Abstract
We document changes in U.S. children's family household composition from 1968 to 2017 with regard to the number and types of kin that children lived with and the frequency of family members' household entrances and departures. Data are from the U.S. Panel Study of Income Dynamics (N = 30,412). Children experienced three decades of increasing instability and diversification in household membership, arriving at a state of “stable complexity” in the most recent decade. Stable complexity is distinguished by a decline in the number of coresident parents; a higher number of stepparents, grandparents, and other relatives in children's households; and less turnover in household membership compared with prior decades, including fewer sibling departures. College-educated households with children were consistently the most stable and least diverse. On several dimensions, household composition has become increasingly similar for non-Hispanic Black and White children. Children in Hispanic households are distinct in having larger family sizes and more expected household entrances and departures by coresident kin.
Introduction
Family household composition is responsive to changes in the economic, technological, cultural, legal, and social contexts in which families are embedded (McLanahan 2004; Ruggles 2015). In the broad sweep of American economic history, families have transitioned from multigenerational households that included adult children and were organized to achieve small-scale agricultural production and manufacturing to nuclear households in which adult children pursue education, wage employment, and economic independence while living apart from their parents (Gutmann et al. 2002; Ruggles 2015). More recently, discourse around U.S. family composition has emphasized how the labor market, legal frameworks, and family policy have interacted with characteristics of social location (e.g., race, gender, and class background) to shape the emergence of a variety of dynamic family forms as contexts for child-rearing over the last half-century (Cherlin et al. 2016; Edin and Kefalas 2005; Furstenberg 2014; McLanahan and Percheski 2008).
We build on this literature by engaging four strands of family demographic research to provide a comprehensive portrait of stability and change in U.S. family household composition from the perspective of U.S. children between 1968 and 2017: (1) coresidence with parents; (2) the number and relatedness of siblings; (3) family complexity, as measured by stepparent and stepsibling coresidence; and (4) household complexity, as measured by coresidence with grandparents and other relatives. We consider childhood coresidence with kin on two dimensions. First, we consider the average number of each type of kin with whom a minor child (aged 0–17 years) lived in each of the five decades between 1968 and 2017. Second, we estimate the amount of change in household composition children would be expected to experience in terms of the number of kin entrances and exits over the same age interval if they were exposed to the patterns of family change observed in each decade—a measure we call the coresidence transition rate. Together, these estimates describe variation in dynamic household composition over recent historical time. Further, we describe emergent patterns of diversity and similarity in family structure by race and ethnicity and by educational attainment. Data are from the U.S. Panel Study of Income Dynamics family unit relationship matrix. This novel data source enables the first profile of children's dynamic and multifaceted family structure in each decade over 50 years.
Background
Significance of Family Composition Change
Attention to historical change in coresident family composition from children's perspective is important because family households provide resources critical to children's well-being and development. First, family households are the primary conduit to the material goods and services children need to thrive (Brooks-Gunn and Duncan 1997; Duncan et al. 1998; McLanahan and Sandefur 1994). Coresident kin also invest time in children, ranging from 3.5 to 4.5 hours per day depending on family composition (Kalil et al. 2014). This time contributes to emotional closeness, and time engaged in shared learning activities is associated with subsequent academic achievement (Fiorini and Keane 2014; Hsin and Felfe 2014; Raley 2014).
Second, children's access to family resources varies depending on family household composition status—that is, the other people with whom children live. As we elaborate later, recent research in family demography has focused on the presence and biological and legal relatedness of various kin types, including married or cohabiting biological or adoptive parents or stepparents; full siblings, half-siblings, and stepsiblings; and vertical and lateral kin extension, including grandparents, aunts, and uncles. This work has drawn on a variety of perspectives to consider whether and why these aspects of family composition appear to influence investments in children. Broadly, these perspectives have emphasized how the number and characteristics of individuals in a family unit, their relatedness, and their competing obligations or allegiances to family members living elsewhere may influence the supply of time, money, and emotional energy available to invest in children net of mechanisms that select adults into varied living arrangements (Cross 2020; Fomby et al. 2016; Hofferth 2006; Kalil et al. 2014; McLanahan and Sandefur 1994).
Further, U.S. children's family household composition status has increasingly diverged by race, ethnicity, and social class over the past 50 years. Although the share of U.S. children living only with stably married parents and full biological siblings throughout childhood has declined overall during the last half-century, this mode of family organization has continued to predominate for White children and children whose parents have a college degree. In contrast, since the mid-1970s, the majority of Black children have lived outside of their parents' marriage (U.S. Census Bureau 2019), and nonmarital cohabitation as a context for childbearing has risen most sharply among adults with less than a college education (Lundberg et al. 2016). Growth since the 1980s in the share of children living with grandparents or extended kin has also been higher in Hispanic families and in families with less than a college education (Pilkauskas et al. 2020). Thus, family composition status is socially stratified in the United States. In the absence of a strong public safety net for children, disparities in family structure–based resources may exacerbate racial/ethnic and class-based disparities in children's development and well-being.
Third, children experience frequent transitions in family household composition. These transitions include events such as the addition of a newborn sibling or an older sibling's home-leaving, but they also may include a biological parent's entrance after a child's birth or a departure following union dissolution, the entrance or departure of a coresident parent's new partner or spouse and their children, or changes in coresidence with extended kin. Additional examples of such transitions include family members' temporary departures and reentry as a result of incarceration, military deployment, or other remote employment. For the most part, above and beyond family household composition at any point in time, change (particularly repeated change) in composition is independently predictive of children's diminished educational attainment, increased risk of behavior problems, and early childbearing and family formation (Cavanagh and Fomby 2019). Yet, persistence in a given family structure arrangement is sometimes less advantageous than transitions into other arrangements (Dunifon and Kowaleski-Jones 2007; Musick and Meier 2010; Van Hook and Glick 2007).
Increasingly, family demography research has distinguished between children's family composition status and family composition transitions to develop more refined models of family structure experience. By age 18, about one third of children born in the 1970s had experienced a parent's household entrance or departure, and about 35% had experienced another adult family member's entrance or departure (Perkins 2019). Using synthetic cohort estimates from children's household composition between 2008 and 2013, Raley et al. (2019) estimated that children in the 2010s would experience just under five changes in household composition by age 18, including household entrances and exits by parents and parent figures, siblings, extended kin, and nonrelatives; change was estimated to be more frequent among Black and Hispanic children and children whose parents had not completed college (also see Cross 2018). These cohort-specific studies illustrate the prevalence of family composition change as a feature of children's family lives, but insight into how exposure to family composition transitions has evolved over recent historical time is lacking.
Dimensions of Family Household Composition and Transition
Parental Coresidence
Biological and adoptive parents' union status and coresidence with children consistently attract scholarly and policy attention because of their robust and enduring association with children's short- and long-run well-being across a variety of domains, including educational attainment (Fomby 2013; Martin 2012; McLanahan and Sandefur 1994; Sassler et al. 2013; Wojtkiewicz and Holtzman 2011; Ziol-Guest et al. 2015), the timing of first birth (Musick 2002; Wu and Martinson 1993) and union formation (Amato et al. 2008; Amato and Patterson 2017; Hofferth and Goldscheider 2010; Ryan et al. 2009), income mobility (Bloome 2017), and life satisfaction (Amato and Keith 1991; Amato et al. 1995). In the United States over the last half-century, this focus reflects the declining postconception marriage rate and a commensurate increase in the share of all births that occurred outside of marriage during the 1970s (Akerlof et al. 1996), the rising divorce rate during the 1970s and 1980s (followed by a plateau and decline in subsequent decades; National Center for Health Statistics 2020), and the emergence of cohabitation as an increasingly frequent and relatively unstable context for childbearing compared with marriage during the 1980s and 1990s (Bumpass and Lu 2000). This period saw steady increases in the age at first marriage and declines in the marriage rate as attitudes about premarital sex, childbearing in cohabitation, and divorce became increasingly accepting (Brown and Wright 2016; Harding and Jencks 2003; Powell et al. 2010; Thornton and Young-DeMarco 2001) and marriage came to be perceived as a capstone event preceded by the achievement of personal economic stability and independence (Cherlin 2009, 2014; Gibson-Davis et al. 2005). Consistent with these changes in adults' patterns of union formation and stability, the share of children ages 0–17 years living with both parents declined from 88% in 1960 to 67% in 2005 and has since plateaued at approximately 70% (U.S. Census Bureau 2019).
A growing literature has separately documented the occurrence of family change in children's households through parents' coresident union formation and union dissolution. Among children born between 1958 and 1965, 21% of White youth, one third of Black youth, and one quarter of Hispanic youth experienced at least one change in parents' union status between ages 13 and 19 (Wu and Martinson 1993). Exposure to family instability through parental union formation or dissolution (excluding transitions from cohabitation to marriage) persisted in subsequent cohorts: more than one third of youth experienced at least one such change by mid-adolescence (Cavanagh 2008) and an average of 0.89 parental union transitions by age 12 (Brown et al. 2016) in the mid-1990s. Parental union instability has largely plateaued since then, with children estimated to experience about 1.01 changes in 2006–2010. Across historical time, exposure to parents' union instability has been socially patterned, such that Black and Hispanic children have experienced more union transitions than White children (Brown et al. 2016).
Sibling Coresidence
Change in U.S. children's coresidence with biological siblings over time is a function of two countervailing forces. On the one hand, contemporary low fertility implies that children today have fewer siblings compared with children growing up during the 1960s and 1970s in families formed in the last years of the baby boom. Between 1970 and 2012, children's average sibsize (the number of children, including oneself, who share the same mother) declined from 4.4 to 3.0 children. This reflects a broad decline in women's achieved family size across racial and ethnic identity categories and levels of educational attainment (Fahey 2017).
On the other hand, adult children today leave the parental home at increasingly older ages and are more likely to return to live with parents during their 20s relative to those in the 1960s, a distinctive historical period when long-running economic growth fueled young adults' early marriage and economic and residential independence (Newman 2012). In 1970, 37% of young adults ages 18–24 and 10% of those ages 25–29 lived in their parents' households. By 2017, these proportions had increased to 51% and 20%, respectively (IPUMS 2020, census and American Community Survey–weighted estimates, author analysis). Young adult home-leaving is also marked by substantial churning, or reentry into and repeated departure from the parental household. Among young adults born between 1980 and 1984, for example, 90% of young adults had left home at least once by age 27, but more than half of those home-leavers returned to live in the parental home at least once in the decade after entering adulthood (Dey and Pierret 2014). Thus, today's minor children have fewer siblings compared with their age peers half a century ago, but they are also likely to live with older siblings for a larger fraction of their own childhoods.
Family Complexity
Carlson and Meyer (2014) described family complexity as occurring when marriage and legal ties, living arrangements, fertility, and parenting are not coterminous. In practical terms, this definition describes systems of family organization other than married, coresident two-parent families in which both parents have all of their biological or adoptive children in common. Historically, family complexity occurred most often through a parent's remarriage following a spousal death. During the mid-twentieth century, remarriage following divorce emerged as the most frequent pathway to family complexity (Coleman and Ganong 1990). Since the 1980s, increases in nonmarital multipartner fertility, frequent childbearing in cohabitation, high rates of cohabiting union dissolution, and the emergence of cohabitation with a new partner as an alternative to remarriage have led to more diverse pathways into complex family organization (Manning et al. 2014). Thus, the concept of family complexity powerfully connects a variety of family types and statuses over historical time.
Here, we consider one dimension of family complexity: coresident stepfamily organization, including the presence of stepparents (married or cohabiting) or stepsiblings in children's households. National estimates indicate that children's residence in a stepfamily has not changed substantially during the last several decades: approximately 10% of children lived with a stepparent in 1980 (Glick 1989) and in 2009 (Manning et al. 2014), and about 25% of children are estimated to spend time residing in a stepparent family (Bumpass et al. 1995). But these stable prevalence estimates conceal substantial changes in the background and union status of stepfamilies over time. Among women and their children with stepfamily experience in the mid-1990s, an increasing share entered that family type through a nonmarital birth rather than divorce or widowhood, and stepfamilies were increasingly likely to be formed through cohabitation rather than marriage (Bumpass et al. 1995). To the extent that cohabitation and nonmarital fertility have become increasingly common among parents with less than a college education over the same period, the experience of family complexity through stepfamily coresidence is likely also to be increasingly socially stratified.
Household Complexity
Household complexity refers to coresidence with vertically or laterally extended kin in multigenerational households. From the perspective of children, this most often refers to living with grandparents or aunts and uncles in a household headed by one's parent(s) or some other kin member. Until the post–World War II period, household complexity was a stable feature of U.S. family organization. In the mid-nineteenth and early twentieth centuries, 8% to 9% of children lived with grandparents, and the share peaked at 10% in 1950. Rising family income, growth in affordable housing, and residential mobility spurred young families' more frequent residence apart from parents or other kin (Ruggles 2007). By 1980, only 5% of children lived with extended kin, although the prevalence of such an arrangement remained higher in Black families than in White families. Kin coresidence began to recover in the decades that followed, largely owing to declining marriage rates and higher rates of household complexity among recent immigrants compared with the U.S. native-born population (Landale et al. 2011; Pilkauskas et al. 2020). The collapse of the housing market and subsequent high rates of unemployment during the Great Recession in 2007–2008 catalyzed what has become broad incremental growth in the prevalence of multigenerational households. In 2018, 9.9% of children lived with at least one grandparent (Pilkauskas et al. 2020). Estimates from a recent national cohort suggest that over the course of childhood, more than one third of U.S. children will reside in an extended kin household, and extended kin coresidence is more frequent among children in Black and Hispanic families than among those in White families (Cross 2018).
High lifetime prevalence combined with lower estimates of point-in-time prevalence of multigenerational households implies that children experience relatively frequent entrances into and exits from coresidence with extended kin. Indeed, among children born during the 1970s, Perkins (2019) reported that 37% of children experienced a change in coresidence with kin other than parents, compared with 35% of children who experienced a change in coresidence with parents. In their estimates derived from recent synthetic cohort data, Raley et al. (2019) reported that by age 18, contemporary youth had experienced 2.0 entrances or departures by coresident grandparents or other relatives, compared with 1.2 transitions in parent coresidence and 1.7 transitions in sibling coresidence.
A growing body of research has examined whether and why household complexity is related to children's well-being. The direction and magnitude of the observed association vary by race and ethnicity and by the reasons extended family members choose to live together. Much of the extended kin coresidence that children experience occurs when the children are young (Cross 2018; Mollborn et al. 2011). Early in the family life cycle, extended kin coresidence allows parents to pool resources and rely on close kin for childcare. These factors may contribute to more frequent household complexity among teen parents and poor families (Kreider and Ellis 2011) and to better outcomes for children's early cognitive achievement and behavior relative to peers living with single parents but not relative to married-parent households without other kin. Gains are also greater for Black and Hispanic children than for White children (Mollborn et al. 2011), perhaps partly because household complexity is more common across income and education levels in non-White families than in White families (Cross 2018).
In sum, multiple dimensions of family household composition status and change are each related to children's well-being, and their prevalence and correlates have changed during the last half-century. Significant recent advances in the conceptualization and measurement of each of these family structure experiences reflect the increasing diversity and dynamic quality of family forms. We apply these innovations to dynamic family household roster data to create period estimates of children's household composition and change from the late 1960s—an era distinguished by relative homogeneity in family organization—to the present. This approach enables a comparison of predominant patterns of children's coresident family structure over the last five decades. However, our focus on coresident family relationships is likely to offer an incomplete portrait of complex and multigenerational families whose social, legal, and biological ties extend across households (Seltzer 2019).
Data and Methods
We use data from the 1968 to 2017 waves of the U.S. Panel Study of Income Dynamics (PSID), the world's longest-running intergenerational household panel survey (Panel Study of Income Dynamics 2021a). The PSID began in 1968 with a sample of 4,802 U.S. families, including an oversample of low-income families headed mostly by Black householders. High wave-to-wave response rates and immigrant sample refreshers in 1997 and 2017 have helped maintain population representativeness over time. The PSID has a genealogical design. Children who are descended by birth or adoption from original householders become eligible to participate when they establish their own family units (i.e., economically independent households). In most waves, approximately 85% of young adults who have established their own family units since the prior wave respond to the PSID interview. The interview was conducted annually until 1997 and has been conducted biennially since then. It is conducted by telephone with a householder who is knowledgeable about the family's finances.1 The active sample includes just under 10,000 economically independent family units.
The family unit roster gathers information about all current family unit residents and those who left the family unit since the last interview, including dates of entry or departure for those who joined or left the household. Information on sex, age, and relationship to reference person (i.e., a primary householder) is reported for each family unit member. Since 1985, the respondent also reports on all marriages and births that occurred to family unit members aged 15 or older2 since the previous interview.
PSID Relationship Matrix
The PSID Relationship Matrix, a public-use data file, uses information from the family unit roster and birth and marriage histories to infer the relationships between each pair of coresident individuals in a family unit in each PSID wave3 and to characterize family unit composition at each wave from any family unit member's perspective, not only from the reference person's perspective. All available information from family unit rosters, birth and marriage histories, and internal records is used to definitively establish the relationships presented in the matrix file. As of the 2017 wave, the file includes more than 3 million records describing reciprocal pairs of family unit members and identifies 50 types of kin relationships, including biological, adoptive, step, social, and unrelated affiliations. Relationship type is determined for 99.92% of pairs (Panel Study of Income Dynamics 2021b).
We used the relationship matrix to describe the prevalence of coresidence and the frequency of change in coresidence with parents, siblings, stepkin, and extended kin from the perspective of minor children (ages 0–17) by decade from 1968 to 2017 (1968–1977, 1978–1987, 1988–1997, 1998–2007, and 2008–2017).4 We restricted the analysis to children whose families were included in the original 1968 sample or who joined through the 1997 immigrant refresher sample and who are descended from those householders through birth or adoption. The analytic file includes 256,956 person-year observations on 30,412 minor children.
We considered six mutually exclusive relationships to measure children's family composition: biological or adoptive parent,5 stepparent, biological or adoptive full sibling or half-sibling,6 stepsibling, grandparent or great-grandparent, and other relatives. Parents' cohabiting partners and their children who are not biologically related to the focal child are coded as stepparents and stepsiblings, respectively.7 In some cases, two or more family units that resided in the same physical dwelling may have been interviewed separately. We extended the family unit relationship matrix by using the PSID Family Identification Mapping System to locate instances in which a child lived in the same dwelling with a grandparent or an aunt/uncle who headed a different family unit. Thus, the indicators of coresidence with biological/adoptive parents, stepparents, and full, half-, or stepsiblings were limited to relationships in family units, but coresidence with extended kin included grandparents or aunts/uncles living in the same physical dwelling but in a separate family unit. (Aunts and uncles appear in the “other relatives” category.)
We describe patterns of kin coresidence overall and stratified by race and Hispanic ethnicity and family socioeconomic status. These characteristics pertain to the household reference person in the first year that a focal child was observed in a PSID family unit. (In most cases, the reference person is a parent of the focal child.) The PSID's measurement of race and Hispanic ethnicity has changed over time. To make equivalent comparisons across decades, we harmonized measurement and used the first racial/ethnic identity reported for the reference person, prioritizing Hispanic ethnic identity. The racial/ethnic categories are Hispanic, any race; non-Hispanic White; non-Hispanic Black; and non-Hispanic other race. We include all groups in the presentation of patterns in the population overall, but we exclude the non-Hispanic other race group from the presentation stratified by race and ethnicity because of small sample sizes (N = 268). To represent socioeconomic status, we used the family unit reference person's years of completed education collapsed into four categories: less than a high school diploma (0–11 years of completed education); a high school diploma (12 years); some college (13–15 years); and a college degree or higher (16 or more years). Analyses stratified by educational attainment excluded children for whom this information was missing about the household reference person (N = 1,819).
Two characteristics of the data affect estimation and representation. First, information about short-term coresidence that occurred between waves is not available, so estimates of kin coresidence prevalence and the coresidence transition rate are likely underestimated for some types of kin relationships. This limitation is more pronounced for data after 1997, when the PSID began biennial interviewing. A short-term coresidence module collected in 2003 demonstrated that 3.2% of family units included a short-term coresident in the two years since the last interview and that short-term coresidents were most often the householder's adult children (23%) or grandchildren (13%) or unrelated individuals (11%). Approximately 5% of short-term coresidents were the householder's cohabiting partner (Fomby and Chen 2021).
Second, the PSID does not capture family coresidence retrospectively. Hence, our analysis follows children from the time they are first observed in a PSID family unit. Children's post-1968 immigrant families added to the PSID in 1997 are followed prospectively without information about their coresident family composition before 1997. Consequently, pre-1997 results describe children whose families were in the United States since at least 1968; results from 1997 onward describe children whose families were in the country since at least 1997. Because the immigrant refresher added in 2017 does not have retrospective family composition data, those families are not included here.8 Despite this limitation, the results are largely consistent with recent estimates of kin coresidence derived from samples that are inclusive of recent immigrant families (Brown et al. 2016; Manning et al. 2014; Pilkauskas et al. 2020; Raley et al. 2019).
Methods
Patterns of Kin Coresidence
To describe the prevalence of children's family structure arrangements in each decade, we calculated as a period estimate the average number of kin of each type that children ages 0–17 lived with during the 10-year interval. For each minor child in each wave in a given decade, we used the relationship matrix to identify the child's relationship to every other individual in their family unit (or household in the case of extended kin) and summed the number of coresident kin of each type. We retained one record per child per wave to obtain the period estimate of the average number of coresident kin of each type in each decade. We weighted the numerator (total number of coresident kin of a given type among all child person-records in the decade) and the denominator (total number of child person-records in the decade) using children's longitudinal individual weights assigned at each wave. Weighted estimates are generalizable to children in families that have been resident in the United States since at least 1997.9 To evaluate statistically significant differences in estimates between decades, we constructed bootstrapped standard errors around each point estimate. Bootstrap samples were generated over 100 iterations, with standard errors clustered on time-invariant individual identifiers. Temporal differences in kin coresidence (two-tailed tests, p < .05) were evaluated between adjacent decades and between the first and last decades (1968–1977 vs. 2008–2017).
Transitions in Kin Coresidence
We estimated the number of coresidence entrances and exits by each kin type that children would be expected to experience between birth and age 17 if they were exposed throughout childhood to transition rates that were observed in each decade. This coresidence transition rate is analogous to population statistics such as the total fertility rate or life expectancy in that it is based on a synthetic cohort estimate of lifetime exposure to change if subjected to population patterns during a given period. Thus, these statistics should be interpreted as characteristics of populations in defined historical periods rather than as a reflection of a child's experience in the 18 years from birth to adulthood.
To estimate the coresidence transition rate, we obtained the age-specific probability of experiencing the entrance or departure of any family member of each kin type from a child's family unit roster between two waves for children who were aged 0–17 in the later wave. These rates are annual probabilities for survey years up to 1997 and probabilities over two years for 1998–2017. We summed age-specific probabilities of recent transitions at each wave to obtain an annual or biennial coresidence transition rate for each kin type. We then summed these annual or biennial rates across all waves within a given decade and divided the sum by the number of years in the period. Because 1969 was the first year to observe any change compared with the prior year, the first period included nine years (1969–1977).
Year-to-year fluctuations in some types of infrequent family change introduced noise into the estimated coresidence transition rate; that is, even small absolute change on a small base implied large changes in family structure that were more likely a result of sampling error. To smooth these estimates, we used a three-year rolling average from birth to age 15 for our age-specific probabilities. For example, for a child who was 1 year old in 1969, we used the average of age-specific probabilities for children who were 1 to 3 years old at that wave. We applied this approach for children up to age 15 (using the rolling average for children aged 15–17). For children who were 16 or 17 years old, we used their singulate age-specific probabilities.
Results
Unweighted and weighted sample characteristics are summarized in Table 1. The unit of analysis is person-year observations for minors aged 0–17 between calendar years 1968 and 2017. Three features in the weighted statistics warrant particular attention. First, the distribution of the reference person's (i.e., the householder's) educational attainment changed over time. In the earliest decade (1968–1977), about 40% of child person-years were spent with a reference person who had not completed high school, and 14% resided with a reference person who had completed at least four years of college at the time of the child's birth. By the most recent decade, nearly half of person-years were spent with a reference person who had completed at least some college. Thus, when considering evolving patterns of kin coresidence by householder's educational attainment, it is important to remember that the composition of families at each level of achieved education also changed. Second, the proportion of child person-years spent residing with a Hispanic householder increased in 1998–2007 compared with earlier decades as a result of the addition of an immigrant refresher sample in 1997, which added post-1968 immigrant families to the PSID in proportion to their representation in the U.S. population. (These proportions declined in the last period because of higher rates of sample attrition not fully accounted for by the sampling weights.) Third, the number of person-years is smaller in the last two decades than in the first three decades because of a planned sample reduction and a shift from annual to biennial interviewing in 1997.
Table 2 summarizes period estimates of coresidence prevalence and entrance and exit transition rates for each type of kin in each decade. Estimates are presented for the sample overall and by the reference person's educational attainment and race and Hispanic ethnicity. Values in italics indicate a statistically significant difference compared with the prior decade. Values in bold in the last set of columns denote that estimates in the most recent decade are significantly different from estimates in the first decade (two-tailed tests, p < .05). Superscripts describe within-period group differences by the reference person's educational attainment and race and Hispanic ethnicity.10
Table 3 complements Table 2 by describing coresident kin prevalence in each decade in categorical rather than linear terms. In particular, it describes the proportion of child-years in each decade shared with zero, one, or two coresident biological or adoptive parents; zero, one, or two or more biological/adoptive siblings; any stepparent; any stepsibling; any grandparent; and any other relative. Results stratified by the reference person's educational attainment and race and Hispanic ethnicity are available in the online appendix.
Biological and Adoptive Parent Coresidence
The first row of Table 2 shows that in the earliest observed decade (1968–1977), children aged 0–17 years lived with 1.78 parents, on average. The average number of coresident parents declined in each of the next two decades and then plateaued in the late 1990s, reaching 1.64 in 2008–2017. Table 3 shows that this decline was nearly entirely attributable to a declining prevalence of coresidence with two parents and an increasing prevalence of living with one parent. In the earliest decade, 80% of child-years were spent living with both parents, and 17% were spent living with one parent. By the most recent decade, these figures were 66% and 32%, respectively. Consistently across decades, only about 2% of child-years were spent with neither parent in the family unit.
Over the same period, the coresidence transition rate for parent entrances and exits rose over the first three decades and declined during the two most recent decades, especially for parent exits (Table 2). Specifically, under the transition rates observed in 1968–1977, children were expected to experience 0.24 parent entrances11 and 0.78 parent departures by age 17. These values peaked at 0.42 and 0.93, respectively, in the late 1980s and 1990s; they then tapered to 0.38 and 0.69 in the most recent decade. Changes in period estimates of coresidence and the coresidence transition rate imply that contemporary children reside with fewer parents compared with children in the late 1960s and 1970s but also are expected to experience fewer parental departures. In between, the greatest turbulence occurred during the late 1980s and 1990s, when parent coresidence was sharply lower and parent entrances and exits were sharply higher than in the prior decade.
Figure 1 shows the trends in the average number of parents a minor child lived with during each 10-year period with the line graph (left axis) and the parent coresidence transition rates for entrances and exits with the bar graph (right axis). Entrances fall above the x-axis, and exits fall below it. The scale of the bars is marked on the right y-axis. The figure highlights the plateau in the number of coresident parents beginning in 1988–1997 as well as the U-shaped pattern in the number of parent exits over the 50-year period.
The panels in Table 2 characterizing family composition by the reference person's educational attainment and race and ethnicity align with expected patterns of social stratification. With each additional level of educational attainment, children reside with more parents, on average, and experience fewer parental entrances or exits compared with peers in family units where the reference person has less education in a given decade. In the most recent decade, for example, children in a family unit with a college-educated reference person resided with an average of 1.85 parents, compared with an average of 1.54 parents among peers in family units with a high school–educated reference person (p < .05). Non-Hispanic White and Hispanic youth were more similar in parental coresidence compared with non-Hispanic Black youth (1.72, 1.64, and 1.32 coresident parents, respectively), but non-Hispanic White and Black youth were more similar with regard to fewer expected parental entrances and exits compared with Hispanic youth (0.68, 0.66, and 0.83 parental exits, respectively). (However, most group differences are statistically significant at p < .05 for each decade.) Despite these level differences, subgroups followed the general pattern of change across the decades described earlier, with parental coresidence declining through the late 1990s and then stabilizing, parental entrances increasing across the decades, and parental exits peaking in the late 1990s before returning to rates observed in the earliest decade. Children in families headed by a high school graduate were an exception to this general pattern: in this group, the number of coresident parents continued to decline across all decades, rather than leveling off after 1997.
Figure 2 presents trends by each category of educational attainment. Panel a displays period estimates of the number of coresident parents, and panel b summarizes expected parent entrances and exits. Ninety-five percent confidence intervals bound each estimate. The figure highlights the emergent educational gradient in coresident parent size across decades. It also shows, however, that growth in the magnitude of the gradient associated with the coresidence transition rate gradient declined and then plateaued after 1997.
Sibling Coresidence
Consistent with recent research (Fahey 2017) and reflecting changes in the U.S. total fertility rate, children's average coresident sibling size declined from an average of 2.35 full siblings or half-siblings in the earliest period to 1.50 siblings in the most recent decade, reaching its smallest size at 1.37 siblings in 1988–1997 (Table 2). Table 3 shows that this change is attributable to an increasing prevalence of child-years spent with no siblings or with one sibling and a declining prevalence of years spent with two or more siblings—a shift that was most pronounced between 1968–1977 and 1978–1987, when the share of child-years spent with two siblings declined from 61% to 42%. Over the same period, the share of child-years spent living with one sibling increased from 26% to 39%, and the share with no siblings increased from 13% to 19%.
The number of siblings entering children's family units also declined through the third decade and then plateaued, implying that the number of younger siblings expected to join children's families by age 17 held mostly steady over the last 30 years. In contrast, the number of expected sibling departures from children's households consistently declined, dropping from 2.11 in the first decade to 0.75 in the most recent decade. Together, these three patterns—a U-shaped trend in the number of coresident siblings, a plateau in expected sibling entrances, and a steady decline in expected sibling departures—suggest that children today have fewer siblings than they have had historically but are expected to spend more of their childhood living with the siblings they have, particularly with older siblings who delay or abbreviate their own departures from the family household. Figure 3 summarizes these trends graphically.
This general pattern holds across racial and ethnic groups and across all levels of educational attainment. However, the magnitude of change in coresident sibling size and the coresidence transition rate—particularly declines in the exit rate—is most pronounced among children with householders who were Black or had less than a high school education. These two groups had the largest sibship sizes in the earliest decade. Change over time is less pronounced among children in Hispanic families.
Family Complexity
In all five decades, coresidence and transitions in coresidence with stepparents or stepsiblings were much less frequent compared with coresidence with biological and adoptive parents and siblings. Over time, though, coresidence with stepparents and, to a lesser extent, stepsiblings increased as a type of family complexity. In the first decade, children resided with an average of 0.02 stepparents and 0.01 stepsiblings, and this prevalence level more or less held across educational attainment and racial/ethnic subgroups. Children with non-Hispanic White householders or householders with less than a college education had higher coresidence transition rates than their peers in non-Hispanic Black, Hispanic, or college-educated households. The prevalence of coresident stepkin increased over the next two decades, rising to 0.08 stepparents and 0.02 stepsiblings in 1988–1997 and plateauing at 0.09 and 0.04, respectively, during the next two decades. The late 1980s and 1990s were also marked by the most turbulence in stepparent transitions: rates observed in that decade suggest that children were expected to experience 0.36 entrances and 0.43 exits by stepparents by age 17. The entrance and exit transition rates for stepsiblings, in contrast, more than doubled in 1978–1987 compared with the prior decade and stayed roughly the same through 2008–2017.
Increases in coresident stepparent prevalence and stepparent transitions were strongly graded by householder's educational attainment. Children living with a householder who had less than high school education resided with 0.13 stepparents in 2008–2017 and were expected to experience 0.42 entrances and 0.43 exits by age 17, compared with (respectively) 0.03 coresident stepparents and 0.16 entrances and 0.14 exits among children living with a college-educated householder in the same period. These patterns are consistent with higher rates of childbearing in cohabitation, union dissolution, and new union formation over the last 40 years among adults with less than a college education compared with adults who have at least a four-year college degree. Group differences by race and ethnicity in each decade were smaller in magnitude. Children in non-Hispanic White families had slightly higher prevalence rates for stepparent coresidence and similar entrance and exit rates after 1977 compared with non-Hispanic Black children (most within-period group differences are statistically nonsignificant).
Notably, the strong educational attainment gradient for stepparent coresidence is absent with regard to stepsibling coresidence. This finding likely reflects that children born to unpartnered parents most often live with their mothers and that despite significant increases in the prevalence of postdivorce joint custody arrangements, most children continue to live with mothers at least half-time. As a result, when fathers enter new unions, children from their prior unions rarely experience full-time coresidence with them (Arsenault and Stykes 2019). Thus, to the extent that the presence of coresident stepparents signals the presence of stepsiblings living elsewhere, the current focus on coresidence potentially underestimates a social gradient in children's stepsibling relationships, a topic requiring further research with data on children's extrahousehold family relationships.
Household Complexity
Coresidence with grandparents (including a small number of great-grandparents) and with other relatives represents children's exposure to household complexity. The late 1960s marked a low point for U.S. children's extended kin coresidence as young families took advantage of an expanding housing market to establish households independent of parents (Pilkauskas et al. 2020; Ruggles 2007). Since then, the number of expected coresident grandparents more than doubled, from 0.06 in 1968–1977 to 0.14 in 2008–2017. The grandparent entrance rate mostly held steady since 1978–1987, but the exit rate declined (0.42 exits in 1978–1987 vs. 0.25 exits in 2008–2017, p < .05). This pattern suggests that while the incidence of grandparent coresidence increased over time, its duration during childhood also increased—that is, contemporary children who live with grandparents are expected to do so longer than children in earlier periods, when grandparent coresidence most often ended by the time children entered school (Kreider and Ellis 2011). The prevalence of coresidence with other relatives—most often, aunts, uncles, and cousins—remained more consistent (0.12 coresident other relatives in 1968–1977 and 0.13 in 2008–2017, a nonsignificant difference).
Change in the prevalence of grandparent coresidence is socially graded by educational attainment. Children with a householder with less than a high school education had the highest rates of grandparent coresidence across decades (0.11 coresident grandparents in 1968–1977 vs. 0.25 coresident grandparents in 2008–2017). In families where the householder had a high school diploma or more education, children resided with approximately 0.03 grandparents in the earliest decade. Over time, a social gradient emerged among children in these families: in 2008–2017, children with a high school–educated householder resided with an average of 0.16 grandparents, compared with 0.11 for children with a householder with some college and 0.05 for children with a college-educated householder (all group comparisons p < .05). The gradient is similar for coresidence with other relatives. Patterns in the grandparent and other relative coresidence transition rates were variable across educational attainment categories, but expected entrances and exits generally peaked between 1978 and 1997 for all groups.
With regard to race and ethnicity, children living with non-Hispanic Black householders experienced the most consistency in coresident grandparent prevalence and the largest declines in other relative prevalence over historical time (0.19 grandparents and 0.43 other relatives in 1968–1977 vs. 0.17 grandparents and 0.19 other relatives in 2008–2017). In contrast, children with non-Hispanic White householders tripled the count of coresident grandparents over the same period (0.04 in the first decade vs. 0.12 in the last decade), and children with Hispanic householders nearly doubled that count (0.09 and 0.17, respectively). The prevalence of coresidence with other relatives for children with Hispanic householders peaked in 1988–1997 (0.29 other relatives) but was mostly consistent across decades. It was generally low for children with White householders across decades until a significant increase in the most recent decade (0.09 in 2008–2017 vs. 0.05 in 1998–2007). Overall, these patterns indicate a substantial change in the racial/ethnic distribution of extended kin coresidence, with increasing similarity in grandparent coresidence, although group differences compared with non-Hispanic White families remained statistically significant in the most recent period.
Figure 4 summarizes these patterns. Panel a shows the increasing similarity in the prevalence of coresident grandparents across racial and ethnic groups. Panel b illustrates that grandparent coresidence entrance and exit rates declined in the two most recent decades among children with non-Hispanic Black or Hispanic householders but mostly held steady for children with White householders. Together, these results suggest different pathways to the current relatively high rates of grandparent coresidence by race and ethnicity.
Discussion
We provide a comprehensive portrait of the increasingly diverse and dynamic nature of U.S. children's family organization over the last half-century. We offer three innovations to build on a substantial body of prior research that has documented growth in the prevalence of varied family forms and become increasingly nuanced in the measurement of children's family structure change. First, we consider patterns of historical change in coresidence with multiple types of kin simultaneously, including biological/adoptive parents, siblings, stepkin, grandparents, and other relatives. Second, we consider both the prevalence of types of kin coresidence and the coresidence transition rate to measure the expected amount of kin coresidence change children experience from birth to age 17. These measures together characterize trends in the dynamic composition of children's family structure arrangements. Third, we apply this approach to the PSID relationship matrix, a novel data resource that enables the first profile of children's dynamic and multifaceted family structure in each decade over 50 years (1968–2017), beginning from a historical moment that was distinctive for its high levels of two-parent nuclear family organization and large family sizes. These data also permit subgroup analysis by educational attainment and race and ethnicity in each decade.
We offer five main findings. First, consistent with prior research, children's households increasingly diverged from two-parent headship and became increasingly complex (Carlson and Meyer 2014; McLanahan 2004). Reflecting a plateau in the proportion of children living with unpartnered parents (U.S. Census Bureau 2019), the drop in the number of children's coresident parents tapered off after the 1990s, but growth in the prevalence of both family complexity (as measured by coresident stepkin, particularly stepparents) and household complexity (as measured by coresident grandparents and other relatives) continued to the most recent decade (Manning et al. 2014; Pilkauskas et al. 2020). Second, instability in children's households—the expected number of family members' household entrances and departures that children are expected to experience by age 17—peaked around the end of the twentieth century and declined in each decade since then. This pattern is consistent for each kin type.
We propose that these two findings illustrate that contemporary family organization is gravitating toward “stable complexity,” a term we coined to describe household kin's varied legal, social, and biological relationships to children in a relatively durable family household structure compared with prior decades. This view is consistent with recent evidence of a decline in the national divorce rate (National Center for Health Statistics 2020) and a plateau in parents' union instability and the prevalence of coresident family complexity since the 1990s (Brown et al. 2016; Manning et al. 2014), as well as a leveling off of the nonmarital fertility rate and share of births occurring outside of marriage in the last decade (Martin et al. 2019). This historical perspective on status and change in children's families can help to inform future research about the pace and evolution of the second demographic transition (Lesthaeghe 2020; McLanahan 2004; van de Kaa 1987) and to investigate the economic, cultural, and structural precursors of this apparent plateau in the shift toward a new distribution of varied family forms.
A third finding also related to household stability is that contemporary children are expected to have experienced fewer household departures by siblings in recent compared with earlier decades. We interpret this to mean that children are spending an increasing share of time coresiding with older siblings who have not yet left or who have returned to the parental household as young adults. Much research has investigated the causes and consequences of delayed or abbreviated home-leaving for young adults (Hill and Bosick 2017; Houle and Warner 2017; Lei and South 2016; White 1994). To date, however, there is not a wide body of work on the question of how these slower departures affect younger siblings. Do older and younger siblings compete for resources, or do adult children contribute resources to families to their younger siblings' benefit? Do young adults provide quality peer relationships for their younger siblings? How does the parenting of younger children change when coresident adult children exhibit greater autonomy? The circumstance of delayed home-leaving offers potentially important insights to the growing literature on sibling relationships and dynamics.
Fourth, trends in stable complexity vary by social class (measured by householders' educational attainment). Although all children became increasingly likely to live in complex families and to have experienced the highest coresidence transition rates during the 1980s and 1990s, levels of complexity and frequency of transition are consistently higher among children in families where the householder has not completed college. Further, although the drop in the number of coresident parents tapered off in the 1990s overall, it continued to decline for children with householders who had a high school education or less through the most recent decade. In part, these patterns may be attributable to compositional changes in the population of less educated adults. However, they also likely reflect a family-level response to the 50-year trend toward wage stagnation, declines in full-time stable employment, and an increasing share of jobs in the service sector—trends that have been particularly consequential for men who lack a college education (Cherlin et al. 2016; McLanahan 2004; Western et al. 2016).
Fifth, trends by race and ethnicity indicate increasing similarity in some forms of family organization, although group differences largely remain statistically significant. In particular, children in non-Hispanic Black households continued to be less likely to reside with two parents than children in non-Hispanic White households throughout the 50-year period, but the two groups grew increasingly similar in their expected parental coresidence transition rates and the prevalence of coresident stepparents and stepsiblings. Non-Hispanic White children also experienced substantial growth in the average number of coresident grandparents and other relatives, whereas the prevalence of these kin in the households of non-Hispanic Black children held steady or declined. Children in Hispanic households in the most recent decade had levels of family and household complexity that were similar to those of non-Hispanic Black children, but they were expected to experience more parental coresidence transitions (entrances and exits) than non-Hispanic Black or White children. To date, much research has focused on what drives racial and ethnic differences in family composition or the extent to which those differences contribute to disparities in child outcomes (Cross 2018; Dunifon and Kowaleski-Jones 2007; Mollborn et al. 2011). Our research highlights dimensions on which non-Hispanic White, non-Hispanic Black, and Hispanic children's family composition are increasingly similar, potentially offering a new avenue of research to consider whether and how the antecedents to family organization have changed within and across racial and ethnic groups.
In sum, this work offers a novel and informative view of coresident family change over historical time from the perspective of children. By documenting the emergence of a diversity of family composition arrangements and patterns of family and household instability, it lays the groundwork for research into the long-range economic and social determinants and consequences of children's coresident family organization.
Acknowledgments
We are grateful to Wendy Manning, Natasha Pilkauskas, Narayan Sastry, and three anonymous reviewers for helpful comments on previous versions of this manuscript and to Meichu Chen, Flannery Campbell, and Carissa Scurlock for outstanding data preparation and programming assistance. All errors and omissions are the responsibility of the authors. This work was supported by research grants from the Eunice Kennedy Shriver National Institute for Child Health and Human Development (R01HD088506) and the National Science Foundation (1623684).
Notes
Between 1968 and 1972, the interview was conducted in person.
Before 2013, information on recent births was collected about other family unit members aged 12 or older.
The relationship matrix includes all individuals who are present in the family unit at a given wave. It does not include family members who are incarcerated, in the military, in long-term medical care, or living at an educational institution (e.g., boarding school or college).
We do not present information on nonkin. Because the PSID is a study of families rather than a study of households, coresident nonkin who do not contribute economically to the family unit are not included in the family unit roster. As a result, information on nonkin is not exhaustive and is biased toward “kin-like” relationships. In the observed period, children lived with an average of 0.01 rostered nonkin and were expected to experience approximately 0.05 entrances and 0.05 departures by rostered nonkin between birth and age 17.
Adoptive parents include stepparents who formally adopted their spouse’s biological child at any point.
We combine full siblings and half-siblings because in most years, the PSID birth history did not collect information about a child’s second parent if that parent was never present in the family unit. As a result, we cannot discern whether two children whose second parent was never present had the same pair of parents and thus cannot definitively distinguish full siblings from half-siblings in all sibling pairs.
Before 1983, the Survey Research Center staff inferred cohabiting status of a family unit head and their partner through a combination of the partner’s position on the family unit roster, the head’s reported marital status, and any available interviewer notes (personal correspondence).
The 2017 immigrant refresher represents approximately 12% of U.S. children whose families entered the United States since 1997 (Ruggles et al. 2020; the authors’ estimates from the 2013–2018 American Community Survey).
Individuals born after sample recruitment who are descended from a PSID householder but who were not observed in a PSID family unit for more than one wave after their birth do not have an assigned longitudinal individual weight before 1994. We imputed weights for these individuals by assigning the average value of the longitudinal individual weights of the reference person and their spouse or partner in the affected individual’s family unit in a given wave. If no spouse or partner was present, the imputed weight was equal to half the value of the reference person’s longitudinal weight (Gouskova et al. 2008).
Bootstrapped standard errors are available from the authors upon request.
Parental entrances occur through three main pathways: (1) most often, through a parent’s initial entrance into the child’s household shortly after the child’s birth; (2) less often, through a parents’ return to the family unit following deployment or incarceration or as a result of parents’ reunion; or (3) initially as a stepparent who eventually becomes an adoptive parent.