Introduction

All demographic events occur in families. The motivation for having a child, the consequences of losing a parent or spouse, and when people move to new locations (whatever other economic and social factors are involved) are all family experiences. Individuals’ health depends on the transmission of genes from parent to child. Individuals rely on family members to care for them if they become ill and to help them remain healthy once they are better. People move into and out of the labor force and adjust their hours worked to fulfill family responsibilities. They move to new locations, including across national borders, to contribute to their families’ economic welfare. Family members share resources and ameliorate the risks of economic uncertainty.

“Family” has colloquial meaning. When the word is used in conversation, the people involved understand its meaning. But “family” is an elusive concept. Views about what a family is may be contested political terrain. Family is also a scientific concept. Scholars compare families across cultures and in different historical periods to learn the causes of change in what families look like and what they do. It is impossible to consider family change without considering the demographic changes in families—for instance, changes in the numbers of children people have, the timing and duration of marriages, and the overlapping lives of parents and offspring.

The field of family demography is relatively new, although demographers have long considered families and family processes. I argue that all demographers—not just family demographers—should pay more attention to who is in a family and to the distinction between who lives together and who is a family member. To improve knowledge about what families do, demographers should study not only behavior but also attitudes, feelings, and preferences.

Demographers have made considerable progress in learning about family and household composition, but there is much more to learn. Changes in the demography of U.S. families illustrate the relevance of new questions about families. Such changes include increases in the number of generations alive at the same time, the prevalence and exposure to informal unions (cohabitation), and the percentage of children born outside of marriage.

The percentage of children who have all four biological grandparents alive has increased substantially since 1900. Only 6 % of 10-year-old children in 1900 had all their grandparents alive. In contrast, 41 % had all four biological grandparents alive by 2000, and estimates suggest that almost half will still have all four grandparents alive by 2020 (Uhlenberg 2005). Individuals have more living vertical kin ties, with more three- and four-generation families, than ever before because of improvements in life expectancy (Bengtson 2001). Advances in health mean that older generations can lead more active lives and, as a result, there may be a wider range of types of social interactions among the generations, such as grandparents helping adult children and grandchildren with routine household tasks and grandparents guiding grandchildren’s development, instead of interaction being limited to a younger generation caring for the infirm older generation (Margolis and Wright 2017).

Changes in couple relationships also motivate new questions about who is in a family. Cohabitation has increased dramatically since the late 1970s. In 1977, fewer than 1 million U.S. couples were cohabiting, but by 2017 there were nearly 8 million (U.S. Census Bureau 2017c). Today, almost two thirds of U.S. women have ever been in a cohabiting relationship (Hemez and Manning 2017). Although cohabitation is more common than it used to be, cohabiting unions remain unstable, with most dissolving after a short time (Copen et al. 2013).

The percentage of children born outside of marriage has also increased in the United States. In 1970, about 1 out of 10 births occurred outside of marriage. By 2016, that fraction was 4 out of 10 (Martin et al. 2018). More than half of nonmarital births are to cohabiting couples (Curtin et al. 2014), but significant racial/ethnic differences exist in the union context of births (Sweeney and Raley 2014).

Cohabiting unions are much more common among women with less than a high school education than among women with more schooling, especially for those with college degrees (Hemez and Manning 2017). This is important because the educationally disadvantaged are also disadvantaged by instability in their partnerships (Musick and Michelmore 2015).

Taken together, these demographic facts may change how families operate. Greater life expectancy implies greater potential for family ties across generations. At the same time, however, family ties have become more unstable because of high rates of cohabitation, the instability of cohabiting unions and marriage, and the growing separation of marriage and childrearing. Americans are the marrying kind, as Cherlin (2009) has noted. They remarry or cohabit after a first relationship ends. Repartnering among parents means that families have become increasingly likely to include children from different relationships—stepchildren or quasi-stepchildren through cohabitation (Guzzo 2017; Wachter 1997). The diverging family destinies that Sara McLanahan talked about in her 2004 presidential address to the Population Association of America (PAA) imply that children whose mothers are highly educated have stable family lives, and children whose mothers have little formal schooling have unstable family lives. The divergence remains a feature of family life in the United States as well as a number of other high-income countries (McLanahan 2004; McLanahan and Jacobsen 2015; Thomson et al. 2014). This divergence may exacerbate inequality because families act as a private safety net for individual family members. Instability in family ties threatens families’ effectiveness in fostering the health and well-being of the next generation and threatens families’ ability to support vulnerable family members.

Demographers need new approaches to understand who is in the family and what families do. In the remainder of the article, I briefly discuss why family is important to demography and what demographers mean by “family,” and then turn to key historical developments in family demography that illustrate principles that demographers should incorporate in new research. I take stock of findings in family demography that motivate new research on families from the perspective of individuals’ family relationships. Although I focus on the demography of U.S. families, the conceptual challenges I identify are ones that matter for everyone who studies families.

Why Family Is Important

Families bear and rear the next generation. Family income, parents’ education, and childrearing practices affect children’s nutrition, health, and the kind of people they grow up to be (see, e.g., Amato and Fowler 2002; Heckman 2006). Stable family ties affect the kinds of family lives children have when they become adults and influence their economic welfare (Bloome 2017; Cherlin et al. 1995; Sassler et al. 2009; Wu 1996). Adults who grew up in disadvantaged families face limited economic opportunities (Blau and Duncan 1967; Duncan et al. 2010), which may make it difficult for them to “achieve” the status of being married, as some single mothers describe it (Edin and Kefalas 2005). The long-term effect of the differences in family experiences and economic resources across generations may contribute to economic inequality because family members pool income (Bloome 2014; McLanahan and Percheski 2008).

Family members provide unpaid labor to improve each other’s lives. Grandparents provide childcare, spouses care for each other, and adult children help aged parents (Compton and Pollak 2014; Fingerman et al. 2016; Luo et al. 2012; McGarry 1998). People can pay others to do this work, but family members often have a comparative advantage because they know each other’s preferences well and—usually—trust each other (Brody et al. 1984; Burton et al. 2009; Pollak 1995).

Demographers in diverse areas within the field take account of the importance of family, even if they do not necessarily call themselves “family demographers.” This is evident in the PAA conference program each year. Because family is important to a wide range of demographic topics, demographers and social scientists in general should do a better job of collecting and analyzing data on families.

Notions of Family

Fertility Versus Family

Demographers initially focused attention on families because families produced and raised children. Early U.S. surveys defined family narrowly. The Growth of American Families (GAF) study in 1955 assessed young white married women’s fertility-related attitudes and behavior (Freedman et al. n.d.). Although highly innovative at the time, the GAF was clearly a study of fertility and not of family. The use of “family” in the title of the study was more than public relations. Demographers themselves equated family and fertility. Arthur Campbell, one of the GAF investigators, wrote in 1965 that by ages 25–30, most “married couples consider their families complete [emphasis added]” (Campbell 1965:20).

Yet even in the 1960s, families were not complete at such an early age. The offspring of these couples would grow up, and many of them would marry and have children of their own. This new generation—the future grandchildren of the women interviewed in the GAF—would surely have been considered family members by their grandmother. Today, with the increased chance that grandparents and grandchildren will be alive at the same time, demographers should reorient their work to consider a broader range of who is in the family.

The current National Survey of Family Growth (NSFG), which is the modern incarnation of the GAF, describes a much broader population and covers a much wider range of behaviors, including same-sex sexual behavior, than the growth histories of the GAF. But the NSFG is still based on a sample of reproductive-aged adults, mainly measures characteristics of heterosexual families, and focuses on relatively young families (National Center for Health Statistics 2016). The NSFG data are extremely valuable for many purposes. But fertility is not the only form of family growth that family demographers study.

A Broader Notion of Family

“Family size” is no longer just the number of children a woman bears. Today, demographers typically treat family size as a much broader concept. Demographers consider how many generations are alive at the same time (Bengtson 2001; Park et al. forthcoming); the number of siblings and cousins available as peers; and the number of different parents who are connected or potentially connected to each child—biological parents, stepparents, and the cohabiting partners of a biological parent (Keyfitz 1985; Verdery 2015).

I adopt this broader notion of family. Family is a social institution with roles defined by long-term rights and responsibilities. Some roles originate in biological connections, but it is the social understanding of these roles that contributes to temporal and group variation in family processes. Family roles may be codified by law, as in laws about who can marry or laws about paternity, but family roles are much more than roles governed by laws. Individuals internalize social understandings of family roles and use the understandings to guide their behavior. This notion of a family does not include any requirement for love or intimacy because historians and social scientists have shown that the extent to which love, intimacy, and privacy define family connections varies cross-culturally (Carsten 2000; Coontz 2005).

Families are made up of relationships between individuals. Families are also collections of individuals whose lives are linked to form units (Burch 1979; Glick 1959). These are complementary ways of thinking about families, as depicted in Fig. 1.

Fig. 1
figure 1

Family relationships and family units. Lines connecting individuals indicate family relationships

Full size image

Couples are linked to each other and to their offspring, as shown in panel a of Fig. 1. Siblings, children who share the same parents, are linked. Grandparents are connected to a grandchild, and the siblings of a parent in the middle generation (aunts and uncles) are also connected to this child. Ties are also created through marriage or conjugal ties—colloquially, “in-law” relationships. Important anthropological work has shown that societies vary in the degree to which people recognize all these connections as family relationships. Societies also vary in the strength of the connections between individuals (Queen et al. 1985).

Families can also be thought of as units. The set of relationships in each of the circles in panel b of Fig. 1 is a family unit. The parents in the middle family unit are members of different family units as well as their shared family unit. Because people in one family are simultaneously part of different families, people in the same family may have different interests. Sometimes researchers try to combine the family-as-relationships and family-as-unit approaches by treating family as a network of relationships and using information about the network boundaries to outline the boundaries of the family unit (Finch 1989; Heady 2010). But this does not circumvent the problem that people in the same family have different family relationships that sometimes pose competing demands and may put family members at odds with each other.

I argue that to better understand families, demographers should focus on individuals’ family relationships. By focusing on individuals’ relationships, we also learn more about households and living arrangements. By studying individuals’ family experiences, we can better incorporate the real-life ambiguities about who is a member of a family and who lives in a household and thereby learn more about what families do and what helps or hinders them.

Key Innovations in the History of Family Demography

This highly selective timeline of developments in family demography ignores the significant work in historical demography and emphasizes developments in the United States (e.g., Dong et al. 2015; Laslett and Wall 1972; Ruggles 1990; Wrigley and Schofield 1981).

Family has always been part of demography, dating back to work by Thomas Malthus in 1798. Malthus referred to expectations about supporting a family as a check on fertility (Malthus 1798/1970). Family demography is much newer than 1798, though.

Although some population researchers were doing family demography before the 1960s, family demography entered the scene organizationally in 1961. That was the first time the International Union for the Scientific Study of Population (IUSSP) and PAA included a special session on families and households at the international population conference held in New York (IUSSP 1963:155; Wargon 1974). Since then, theoretical, data, and methodological innovations have improved our understanding of families and households. These innovations also illustrate principles demographers would do well to adopt in new research on families.

Theoretical Developments

Economic theories of family behavior influence the work of family scholars in all fields, not just in economics. In the early 1970s, Theodore W. Schultz organized two conferences on the New Home Economics, with support from the National Bureau of Economic Research and the Population Council. The New Home Economics applies economic principles of production, consumption, supply, and demand to family behaviors. Family members make decisions about how to allocate resources. These resources include time and money, which are sometimes substitutes and sometimes complements. Families—and households—are both producers and consumers (Schultz 1973).

These conferences had wide-reaching effects for three reasons. First, the participants addressed a broad set of topics—marriage, fertility, investments in children, the value of time, and household decision-making. Second, the conferences involved sociologists and demographers as well as economists. Third, the empirical orientation of the presented papers fostered closer connections between economists and researchers in other disciplines.

Since the Schultz conferences, Gary Becker (1981) published A Treatise on the Family, which consolidated theoretical understandings of how families operate within and between generations. These theoretical advances in economics continue to evolve in ways that have helped us understand families better, especially the unpaid work of family members, mainly women.

From the disciplines of sociology, human development, and developmental psychology, family demographers learned to consider the interactions between what happens in one individual’s life and what happens in another’s when those people are connected by family ties. In 1974, Glen Elder published Children of the Great Depression, a major step in his articulation of the importance of the interaction of age and historical period (Elder 1974). Demographers were already well aware of age, period, and cohort effects (for a review, see Hobcraft et al. 1982). But work by Elder and his collaborators brought these ideas to the study of the linked lives of family members, the idea that what happens in one person’s life affects and is affected by what happens in the other person’s life because they have (some) shared interests. Psychological and social observations combined with longitudinal data on individuals and their families point to the importance of considering early-life conditions and individuals’ attitudes as well as behavior to better understand subsequent well-being (e.g., Doty and Mortimer 2018; Gaydosh and Harris 2018).

Data Innovations

Three data innovations also contributed to the development of the field of U.S. family demography. First, in 1968 the Panel Study of Income Dynamics (PSID), initially called Five Thousand American Families, began. The PSID was designed—and continues today—to provide data on the factors affecting changes in families’ economic welfare (McGonagle et al. 2012; Morgan et al. 1974). Its remarkable features are that it follows and interviews all members of the original 1968 coresiding families and their descendants, documents changes in household composition (e.g., due to divorce and adult children leaving home), measures individual household members’ characteristics, and collects detailed economic information. The PSID design incorporates the adult offspring of original PSID respondents into the sample, and this genealogical design enables researchers to reconnect parents, offspring, and often grandchildren to other family members even if those family members do not live together.

Second, more complete measurement of family relationships within households also improved our understanding of individuals’ family experiences. In the late 1970s, Paul Glick and Arthur Norton began to identify cohabiting couples by inferring that Persons of the Opposite Sex Sharing Living Quarters (POSSLQs) were a couple (Glick and Norton 1977/1979). This initiated improvements both outside and within the U.S. statistical system. We can now directly identify cohabiting partners and link parents to each child living with them through parent pointers and relationship grids (Brandon and Bumpass 2001; Casper and Cohen 2000; Kennedy and Fitch 2012; Kreider 2008; Manning et al. 2014).

Direct identification of partners also has dramatically improved knowledge of same-sex partnerships, marriages, and families. Same-sex cohabiting couples were first identified in the 1990 census when the Census Bureau revised the question about the relationship to the householder to include “unmarried partner” as a response option. This change allowed same-sex partners to be distinguished from same-sex roommates (Black et al. 2000). Since then, census and other federal surveys have collected data that enable researchers to describe coresident families of same-sex couples who live together, including whether the couples have children in their households. Since 2005, census data have distinguished same-sex married couples from same-sex cohabiting couples (U.S. Census Bureau, Fertility and Family Statistics Branch 2013). Census surveys do not currently ask about sexual orientation, but the NSFG and surveys outside the federal statistical system do. This information provides a denominator for estimates of the percentage of those who identify as gay or lesbian who are in same-sex cohabiting unions or married, and the data support efforts to understand variation among the LGB population as well as between those who are LGB and those who are not (Gates 2011).

A third innovation in data occurred when Larry Bumpass and James Sweet designed the National Survey of Families and Households (NSFH) (Sweet et al. 1988), which is unique in the United States for its holistic coverage of multiple domains of family life. The survey asked about housework, paid work, work schedules, time with children, and family activities. It rostered parents, children, stepparents, and stepchildren. The survey asked questions about family members in and outside of the household. The design addressed “his” and “her” family because each spouse or cohabiting partner in opposite-sex unions reported their own attitudes.

Valuable Methods for Family Demography

In 1987, the same year the NSFH was fielded, Bongaarts et al. (1987) published a collection of papers demonstrating the value of multistate life tables and simulations for family demography. These tools allow researchers to describe individuals’ cumulative exposure to different family experiences—for instance, children’s years in single-parent households and the expected numbers of family members under different demographic regimes. Simulations provide answers to questions for which existing data are insufficient, such as what percentage of children have living grandparents, and to predict the future by addressing the likelihood that adults will have stepfamily members in one or more generations.

Growth in Publications on Family Topics

These developments informed publications on family topics. Emily Klancher Merchant’s analysis of the family content of PAA’s journal, Demography, covers the period 1964–2010 (Merchant n.d.). The data include articles on family and household structure, marriage, cohabitation, divorce, widowhood, parents, and siblings (topics 17 and 26 in Merchant’s coding scheme). The graph in Fig. 2 shows three-year moving averages. The percentage of articles on family topics increased over that period, especially after the mid-1970s.Footnote 2 The increase in published research in family demography is partly the result of new longitudinal data from the Health and Retirement Study (n.d.), the National Longitudinal Surveys of Youth (NLSY) (Bureau of Labor Statistics 2016a, b), the National Longitudinal Study of Adolescent to Adult Health (Add Health) (Harris and Udry 2016), and the Fragile Families and Child Wellbeing birth cohort study (Brooks-Gunn et al. 2011). Study designs that take account of individuals’ family relationships by direct observations of parents and offspring in the same family have also advanced research—for instance, through the Child Development and Transition to Adulthood Supplements to the PSID (Johnson et al. 2018) and the Children of the NLSY followed into adulthood (Cooksey 2018).

Fig. 2
figure 2

Percentage of articles in Demography on family topics (content 17 and 26), 1964 to 2010. Source: Merchant (n.d.).

Full size image

Lessons Learned

These innovations in theory, data, and methods provide demographers with the tools to better understand change over historical periods and across cohorts in who is considered part of a family, the way family life is organized, and what families do. The lessons learned are ones that demographers are grappling with today. Individuals have linked lives, but people in the same family do not have all of the same links because people are simultaneously members of multiple families. Family members have different but also shared interests. Important family ties exist between noncoresidential family members. And sometimes people who live together in the same household are not part of the same family.

Families and Households, and Individuals’ Linked Lives

Population researchers should intensify their attention to individuals and their family relationships. Researchers should consider the implications for individuals of having linked lives. This does not mean that researchers should collect data only on individuals. “Relationship” requires information about more than one person. But by focusing on individuals, we can better represent that their families and households (or living arrangements) are different things.

I suggest three concerns that new research is beginning to address to improve an understanding of who is in a family and how families work. These concerns build on lessons learned from family demography: (1) family members may live apart; (2) even when people live together (as cohabiting couples do), they may not be a family; and (3) weak family links mean that some family members come and go in individuals’ lives. I illustrate these concerns with examples that show how more careful attention to these issues may shed light on variation in how effectively families can support their members—in other words, how effective family safety nets are.

Family Members May Live Apart

Parents and young children often live apart because of divorce and childbearing outside of cohabitation and marriage. Figure 3 shows the dramatic rise in the percentage of U.S. children living with a single parent. In 1960, 9.1 % of children lived with a single parent, but by 2017, the percentage had tripled to 27.1 %.

Fig. 3
figure 3

Living arrangements of children under 18 years old: 1960 to 2017. Source: U.S. Census Bureau (2017b).

Full size image

The trend shown in Fig. 3 masks considerable diversity. Black children are much more likely than children in any of the other racial/ethnic groups to live apart from at least one of their parents—usually the child’s father. Figure 4 shows that 55.5 % of black children are living with a single parent, compared with 20.4 % of white children. Ties between children and nonresident parents are generally weak, but when they exist, they appear to enhance children’s well-being (Amato and Gilbreth 1999; King and Sobolewski 2006).Footnote 3

Fig. 4
figure 4

Percentage distribution of children's living arrangements by race/ethnicity, 2016. Source: U.S. Census Bureau (2016).

Full size image

Data innovations allow demographers to challenge the assumption behind much of the early research on single motherhood that fathers are truly absent from children’s lives. At least early in their children’s lives, fathers are usually not absent. Evidence from the Fragile Families and Child Wellbeing Study, for example, shows that 30 % of nonresident fathers of children born to unmarried parents are romantically involved with their child’s mother at the time of their child’s birth, and another one half of the fathers are living with the mother and their child; only 10 % have little or no contact (McLanahan and Beck 2010). Despite the innovation of the Fragile Families design in which unmarried fathers are interviewed at the time of their children’s births, fathers’ connections to children in low-income, unmarried families are hard to observe with conventional methods.

The Wisconsin Moms Study (WISCMoms, previously Wisconsin Mothers with Young Children Study) takes a promising new approach to collecting data on the family lives of vulnerable children. WISCMoms is a study of biological fathers’ and stepfathers’ history of marriage and cohabitation with the children’s mother, and the number of nights (if any) that any of the men stayed in the mother’s household in the past 30 days. WISCMoms combines an event history calendar with child rosters to link all of a woman’s births to the children’s fathers and to show movements of fathers, boyfriends, and stepfathers into and out of the household (Berger et al. 2012). This approach links hard-to-observe family members to explore how they share resources. Preliminary findings from the pilot study indicate that significant percentages of fathers who live with the mother and child do not have keys to their home, suggesting inequality between the parents. In addition, many mothers reported that adults other than the children’s fathers or the mother’s boyfriends stayed overnight in the past month. Thus, children in vulnerable families are exposed to a range of adults whose presence in their lives may not be noted in traditional surveys. The WISCMoms study provides data that researchers want using concepts that make sense to respondents.

Once the children are grown, they are unlikely to live with their parents. Parents who have adult children at home receive a lot of attention in the popular press, but only 1 in 5 parents of an adult child age 25 or older lives with a child (author’s calculations using the 2013 PSID Rosters and Transfers Module). These offspring may not be in their parent’s household, but they remain part of their parent’s family. That most parents and adult children live apart raises the question of what behaviors or bases of solidarity connect them to each other (Keyfitz 1987).

Obligations are one dimension of solidarity. Most people think of parents and adult children as part of the same family even though they do not live together. A 2010 Pew survey asked, “Suppose someone you know had a serious problem and needed either financial help or caregiving. How obligated would you feel to provide assistance if that person were your [NAME ON LIST]:” parent, grown child, and several other relationships.Footnote 4 Most respondents said that they felt very obligated to help a parent or a grown child. Eighty-five percent of adults felt very obligated to help their parents, and 78 % felt very obligated to help their grown child (Parker 2011). Determining whether these felt obligations actually provide a safety net for family members requires information about both obligations and behavior—that is, the help parents and adult children actually give each other. Some evidence suggests that individuals’ attitudes about family obligations predict their behavior (Ganong and Coleman 2005; Silverstein et al. 2006), but the relationship between obligations and family support is not well understood (Seltzer et al. 2012).

The Add Health Parent Study data (2015–2017) released in the summer of 2018 will let researchers study the connections between transfers of time and money on the one hand and parents’ feelings about who they can count on—and who counts on them—on the other hand (http://www.cpc.unc.edu/projects/addhealth/design/parents-phase-2). Because the design includes parents, spouse/partners, and siblings, some of whom are half- or stepsiblings, the Add Health Parent Study will provide new opportunities to study individuals’ linked lives across relationship domains and across households. The study does this in the context of the multiwave panel of observations of the National Longitudinal Study of Adolescent to Adult Health. The rich data on individuals’ health, education, work, and family relationships will support studies of the effects of family relationships on a range of outcomes from adolescence to middle age.

Cohabiting Couples: Are They Part of the Family?

Americans think of cohabiting couples who live with their children as a family (Powell et al. 2010). Among opposite-sex cohabiting couples, almost 40 % live with at least one child under age 18 (U.S. Census Bureau 2017c). Many of these children are from parents’ previous relationships. Between one third and two fifths of cohabiting parents with minor children in their households have no shared biological children (U.S. Census Bureau 2017a). Their households look more like stepfamilies than like first-married biological families.

Parents and children may disagree about whether a cohabiting partner is part of their family because they have different points of view about the cohabiting relationship. Brown and Manning (2009) found considerable disagreement in mothers’ and teenagers’ reports about whether their mother was living with a partner. Sweeney (2012) showed that this is because teenagers sometimes refer to their mother’s partner as an “other nonrelative” or as a stepfather. The terminology that people use to describe cohabiters may signal whether they see the person as part of their family.

Parents and adult children also may disagree about whether the cohabiting partner of the adult child is part of the parent’s family. A vignette study shed light on this by showing that Americans think it is more appropriate to support adult children who are married than those who are cohabiting (Seltzer et al. 2010, 2012). In the study, a national sample of adults responded to a vignette about whether a mother and adult child should share a home if one of them became homeless. The vignette varied the offspring’s union status. Thirty-one percent of respondents said it would be a good idea for the mother and married offspring to move in together, compared with only 17 % if the offspring were instead cohabiting (Seltzer et al. 2010). This suggests that cohabiting partners may not be viewed as “in the family” to the same extent as spouses are (Seltzer et al. 2012). Parents may not think of their offspring’s cohabiting partner as part of the family because cohabiting relationships are usually short-lived, as noted earlier.

Family Members Come and Go

Stepfamily ties may also be short-lived because they often break down if the marriage that brings the stepparent (or child) into the family ends (Coleman et al. 2005; Noël-Miller 2013; Schmeeckle et al. 2006). Figure 5 shows that U.S. adults feel much greater obligation to their biological parents and children than to their stepparents and stepchildren. But obligations to any parent or child are considerably greater than to nonfamily members, including best friends.

Fig. 5
figure 5

Percentage who feel “very obligated” to provide financial assistance or caregiving to their biological parent, stepparent, grown biological child, grown stepchild, or friend. Source: Parker (2011).

Full size image

Behavioral data are consistent with these attitudes. Stepparents and stepchildren are less likely to live together or near each other later in life (Seltzer et al. 2013). Stepchildren are also less likely to provide care to older parents (Pezzin et al. 2008; Pezzin and Schone 1999). Stepparents and stepchildren may be less connected to the family because their roles in the family are poorly defined (Cherlin 1978) and because they face competing demands from biological family members for whom they have responsibilities (Ganong and Coleman 2017). That stepfamilies formed after widowhood have weaker ties than in families with no stepkin suggests that factors beyond competing demands or the legacy of divorce threaten family solidarity in stepfamilies (Seltzer et al. 2013).

Loose ties between stepparents and children are important for two reasons. First, both simulation findings and observational data point to increases in stepfamilies in the future (Wachter 1997; Wiemers et al. 2018; Yahirun et al. 2018). Second, stepfamily ties are more common among disadvantaged individuals (Parker 2011), as illustrated by Fig. 6 using data from the 2014 Health and Retirement Study (HRS). Among persons in their late 50s, those with no more than a high school education (18 %) are more than twice as likely to have stepparents as those with a college degree (8 %). Evidence on young families also shows this educational difference in the prevalence of stepfamilies (Thomson et al. 2014). Unfortunately, those with the least education are more likely to need family assistance because of their higher rates of unemployment and health problems (Hout 2012). These problems also contribute to looser family ties (Cherlin 2014).

Fig. 6
figure 6

Percentage of persons ages 55–60 with any stepparent, by education, 2014. Source: Health and Retirement Study.

Full size image

Changing Families Have Changed Family Demography

Demographers have made considerable progress in studying how families have changed. But this progress points to new questions and new problems to solve. I identify three areas in which family demographers have made progress but where more research is necessary: studies of the long-term effects of early family experiences, simulations that suggest what the future will look like, and studies of resource sharing among family members who live apart.

To examine the long reach of individuals’ early family experiences on well-being later in life places large demands on data, requiring patience in allowing panel designs to mature. The design of the PSID, which follows all members of the original PSID households and their offspring, as well as the survey’s longevity allow us to learn how family ties across multiple generations affect individuals’ life chances (Song 2016; Song and Mare 2017).

Links between administrative and survey data also extend family histories beyond what is possible for most surveys. For example, Warren and colleagues are linking the 1940 census to individuals in five major U.S. aging and health surveys to study the effects of family characteristics on later-life outcomes (Warren 2016). These are characteristics of only coresident family members, but the design is a big step toward learning more about the long reach of early family life. Change in the U.S. population due to immigration means that the 1940 census-survey match will not be representative of the U.S. population today. Understanding cultural and economic constraints on family behavior requires attention to variation in the experiences of all U.S. families, not just those who began life in the United States.

Similarly, increases in stepfamily relationships mean that the PSID rules that follow biological children but not stepchildren limit the extent to which the PSID data can address these increasingly common family ties. The 2013 Rosters and Transfers module included in the PSID provides information about these stepfamily relationships for a point in time (Schoeni et al. 2015). The enriched data provide new insights into the prevalence and consequences of stepfamily ties for adults of all ages, not just those with young children or those who are middle-aged and older (Wiemers et al. 2018).

Some research questions can be addressed by exploiting the demographers’ toolkit to simulate the existence of different types of family relationships, including stepfamily ties connected by formal marriage and those connected only by cohabitation. Furstenberg et al. (2015) described efforts to use this method to inform an understanding of the interplay of family change, economic inequality, and demand for health care. The simulations can include data from a variety of sources on births, deaths, unions, and the dissolution of these relationships, as well as individuals’ characteristics and policies affecting employment and access to health care.

The demographic skeleton that simulations produce is not enough. Researchers should go beyond the bare bones to incorporate assumptions about family obligations and behavior. This cultural and behavioral overlay would provide a sense of the effectiveness of family networks for helping or hindering individuals. For instance, the evidence that stepparents and stepchildren are less likely to help each other is one type of information that could be built into simulations to estimate who would have a potential caregiver or who would be able to provide care. Researchers also should evaluate the quality of the simulations by comparing their results to observed family interactions, where possible.

A third way that demographers are making progress is by showing that the existence and characteristics of family members who live apart matter for conclusions about individuals’ health and well-being. Findings of LaFave and Thomas (2017) demonstrate the value of studying all family ties, regardless of whether the family members live together. Using data from the Indonesian Family Life Survey, they showed that the resources of both coresident and noncoresident family members affect a child’s health and human capital.

Social and economic contexts matter too. Findings for U.S. families may differ from those for families in Indonesia because of the countries’ significantly different systems of public support for families. But accumulating evidence for the United States also suggests that individuals’ lives are affected by the resources of family members who do not live with them. Choi et al. (2016) found that the incomes of family members outside a person’s household are positively associated with that person’s consumption. Dalton and LaFave (2017) showed that nonresident adult offspring reduce their consumption and spend down savings in response to parents’ health crises. A fruitful next step would be to incorporate more data on the types of interactions and behaviors that connect family members across households, which would illuminate the processes by which nonresident family members’ resources affect individual welfare.

These efforts are very promising for showing how family membership affects individuals regardless of whether they live together, but demographers still lack data with which to address important questions about individuals’ linked lives (Seltzer 2015). For instance, existing data cannot address how mothers adjust their employment when they first become grandmothers. Grandparents, especially grandmothers, provide significant childcare (Luo et al. 2012). If providing childcare has a similar effect on employment as the effect of providing care to aged parents, new grandparents’ help with childcare may interrupt the grandparent’s labor force participation by reducing hours worked or prompting an exit from the labor force (Fahle and McGarry 2018).

Researchers cannot now observe directly U.S. grandparents’ employment adjustments or the longer-term economic consequences for the grandparent generation of childcare provided at the time of the first transition to grandparenthood. Data requirements include prospective information on two adult generations, either by direct observation or through proxy reports or linked records. Neither the PSID nor the HRS, the two most commonly used national data sources on linked generations, meet these requirements. The PSID has employment information and fertility histories for parents and offspring in the genealogical design, but its fertility histories are not sufficiently complete to support an analysis of the first transition to grandparenthood. The HRS regularly updates information on whether the respondent’s offspring have children (the respondent’s grandchildren), but observes the transition to grandparenthood only for those who make the transition late in life. HRS data show that among those born in the mid-1950s, 38.2 % of women and 27.5 % of men ages 51–54 already had a biological grandchild before they became age-eligible for the HRS sample (Yahirun et al. 2018). Earlier ages of childbearing among those who are less well educated imply that the transition to grandparenthood is earlier for these parents as well; for them, any interruption of paid work would occur during prime working ages and thus may have more serious long-term economic consequences for the grandparent generation than for those who are well educated. Longer lives and educational variation in the timing of family transitions raise new questions requiring new data about how families foster the next generation and at what costs.

Family, Friends, and the Family Uncertainty Principle

A richer and more encompassing view of families is indispensable to population research, not just in the United States, but in all settings. The concept of family has standing in legal and social institutions, as well as in individuals’ minds. Yet it is very difficult to draw clear, scientifically grounded boundaries to distinguish between who is in a family and who is not. Some scholars have argued that researchers should focus on personal communities or the relationships important to an individual at a given time rather than on families (Pahl and Spencer 2010). But this strategy ignores that obligations to family members are more universally recognized than obligations to friends, as I have shown. It also ignores that family obligations are long-term commitments and gives short shrift to the community and cultural processes that help define and enforce family roles. A focus on personal communities, though, points to an avenue that may shed light on how individuals resolve ambiguity about stepfamily relationships. Researchers could compare friend ties and stepfamily ties to investigate how individuals’ interactions and others’ perceptions of these relationships clarify the obligations between friends or between stepkin.

Periods of transition—for instance, from lower to higher exposures to stepfamily relationships or to nonmarital cohabitation—offer a chance to learn how obligations are established and enforced in family or family-like relationships. Transition periods are marked by what we might call the “family uncertainty principle.” This is different from uncertainty about biological relationships, such as when the biological father of a child is in question. Instead, family uncertainty is the uncertainty that social scientists have about the family relationships and units that we study and the social uncertainties faced by the people whom we study. As demographers reach beyond old-style family and household demography and acknowledge the diversity, complexity, and geographic spread of contemporary families, scholars face two big dilemmas. First, the more family ties we recognize, the harder it is to see them with existing observational tools. Second, even if these ties can be observed, we face greater uncertainty about whether these observations match the family members’ own perceptions of their family ties.

Who is in a family is a dilemma for the people we study as well as for researchers. Family relationships are special in part because they imply long-term obligations. When it is ambiguous as to whether a relationship is a family relationship, such as in cohabiting or stepfamilies, individuals (and researchers) are uncertain about whether obligations will be fulfilled. This uncertainty makes relationships even more unstable.

Although the family uncertainty principle challenges meaningful advances in family demography, demographers have met hard challenges in the past and need to do so again. A better understanding of what family means, how families are organized, where they live, and how they behave is essential for understanding basic demographic processes and extending the reach of population research to address compelling problems in health and socioeconomic inequality.

I have identified recent promising developments that help demographers better understand what families are and the causes and consequences of complex, noncoresident, unstable, and diverse forms of family life. This research represents significant progress in the nearly 60 years since the first official recognition of family demography. It is a challenge to family demographers to continue to improve understanding of the behaviors and feelings that make family relationships unique compared with relationships with friends, coworkers, and neighbors.