Intergenerational family support processes from young adulthood through later life: Do we need a new national survey?

Abstract

I argue that the United States needs new survey data on intergenerational relationships in light of the dramatic demographic changes in parent-child and couple relationships that were not anticipated when many major family datasets were designed. Increases in nonmarital childbearing, the instability of parents’ relationships and high rates of repartnering challenge conventional approaches to data collection on families. Large race-ethnic and socioeconomic differences in the extent of these changes and their impact on intergenerational support may contribute to growing inequality. A new study must collect data on both household relationships and relationships among family members who live apart because most U.S. parents and adult offspring do not co-reside. The survey should obtain information on the timing of family transitions and include multiple cohorts to take account of differences in societal conditions that influence family experiences. A longitudinal design would show how parent-child relationships unfold over time and build on past histories. The paper identifies the dimensions of intergenerational ties that should be measured and explains why existing data cannot address the need for a new study.

1. Introduction

Families bring children into the world and raise them to be the next generation of productive adults. Families transmit advantage or disadvantage to younger generations, influencing the development of children's cognitive and social skills, helping them through school, and launching them into adulthood. Emotional support, caregiving time, and material support come from family members throughout life. Even when family members are not actively transferring resources to each other, they still may be available to help in an emergency. How effective families are in helping their members do well in the world depends on characteristics of the family – how close family members are, how much they trust each other, and how well-off they are. Political, economic, and social institutions facilitate the efforts of some families, but hinder the efforts of others.

Conventional representations of U.S. families focus on two dyads – the parent-child dyad and the couple dyad. These representations combine the parent-child dyad and couple dyad in the nuclear family in which a married couple has offspring who are biological children and they all live together in the same household until the children reach adulthood and move to independent households, marry, and have children of their own. Demographic changes in both couple relationships and parent-child relationships challenge this conventional view of U.S. families and motivate the need for a new panel survey of family and household relationships. This paper develops the case for new data, considers design elements that a new study should incorporate, identifies the dimensions of intergenerational ties that should be measured, and explains why existing data cannot address the need for a new study. I focus on parent-child relationships in adulthood, but these relationships must be understood in the context in which couples bear and rear children. I build on the Seltzer and Bianchi [35] review of intergenerational relationships, an assessment of existing data on intergenerational ties (Bianchi et al. [2]), and a report of the NICHD-funded project on Explaining Family Change and Variation (Morgan et al. [29]).

2. Changes in U.S. family demography and inequality

Changes in U.S. families motivate new data collection on parent-child relationships in adulthood. Many of these changes were not anticipated in the design of existing studies or have not been addressed in U.S. surveys since the late 1980s. This section describes changes in the demography of U.S. families, highlighting recent changes that should be addressed in new data collection.

U.S. families have experienced dramatic change in their structure and composition over the past century, and especially in the past 40 years. The long-term decline in fertility has continued into the 21^st century, with the exception of the Baby Boom of the middle of the 20^th century. Three and four generation families are increasingly common, and grandchildren are more likely to reach adulthood with living grandparents as a result of greater life expectancy (Uhlenberg [43], Table 4.1). High rates of cohabitation, instability of cohabiting unions and marriage, and the growing percentage of children born outside of marriage mark the separation of marriage and childrearing, particularly since the 1970s. Americans continue to marry or cohabit after a first (or second) relationship dissolves. Parents have become increasingly likely to have their children in different unions, a phenomenon now referred to as multi-partner fertility.

The increase in childbearing outside of marriage and instability of cohabiting unions contribute to a difference in women's and men's family relationships. Mothers and children remain tied to each other by sharing a home with each other and sometimes the mother's new partner. Fathers and children face the challenge of maintaining a relationship without living together, and these difficulties often result in distant or lost ties. The weakening of father-child relationships diminishes children's access to the father's social and economic resources during childhood and the transition to adulthood and also threatens the father's ability to rely on adult offspring for help if he becomes ill or infirm later in life (Cooney and Uhlenberg [11]; Lin [23]).

Significant socioeconomic and race-ethnic differences in the demographic structure of families underlie these broad trends. Although these differences have long characterized U.S. families, since the late 1970s the socioeconomic divide in family experiences has been growing, resulting in a wide gap between the structure of families formed by those who have college educations and the families formed by those who are less well educated (Cherlin [8]). The college-educated marry, bear children in marriage, and raise these children in stable marriages. Those with less schooling are more likely to have children outside of marriage, to cohabit instead of marrying, to stay with their cohabiting partners for short periods, and then form a new relationship, often having children with multiple partners.

Unstable family lives, competing loyalties, and ambiguity about family responsibilities may make it more difficult for parents and offspring to help each other, to launch young adults, and to provide caregiving when the public safety net does not. For instance, stepchildren leave home earlier than biological children as they transition to adulthood (White [45]), and their parents contribute less to their college expenses than children from intact families (Lopez Turley and Desmond [24]), which may explain the lower rates of college completion among those launched from families with a history of instability (Fomby and Bosick [16]). At the other end of life, stepmothers and adult children are less likely to live together or near each other (Seltzer, Yahirun, and Bianchi [38]). Stepparents do not expect to provide as much support to stepchildren as to biological children, and children from step and cohabiting families are less likely to be caregivers or involved in older stepparents’ lives (Aquilino [1]; Eggebeen [13]; Pezzin, Pollak and Schone [31]).

Race-ethnic differences in the demographic context of parenthood persist (Raley and Sweeney [33]; Sweeney and Raley [42]). Compared to non-Hispanic Whites, Hispanics and non-Hispanic Black women bear a much higher percentage of children outside of marriage. Nearly half of Black women's nonmarital births are to lone mothers (i.e., those who are not living with a partner), whereas much smaller percentages of nonmarital births are to lone mothers among Whites and Hispanics. Blacks and Hispanics become parents at younger ages than Whites (Sweeney and Raley [42]). Compared to Whites, Blacks and Hispanics also are more likely to have stepkin due to their higher rates of union instability and nonunion fertility (Parker [30]).

Race-ethnic differences in transitions to parenthood translate into differences in grandparenthood. Approximately 80% of Blacks and Hispanics have become grandparents by ages 55–64, but only 70% of non-Hispanic Whites are grandparents by then. This race difference disappears at later ages (Seltzer and Yahirun [37]). Younger grandparents may be healthier and better able to be actively involved in grandchildren's lives, for example as child care providers, but grandparents also are more likely to be employed at these ages and employment may compete with grandparents’ availability to provide care. Growing race-ethnic diversity in the U.S. population points to the importance of understanding the explanations for these demographic differences in parent-child relationships and their consequences.

Trends and differences in family demography cannot be understood without also paying attention to the rise in economic inequality among families since the 1970s. In part this is due to deterioration in the earnings opportunities of those with high school degrees compared to the growth in earnings opportunities of those with college educations. And in part this is due to the fact that families in which a parent is college-educated are likely to have two earners, whereas families in which the parent has less schooling are increasingly likely to have only one earner owing to high rates of marital instability (Fischer and Hout [14]). Wealth also has become more concentrated (Wolff [47]). Because parents’ wealth, especially home ownership, is one way that parents finance their offspring's college educations (Lovenheim [25]), wealth in one generation affects the welfare of the next generation. Over the past three decades these transfers from parents to offspring have become more important for supporting young adults through college (Wightman et al. [46]).

A new survey would be an invaluable source of information about the mechanisms through which parents and children help each other manage life in the 21^st century. No single data collection effort can provide the information needed to fully explain the relationship between rising economic inequality among families and the divide between the family lives of those who are educationally advantaged and disadvantaged, but gaps in the design and content of existing data are barriers to understanding these important phenomena. Relationships among adult family members are informal and difficult to monitor or enforce as a matter of policy or through official records. Yet the time, emotional support, and financial transfers that parents and adult offspring give each other are central to the welfare of individual family members and to population-level inequality. Policies at the state and federal level make assumptions about who is in a family to allocate rights and responsibilities. Examples of relevant policies include: eligibility for health insurance coverage, whose incomes to consider for allocating financial aid to college students, whose incomes can be claimed for debt recovery, and who has the right to make health care decisions for someone who is ill or incapacitated. Evaluating how well assumptions about who is a family member fit different subgroups in the population requires a more complete description of family ties than can be addressed with existing data.

3. Design elements for a new survey

The design of a new survey must take into account three elements of intergenerational relationships in the United States: (1) the distinction between who lives together and who is in the family; (2) variation across the U.S. population in the timing of family transitions; and (3) cohort differences in family experiences. The design must be longitudinal because parent-child relationships unfold over time and build on past histories. A promising strategy would be to sample individuals in households and follow the individuals over time, adopting principles like those in the Panel Study of Income Dynamics (PSID) with a genealogical design in which offspring of the original respondents age into the sample and become respondents themselves, perhaps at age 15, to contribute information about youths’ attitudes about their family relationships and parents’ support during the transition to adulthood.

The sample should be of the national population because existing data cannot describe key dimensions of the entire population that are needed for estimates of policy-related characteristics of the population. The sample must be refreshed periodically and weighted so that respondents and their offspring are representative of the current U.S. population. Researchers would need to make difficult decisions about whether or not to oversample theoretically and policy-relevant subgroups, such as grandparents raising grandchildren and severely economically disadvantaged families. Determining the criteria for oversampling, the periodicity for sample refreshment, and developing appropriate weights for population-level estimates would be challenging, but necessary tasks in the design of a new national survey. Important theoretical questions about parent-adult child relationships could be addressed with data collected from populations in more restricted geographical areas. I do not address these concerns although such designs hold promise.

4. Who lives together vs. who is in the family

The distinction between household composition and family membership matters both early and later in life. Children who grow up in the same household may have different fathers; and step and half siblings may never live together in the same household. As result of these demographic changes, family relationships increasingly span household boundaries in childhood. Just over 31% of children under age 18 lived apart from at least one of their biological parents in 2013 (U.S. Census Bureau [44]).

Ties between adult offspring and parents – and grandparents – also span household boundaries. Among mothers over age 50, 22% live in the same household as at least one of their adult children. From the adult child's perspective, co-residence is less likely; 11% of adult children live with their mother (Seltzer, Yahirun, and Bianchi [38], Table 1). Grandparent-grandchildco-residence is even less likely. From the oldest generation's point of view, 7% of women age 55+ lived with a grandchild, according to data from the 2007–09 American Community Survey (ACS) (Seltzer and Yahirun [37]).

That such a small fraction of adults live with parents means that a new survey of intergenerational ties must measure cross-household relationships. Existing data provide a model for measuring transfers of time and money, and even the quality of relationships between parents and children who live apart (Seltzer and Bianchi [35]; Bianchi et al. [2]). However existing data also point to challenges that should be addressed in the design and content of a new survey. The first of these is to identify who is in the family. This is essential for determining who has the potential to share a household or to give or receive support from a parent or child.

The ACS estimate of women's co-residence with grandchildren illustrates the need for a roster of family members in new data on intergenerational support. The ACS, like most U.S. surveys, collects information about relationships among those who live in the same household, but the ACS does not obtain information about family members who live in different households. Thus, it is not possible to use these data to estimate the percentage of grandparents who live with a grandchild because the ACS does not identify all grandparents. Studies that use designs like that of the ACS cannot distinguish between those who could be living in a multigenerational household but who do not out of choice or constraint and those who live in a single generation household because the parent (or grandparent) is no longer alive. Basic description of differences by education and race-ethnicity in intergenerational co-residence requires information about who is “at risk” of living in an intergenerational household. Information about who gives or receives financial transfers, as in the Survey of Income and Program Participation (SIPP), also would be more useful for understanding intergenerational relationships if researchers could identify those who have an alter to whom a transfer could be given (or from whom a transfer could be received).

Newly released data from the 2013 Roster and Transfer module of the Panel Study of Income Dynamics (PSID) help to address this problem (Bianchi et al. [5]). The new module obtains information about living parents and stepparents, in-laws for those who are married or in long-term cohabiting relationships, and adult children and stepchildren. Little is known about how accurately a single respondent can report about a spouse or cohabiting partner's kin network, although Soldo and Hill [41] found more missing data on reports about in-laws than about the respondent's own parents. Even less is known about how qualitative changes over the course of a person's relationship with a stepparent or stepchild affect reports about that person's existence. The PSID relies on one respondent per household to report about the existence and characteristics of parents and offspring and to report about transfers to/from the respondent and his/her spouse or long-term cohabiting partner. The Health and Retirement Study (HRS) also relies on one respondent per household to report about the existence and characteristics of offspring and financial transfers. In earlier years, the HRS also relied on one respondent to report about their parents and parents-in-law, but in recent years the survey has asked each spouse/partner to report about his or her own parents.

In addition to the challenge of obtaining high quality reports about parents and adult offspring from a single respondent, a new survey must address the challenge of when to ask only one person about transfers and relationship quality and when two or more reports are needed (for instance, reports from a parent-child dyad or reports from a parent and all of his/her offspring). Relationship quality and trust, like other attitudes and preferences, may require each person's report. This is especially difficult because dyads with poor relationships are much less likely to participate as matched pairs in a survey than dyads with good relationships (Seltzer and Bianchi [35]). The genealogical design of the PSID with the 2013 Roster and Transfer module in which parents and adult offspring who have become household heads both report about the same intergenerational transfers of time and money will shed light on this. But the PSID data cannot address the need for dyadic reports about qualitative aspects of relationship quality, such as emotional closeness and conflict, because the module did not include those measures (Bianchi et al. [4]).

Finally, there are likely to be differences in the quality of reports about intergenerational transfers of time and money between parents and offspring who live in the same household and those who live apart. Aged parents give less money to co-resident adult children than to adult children who do not co-reside (McGarry and Schoeni [28]), but researchers do not know if co-residence affects the concept that parents use to identify what is a transfer and what is simply a matter of sharing a home. To meet the challenge of asking good survey questions about transfers between co-resident kin requires developmental work to learn more about how individuals think about and report on these transfers.

5. Age and the timing of family transitions

New data collection on intergenerational support processes should use a design that observes both early and late family transitions because age and the timing of transitions affect individuals’ well-being. Chronological age defines the right to vote, to marry, and to make independent decisions. Rules about age also affect school entrance, employment, and eligibility for certain types of health insurance and Social Security. These formal rules have a counterpart in informal understandings about the appropriate ages for some family transitions such as leaving the parents’ home, becoming a parent, and marriage (Smith [40]). Despite age rules, individuals vary in the ages at which they become parents and grandparents, and these age differences shape their experiences in these roles. For example, adults who are older when they have children have more secure employment and greater economic resources than those who are younger when they become parents and their greater economic resources provide advantages to their offspring (Mare and Tzeng [27]).

The timing of grandparenthood also affects the economic resources grandparents have to invest in adult children and grandchildren. One of the most important transfers from parents to young adult offspring who have become parents is the time that grandparents spend taking care of grandchildren (Compton and Pollak [10]). Availability to provide child care may depend on grandparents’ age. Men who are past retirement age are more likely to babysit grandchildren then men who are still in their prime working ages (Kahn et al. [22]). In addition, accumulating evidence suggests that grandparents’ characteristics, including age, may have an independent effect on children's achievement (Fomby et al. [17]; Chan and Boliver [7]).

Demands for intergenerational support may be greater for a middle generation who has an aging parent as well as children who have not been launched, the so-called “sandwiched” middle generation. This may be particularly costly for the middle generation if they are at a critical stage in their employment trajectories and in saving for retirement. There is some evidence from the PSID that for women age 50–65 the percentage who might experience the demands for assistance to an older and younger generation at the same time increased between 1988 and 2007 from 35% to 45% (Seltzer and Bianchi [35]).

In light of the importance of ages at which individuals experience family events, a new study of intergenerational relationships should be designed to observe the transition to grandparenthood. The study should collect information about financial transfers that grandparents make to adult offspring leading up to their producing a grandchild (Cox and Stark [12]) and the time help associated with the birth of a grandchild. Because those without a college education and minorities become grandparents at younger ages than those with college educations and Whites, the Health and Retirement Study (HRS), which represents those over age 50, cannot describe fully this aspect of intergenerational support because the data first observe respondents when most are already grandparents. The Panel Study of Income Dynamics (PSID) data also lack information about how parents make the transition to grand-parenthood and their involvement in grandchildren's lives. Although the new Child Development Supplement (CDS) to the PSID has the potential to address some aspects of grandparent involvement from the grandchild's point of view, the planned six-year interval between observations means that these data also cannot describe educational and race differences in the transition to grandparenthood.

6. Cohort differences in family experiences

Macro-level variation, such as shifts in the cultural acceptance of cohabitation and nonunion childbearing, economic growth and recession, and technological innovation, contributes to cohort differences in who is recognized as a parent or child, the needs and resource of each generation, and how family members communicate with each other. The design of a new survey on intergenerational processes should include multiple cohorts to allow researchers to examine the association between social context and intergenerational ties. The PSID includes multiple cohorts as does the HRS, which incorporates new cohorts every few years.

Other U.S. national surveys are single-cohort studies, such as the 1995 National Longitudinal Study of Adolescent to Adult Health (Add Health), and the National Longitudinal Surveys (NLS), particularly the 1979 National Longitudinal Survey of Youth (NLSY79) and the 1997 National Longitudinal Survey of Youth (NLSY97). These provide valuable information about parent-child relationships. The Add Health survey is unusual in its coverage of sibling relationships as well as parent-child relationships, an important addition because parent-child relationships may vary within the same family (Pillemer and Suitor [32]). Some national cohort surveys include direct observations from two generations (e.g., the NLSY79, the first wave of the NLSY97, and the Add Health,¹) another strength of these designs. Despite these strengths, the surveys cannot be used to describe the intergenerational relationships of the U.S. population as a whole. The cohort studies also vary in the content and measurement of central concepts, further limiting the potential for comparisons across cohorts.

7. Who is in the family vs. who acts like family

Although the goal of this paper is to evaluate the need for new data on intergenerational support processes, some people do not have the potential for intergenerational ties. Increasing rates of childlessness (Hayford [19]) point to the importance of measuring alternate sources of support and caregiving. Collecting information about who provides assistance, whether or not the assistance is from a child or parent, will shed light on the relative value of family vs. non-family relationships for individuals’ well-being.

8. Dimensions of intergenerational support

Intergenerational support is routinely measured in large surveys by co-residence or shared housing, transfers of money, and time help (e.g., caregiving in old age, help with household tasks, and child care for grandchildren). A new study should consider whether or not questions about domains of assistance in existing studies should be revised to take account of technological innovations and the new demands of everyday life, such as assistance with managing health insurance choices and health care. Large surveys typically pay less attention to the quality of relationships, such as emotional closeness, conflict, and trust, but these are important for understanding the motivations for family support. Family support may be routine and regular, an infrequent response to family crises, or tightly tied to key life course transitions, such as going to college, establishing an independent household, or becoming a parent.

Testing some theories about intergenerational support requires data on what all of the children in a family get from their parents and what they each give to their parents (Bianchi et al. [3]). This is an ambitious requirement, but one that is met by both the PSID and HRS. Yet these surveys do not include questions about side-payments or help that one sibling may provide another, for instance to compensate for assistance that the other sibling is giving their parents. How adult siblings help each other and how, if at all, they coordinate care of aging parents are important questions that require new data.

Ideally a new survey would allow researchers to estimate unmet needs for family support and the extent to which family support complements or substitutes for support from those outside the family and from the state. Few national surveys have measured directly individuals’ attitudes about whether and how obligated they are to help family members and whether or not they can count on family members to help them (Seltzer, Lau, and Bianchi [36]). Yet naïve notions of the “family safety net” are elements of policy debates about the responsibility to help family members who are in financial trouble or need care.

9. What surveys are leading contenders as alternatives to a new study?

Four existing surveys hold considerable merit for studying intergenerational family support in adulthood but they have significant weaknesses that would be better addressed by a new study rather than augmenting these designs. Because I view the need for national data as critical, I consider only national surveys in this section: the Health and Retirement Study (HRS), the General Social Survey (GSS), the Panel Study of Income Dynamics (PSID), and the National Survey of Families and Households (NSFH).

Great strengths of the HRS (http://hrsonline.isr.umich.edu/) are the longitudinal design with repeated measures at short intervals, the multi-cohort sample, and the detailed roster of all of the respondent's and spouse or cohabiting partner's off-spring. The roster provides the denominator for measuring the incidence of transfers up and down the generational ladder from either the parents’ or children's point of view. The creation of the Rand Family Data file to help researchers make better use of the child-level data is a significant resource (http://www.rand.org/labor/aging/dataprod/family.html). The repeated measures of transfers of time and money and of parents’ and offspring's economic and demographic characteristics facilitate the study of cumulative investments, within family variation, and the effects on intergenerational support of changes in either parents’ or children's circumstances. HRS also has experimental modules that introduce new measures of social and psychological characteristics of respondents and sometimes families.

A weakness of the HRS design for studies of intergenerational support processes is the restriction to a sample of the population over age 50. Although some aspects of intergenerational ties can be reported about retrospectively, such important phenomena as how parents respond to first becoming grandparents, cannot be studied with these data. Early life measures require retrospective reports. Although some information can be effectively obtained retrospectively, other information, such as attitudes, cannot. The HRS also includes only one family respondent who reports about offspring. Thus offsprings’ characteristics are only described by proxy reports. This is a particular disadvantage for models of intergenerational relationships that consider the quality of relationships and whether or not adult offspring think that they can count on their parents for support or think that their parents are counting on them. These expectations may affect employment and residential mobility decisions in both generations. In married or cohabiting couples, one respondent reports about how each child is related (biological, adopted, step) to the spouse or partner, a reporting task that has become increasingly difficult over time as spouses or cohabiting partners have become more likely to have children from prior relationships.

The HRS design could be modified to include interviews with offspring (Bianchi et al. [3]). Recruiting other family members through their ties to a respondent is a difficult task, and is likely to be more successful for dyads with good relationships than for those with poor relationships. An alternative is to include younger respondents in the HRS sample and attempt to match offspring and parents. Because of the geographic clustering of the sample design, this is likely to result in a sample of matched parents and offspring that over-represents families whose members choose to live in close proximity because they want to be near each other or share other preferences or because parents and offspring are similar in their labor market opportunities.

The GSS is the only long-running national survey that collects data on family-related attitudes for the U.S. population (http://www3.norc.org/GSS+Website/). The repeated cross-sectional design² is an invaluable resource for charting social change and for comparing the United States to a wide range of other countries in the International Social Science Programme of parallel surveys. The GSS, however, includes limited information about family experiences and parent-child relationships except in special topics modules. Not since 2002 has the GSS included questions about the availability of different types of kin and respondents’ perceptions of whether or not they could rely on family members for support if they needed it. The GSS does not include characteristics of the family members, such as education, employment status, health, that would affect their ability to provide help if it were needed.

The PSID is a sample of individuals living in U.S. households in 1968 (http://psidonline.isr.umich.edu/). These individuals, their children, and grandchildren have been followed since then with observations every year until 1997 and then every other year. As young adults who have the PSID “gene” leave their parents’ households they become respondents as well. This unfolding genealogical design is a great strength because it means that parents and adult offspring are observed whether or not they co-reside. The long history of detailed economic and demographic information combined is a valuable resource for the study of intergenerational and multigenerational processes (Mare [26]). The value of the data is enhanced by the multi-cohort design and the inclusion of respondents of all ages so that parent-child ties can be examined in early, middle, and late adulthood.

Another strength of the PSID is the Transition to Adulthood (TA) component that interviews offspring who were first observed as minor children in the Child Development Supplement (CDS) sample that obtained detailed information about children's home and family lives. Inclusion in the TA study is based on chronological age and is independent of whether or not the young adult lives with parents.

A significant weakness of the PSID is that only in 1988 and in 2013 did the PSID include a module that asked explicitly about transfers between parents and adult children. The 2013 module also provides detailed information about the characteristics of parents, stepparents, in-laws, adult children, and stepchildren so that investigators will know who gives or receives transfers among those who could potentially be involved in such exchanges. The new module will allow national estimates of family support, but there are no plans to repeat this.

Because the PSID is a sample of individuals in 1968 families the sample is not representative of the current U.S. population. In particular, immigrants, Hispanics, and Asians are not represented or are represented in too few numbers to allow independent estimates of their family support processes despite efforts to include an immigrant refresher sample in the late 1990s. Because many studies suggest that intergenerational support processes differ by race-ethnicity and by nativity, this is a disadvantage of the PSID. The race-ethnic diversity of today's U.S. population warrants a sample that is more representative.

The PSID has not asked any questions about the family safety net, that is whether or not parents and adult children count on each other for help – the latent or insurance function of kin – since 1980. There is only very limited information about relationship quality, an important construct in theories about intergenerational ties. Without more frequent measures of how parents and offspring interact with each other and a broadening of the content to include measures of relationship quality and attitudes about family support, the PSID could not substitute for a new national survey.

The PSID rule that adult offspring do not become respondents until they head their own households also would have to be modified to improve the value of the data for studying intergenerational ties. Otherwise parent-child reports about their relationship can only be combined if the adult child has lived independently at some point. Because co-residence is a type of intergenerational support, the PSID would be improved by following the rule used in the TA component of the study in which offspring become respondents based on age rather than living situation.

The NSFH is a study of a sample of adults of all ages first interviewed in 1987–88 (http://www.ssc.wisc.edu/nsfh/). The strengths of the design are the wide age range covered in the initial sample, reports from the main respondent and his or her spouse or cohabiting partner about their parents, stepparents, and all of their children (shared or from previous relationships), and the wide range of family attitudes and dimensions of relationships covered in the survey. Although the study is not ongoing, these data are still widely used to study intergenerational relationships because the survey included so many domains of family life. A serious weakness of the NSFH for studying these issues is that the sample was selected close to 30 years ago. These data cannot be used to investigate the dramatic divergence in family experiences of the college-educated and those with less schooling. The two follow-up waves to the 1987–88 interview were conducted at long intervals (one in 1992–94 and the other in 2001–02), limiting researchers’ ability to describe the temporal sequence of family events or how changes in attitudes were related to changes in family support. Because the NSFH is not ongoing it has very limited potential for augmentation, but data from the NSFH could be compared to data from a new national panel survey to assess change in many intergenerational support processes.

Modifying existing studies to broaden the interview content to include more measures of intergenerational ties would require that some existing content be deleted. Deleting existing content runs the risk of damaging the considerable strengths of existing panels. The major revisions in content would likely compromise the measurement and regular periodicity of current measures as well as the new content on intergenerational relationships. Disadvantages of modifying existing studies must be balanced against advantages of potential cost savings and the valuable prospective life history data in the existing panels, such as the HRS and PSID. A new panel survey would take time to mature sufficiently to allow examinations of how intergenerational processes unfold as individuals age. Understanding Society, the UK Household Longitudinal Study, is one model for combining a PSID-like study, the British Household Panel Survey, with an expanded sample and modified content (Buck and McFall [6]).

10. Could administrative records substitute for a new survey?

The high cost of survey data collection coupled with the expanding opportunities to link administrative records to study the intergenerational transmission of education, occupation, and earnings (Chetty [9]) require that the research community consider administrative records as an alternative to new survey data collection. For three reasons administrative record linkages cannot substitute for survey data on family support. First, administrative records cannot account for the time that parents and adult children devote to each other. Time is a valuable resource that parents use to help offspring with household tasks and child care. Adult children are much more likely to give time to parents than money (Schoeni [34]). This unpaid time would be worth considerable money if it were provided through the market (Folbre [15]). That women are more likely to do unpaid family work means that administrative records would significantly under-represent women's contributions to intergenerational family support. Administrative records also could not measure the symbolic and emotional meaning of time investments and the effects of these investments on individuals’ well-being.

A second reason that administrative records cannot substitute for survey data on family support is that administrative records are largely household-based (e.g., tax records and decennial census enumeration forms). Because most parents and adult children live in separate households, administrative records would miss all or most of the intergenerational ties that occur across households.

Finally, administrative records cannot measure attitudes and expectations about family support or the psychological outcomes of good or bad family relationships. There is great potential for a strategy that combines administrative records with survey reports, but without the survey observations the administrative record study would exclude major elements of intergenerational relationships, their determinants, and consequences.

11. Conclusion

Changes in the demography of U.S. families and weaknesses of existing data sources motivate the need for a new panel survey to examine the determinants and effects of intergenerational support. Although existing data have many strengths for studying family support processes they suffer from a lack of representativeness of today's U.S. population, are restricted to ages that limit the range of intergenerational ties that can be observed in the data, and content that ignores important dimensions of intergenerational support. Without new data, researchers cannot address such important questions as whether or not mothers reduce their paid employment to provide child care when they first become grandmothers, the effect of these decisions on the quality of intergenerational relationships, and on the adult offspring's availability to assist her in later life if she becomes frail. Existing data also cannot be used to investigate how facing this decision about a trade-off between time at work and time with family at a younger age, as is more common for those without a college degree or who are Black or Hispanic, affects family and individual well-being, compared to facing the decision at an older age. No single study or type of data can fulfill all needs, but a new family panel study would fill significant gaps in our understanding of intergenerational relationships.