent panel, which began in 2014, was dramatically redesigned, including a change to annual interviews.
We use Wave 2 so that we can produce descriptive statistics for the SIPP that are comparable to those for the ACS (Table A3, online appendix). We only use the 1996, 2001, and 2004 SIPP data to examine trends over time because annual ACS data are available starting in 2005. Figure A1 in the online appendix shows results using the 1996, 2001, 2004, and 2008 SIPP panels and 2008–2018 ACS data.
The household roster does not document all relationships to the child. However, we can identify grandparents in guest extended family and guest nonkin households. We find that only 0.05% of children in guest extended family/guest nonkin households combined also live with a grandparent.
Host and guest children are rarely in the same household (i.e., many children double up as guests of adults who do not have minor children or double up as hosts to adults who do not have minor children). In 2018, just 1% of children lived doubled up in a household in which at least one child was a host and at least one child was a guest.
In an extension, we consider whether the differences we observe by maternal education, marital status, and race/ethnicity are driven by differences in population composition (see Figure A2, online appendix). We estimate predicted probabilities based on weighted regressions predicting household type by our key demographic characteristics (maternal education, marital status, and race/ethnicity), also controlling for the other demographic characteristics in Table 1 (maternal age at child’s birth, disability status, earnings, and employment status; whether the mother was born outside the United States; child’s age; and child’s number of coresident siblings). We predict the probability of living in a particular household type by setting all covariates to the mean, except the focal characteristic. The regression-adjusted means produce findings that are substantively similar to our main analysis, with three exceptions. Compared with the unadjusted estimates, in the regression-adjusted means, (a) children with college educated mothers have higher rates of guesting, (b) children of Black mothers have lower rates of guesting, and (c) children of Asian mothers have higher rates of guesting. Differences between the unadjusted and regression-adjusted estimates for children of college-educated mothers and Black mothers are primarily driven by maternal marital status, whereas differences for children of Asian mothers are primarily driven by maternal immigration status and, to a lesser extent, marital status.
Although these point-in-time estimates are static, children’s residence in doubled-up households is not. In the longitudinal SIPP data, approximately 1% of children lived as both a host and a guest at different points during the three-year period. Table A4 (online appendix) shows the proportion of children in a given household type who also lived in each other household type.
We believe that this decline is likely driven by lessening parental needs as children age. We think it is unlikely that grandparent mortality explains this decline, given that we do not see the same pattern for host multigenerational households. Additionally, grandparents are quite young (in the United States, the median age at transition to grandparenthood is 49 for women and 52 for men; Leopold and Skopek 2015), and life expectancy in 2017 was 79 years (Arias and Xu 2019).
Table A4 in the online appendix shows that hosts have higher rates of living in nuclear family households than guests. Just 45% of children who doubled up as guests also lived in a nuclear family household at some point, compared with 72% of children who doubled up as hosts.
Our household type categories are not mutually exclusive: some children doubled up in multiple types over the three-year period.