Measuring Networks beyond the Origin Family

Abstract

Studies of social mobility typically focus on the associations between the socioeconomic characteristics of individuals and families in one generation and those same characteristics for the next generation. Yet the life chances of individuals may be affected by a wider network of kin than just the nuclear family, including grandparents, aunts and uncles, siblings, and even more remote kin. In planning new studies of intergenerational social mobility, researchers should consider the ways that more remote kin may affect socioeconomic success and hardship and design data collection strategies for collecting data on wider kin networks. Administrative record linkage and survey research have complementary advantages for identifying kin networks. Successful implementation of these approaches holds the promise of a much richer set of studies of intergenerational social mobility than most researchers have attempted thus far.

This article provides a rationale for considering the role of kin and other networks beyond the nuclear family in any new study of social mobility and discusses some relevant data collection design considerations. To orient this discussion I make several assumptions: (1) the major purpose of a mobility study is to provide data on the inter and intragenerational mobility on social and economic characteristics of both individuals and families, variously conceived; (2) the resulting data are intended to be widely useful for making comparisons within the population covered by the study and between this population and those in other times and places; (3) it will be a study with one or more large samples but shorter data collection instruments relative to other relevant studies that may take an in-depth approach to kin and other relationships; (4) replication of earlier mobility studies will be one important concern, although this is to be balanced by considerations of changes in American society and changes in social science understandings of social mobility; and (5) it is desirable to maximize the value of the data through linkage between the mobility study records and data on respondents contained in other studies or government statistical systems. In what follows, I first present some arguments for considering a wider set of kin networks and influences than is typical in mobility studies, briefly discuss nonkin networks and contacts, and then turn to practical issues of research design that arise in this type of data collection effort.

Rationale for Including Networks beyond the Origin Family

The main purpose of social mobility research is to examine the continuities and discontinuities in the socioeconomic statuses and positions of individuals and other social units both within and across generations. Even when the focus is explicitly on the educational attainments, jobs, occupations, earnings, and incomes of individuals, it is necessary to link individuals to other family members because the socioeconomic standing of individuals is inseparable from their family context. Most importantly, the “socioeconomic origins” of individuals are usually defined in terms of some combination of socioeconomic positions and statuses of their parents or parent-substitutes while the individuals of interest were children or adolescents. Nearly all intergenerational social mobility studies take a “two-generation” perspective, focusing on associations between social standings of parents and their offspring, whether assessed retrospectively, by sampling individuals and asking them about the characteristics of their parents, or prospectively, by sampling parents and asking them about or following up to observe the characteristics of their offspring once they reach adulthood. This perspective embodies several key assumptions: (1) the net associations between the socioeconomic positions or statuses of other kin and those of the individual of interest are small relative to the associations between the characteristics of nuclear family members and those of the individual of interest; (2) the relevant kin network for social mobility is known in advance and stable over time and place; (3) socioeconomic influence requires coresidence between the individual in question and his or her family; and (4) the “effects” of parental socioeconomic status are satisfactorily measured by parent-offspring gross or net associations, that is, by measures conditional on observed parent-offspring pairs rather than through models that take account of intergenerational demographic processes.

A different view of intergenerational effects is that, under some conditions, key influencers may include persons other than parents, including aunts, uncles, siblings, and grandparents. Rather than assume that the main goal of social mobility research is to examine variation in the link between parents and offspring, an alternative research strategy is to regard the key kin links for social mobility and immobility to be problematic, that is, as variable across time and place and the proper subject of investigation in their own right. Which family members are the key influencers within and across generations is likely to depend on patterns, differentials, and trends in kin availability; longevity; marital instability; family size; coresidence patterns; economic organization; and other institutional arrangements. Further, it may be desirable to ask how the distributions of families who are the units of conventional mobility analyses come about and whether the units of analysis on mobility studies are created by processes that are related to mobility itself (Mare 2011).

One precedent for considering a variety of kin relations is the widely appreciated variation in nuclear family structure that results from nonmarital fertility and marital instability. The absence of socioeconomic data for biological fathers in retrospective mobility surveys, once treated as a nuisance factor in mobility analysis, is now regarded as a common result of trends in family structure that have widened the range of family types to be distinguished in intergenerational studies and the range of persons for whom socioeconomic statuses should be measured (B. Duncan and Duncan 1969; McLanahan 2004; Biblarz and Raftery 1993). Another precedent is the evolving treatment of women and mothers in the analysis of intergenerational mobility. Once regarded as providing a limited or redundant contribution to family economic status, mothers are now treated as having an independent and growing role in the economic standing of families and the intergenerational transmission of socioeconomic position (Beller 2009). The incorporation of family structure and maternal influences into models of socioeconomic mobility not only provides a more nuanced picture of how families transmit their advantages across generations, but also reveals different trends in social mobility. Beller (2009), for example, shows that, whereas the association of father’s and offspring’s occupational status was relatively stable during the latter half of the twentieth century, the combined association of both parents’ statuses with offspring’s status increased markedly over this period. These kinds of changes in how we study social mobility reflect both growing sophistication in our understanding of social stratification and also our recognition of major changes in family demography and in the economic roles and statuses of women.

In recent years there has been growing interest in the role that grandparents may play in the intergenerational transmission of socioeconomic status (e.g., Mare 2011, 2014). Under some conditions, the socioeconomic positions and statuses of grandfathers and grandmothers (G1) may have net effects on the statuses and positions of grandchildren (G3), even after statistically controlling for the statuses and positions of the parents (G2). Estimates of net effects averaged over entire populations are mixed and may, for some populations, be very small (e.g., Hodge 1966; Warren and Hauser 1997). Where significant estimated effects do appear (e.g., Chan and Boliver 2013; Hertel and Goh-Samberg 2014; Pfeffer 2014), there inevitably remains the question of whether they arise from uncontrolled G2 characteristics. But regardless of the average size and causal status of net associations between G1 and G3 statuses and positions, the investigation of multigenerational effects is of interest for a number of reasons. To the extent that these associations exist, they provide a richer picture of intergenerational associations. The combined associations of G1 and G2 with G3 may be stronger than G2 – G3 association alone and may thus indicate greater persistence of social inequality. Additionally, measuring socioeconomic status in more than two generations reveals the standing of families over the longer run. Just as the distinction between permanent and transitory income is a meaningful way of thinking about individual and family inequality in an intragenerational context where incomes are measured at several points in time, it may be useful to think about within-family average levels and fluctuations in socioeconomic standing over multiple generations (as well as multiple siblings and their immediate families) (Friedman and Mare 2012). With more than two generations of data, moreover, it becomes feasible to consider the trajectories of families, including cumulative advantages and disadvantages across generations (DiPrete and Eirich 2006; Mare 2011; Mare and Song 2012; O’Rand 2002). Absent net G1 – G3 associations, such calculations may have limited analytic value. But to the extent that there is greater continuity in family socioeconomic status than a simple two-generation model would imply, a wider set of multigenerational descriptive statistics may be informative.

Perhaps more importantly, grandparent effects may vary meaningfully across time and place. In the United States, coresidential three-generation families declined in prevalence during the early twentieth century, with the decline in agriculture and the growth of salaried urban and suburban populations. In the late twentieth and early twenty-first centuries, however, despite a continued growth in residential independence of the grandparent generation, the actual prevalence and possibly the need for grandparents increased substantially. Increased survival of grandparents to advanced ages has made it much more likely that grandchildren will know and be in contact with their grandparents well into their adulthood. At the same time, the rise of single-parent families and the continued pressure on single parents to work while raising their children have increased the need for help from other family members. Additionally, the talents and resources that grandparents possess are highly variable, depending on their own socioeconomic histories, and thus grandparents may contribute socioeconomic variability as well as benefits to their grandchildren (Mare 2011; Uhlenberg 2009).

Non-Coresidential Kin and Kin Availability

In most traditional mobility studies, questions about family of orientation and family of procreation focus almost exclusively on family members who live together. In the Occupational Changes in a Generation II survey, for example, respondents were asked who the head of the family was when they were 16 and the educational attainments and occupations of the family head. (They were also asked about the educational attainment of their mothers.) But to focus exclusively on coresidential kin may be problematic generally and especially in the contemporary period in which family structures and family relations are so complex. Key family influencers may include persons with whom respondents did not live full time at important points in their lives. A father, for example, may be physically absent yet economically and socially very relevant to a young person’s environment and future opportunities. More generally, family and household are not identical, because households may include nonfamily members and, more importantly, key family members may not live in the household (Morgan et al. 2008). Whereas coresident kin may generally have a stronger effect on individuals than absent kin, this is an empirical matter that may vary across time and place.

A related issue is the availability of kin at various points in an individual’s life. Persons vary in the potential help they may receive from kin simply because they vary in how many and what types of living (or recently deceased) kin that they possess. Variation in kin availability not only limits what kin can affect an individual but also the degree of effect of the kin who are available. For example, the influence of any one grandparent may depend on how many other grandparents are still alive and available to provide influence. Relatedly, the degree of hardship associated with the absence of a parent may depend on the number, availability, and characteristics of nearby grandparents, aunts, or uncles.

To a largely unknown degree, the lack of attention to wider kin networks may have led to underestimates of the degree of family background influence on mobility and attainment, as well as a failure to explore important mechanisms of family influence. These considerations imply that, for key points in respondents’ lives (at a minimum in adolescence and at the time of data collection), the mobility study should obtain full rosters of who was/is present in the household and, if possible, inquire about the existence, location, and socioeconomic characteristics of key absent kin. To my knowledge, there are no well-established protocols for assessing kin availability that would be appropriate for a mobility study, although potentially useful resources are the household and family enumerations carried out in the National Survey of Families and Households (Bumpass and Sweet 1997).

The Demography of Social Mobility

Cross-cutting the concern with the effects of kin networks, especially those that extend across more than two generations, is the issue of how intergenerational effects come about. A basic insight about the social mobility process is that the way in which the socioeconomic distribution in one generation (period) is transformed into a distribution in a later generation (period) is through a process of demographic reproduction as well as intergenerational association of socioeconomic status. The mobility table itself is inadequate for showing how a socioeconomic distribution persists or changes because the mobility process is interdependent with differentials in timing and levels of fertility, mortality, and migration (O. D. Duncan 1966). O. D. Duncan’s (1966) observation on this issue has led most mobility researchers to focus on the mobility process itself and to avoid inappropriate inferences about population transformation. An alternative line of work, however, focuses on how demographic processes combine with social mobility to produce intergenerational change in educational and other socioeconomic distributions (e.g., Matras 1961, 1967; Preston 1974; Lam 1986; Preston and Campbell 1993; Mare 1997; Mare and Maralani 2006; Maralani 2013). These latter studies are based on the assumption that such an approach is needed not just for assessing population change but also for obtaining cogent estimates in the effects of the characteristics of an individual in the parent generation on the number and characteristics of his or her offspring. Individuals affect the next (and subsequent generations) through both the advantages and disadvantages they provide their children and also the partners they marry and the number of children they have. Such a concern becomes particularly acute when we consider multigenerational effects because the growth or extinction of a family across multiple generations may significantly modify the impact of an individual’s or family’s characteristics several generations forward (Mare 2011; Mare and Song 2012; Song and Mare 2013). Song and Mare (2013) show that even further complications arise when one takes a “two-sex” view of population change, inasmuch as intergenerational mobility then arises from transmission of status, differential net fertility, and the constraints and opportunities provided by socioeconomically differentiated marriage markets. Almost all mobility research takes a very limited view of the ways that demography interacts with social mobility. Thus far, only a handful of studies have fully considered multigenerational effects, differential fertility, and the mating process.

It is likely that the new mobility study will take an orthodox approach to the study of mobility, that is, obtain data for estimating parent-offspring associations in socioeconomic statuses and positions via record linkage or respondents’ retrospective reports of parents’ characteristics. The kinds of demographically informed mobility studies that I have briefly discussed in this section, however, can be facilitated with several extensions of standard data collection practices. It would be valuable to obtain data on individuals’ number of siblings, already a staple of mobility surveys, their parents’ number of siblings, and the socioeconomic attainments of spouse’s parents. Finally, these studies would benefit from obtaining information on grandparents and rosters of coresidential and non-coresidential kin, as discussed further below.

Other Networks

Social mobility and socioeconomic achievement may be principally affected by the characteristics of kin, but other influences may be important as well. Researchers have long recognized the role of “significant” others in socioeconomic achievement (e.g., Sewell and Hauser 1975), including teachers and peers, as well as other family members. An abundant literature considers the role of “neighborhoods,” variously defined, some of it even postulating multigenerational neighborhood effects on cognitive outcomes for youth (Sharkey and Elwert 2011). Further, it is important to recognize that teachers, older peers, clergy, and coworkers may be agents of intergenerational influence. The study of detailed mechanisms through which socioeconomic inequality in one generation are transmitted to the next generation remains an essential area of research. Given the broad goals of a new mobility study, however, I believe that the detailed study of mechanisms should be subordinate to obtaining the best possible data on continuities and discontinuities in the socioeconomic statuses and positions of families. Other types of networks and the identities and characteristics of influencers have been successfully explored using detailed longitudinal data on a more limited set of cohorts than is likely to be covered in a new national mobility study. It is likely that this approach will continue to be best for future work in this area. Nonetheless, it may prove valuable to obtain information on respondents’ places of residence, school attendance, or work at selected points in life (e.g., birth, adolescence, young adulthood) for the eventual purpose of linking respondent data to neighborhood, school, and work contexts.

Design Considerations in Obtaining Grandparent Data

In this section I briefly discuss some design options for obtaining data on grandparents within the context of ongoing infrastructural development for the study of intergenerational mobility that will enable both replication of past mobility studies and also continued monitoring of social mobility in a much richer way than has been feasible in the past (Grusky, Smeeding, and Snipp, this volume). The centerpiece of this infrastructural project is the cross-sectional and longitudinal linkage of large-scale administrative and other government-sponsored data sources, including the decennial censuses, the American Community Survey (ACS), Social Security earnings records, tax records from the Internal Review Service (IRS), and possibly other records of government program participation. But this basic vehicle for studying mobility can be further enhanced with links to sample survey data that may include richer data on a much more limited set of individuals and families. Such data may include established surveys, such as the Survey of Income and Program Participation (SIPP), or new surveys that have not yet been envisioned.

Record linkage

A key feature of this plan is that the data resources for studying intergenerational mobility grow with time. As time passes, obviously historical coverage grows as well, as the mobility experience of successive cohorts is recorded. Just as important, however, the web of potential kin relations that may be derived from the data grows in extensiveness and complexity. As the data for successive generations are added to the data it becomes possible to extend observed kin relations beyond two generations of family members who live together. Individuals can be linked to their offspring and their offspring’s offspring. Siblings can be linked together not only when they coreside as children but subsequently after they enter distinct households. Their spouses and children become part of the observable kin web as well, creating aunt/uncle/niece/nephew/cousin links and in-law links. The number, coverage, and richness of these observable kin links will of course depend on whether linked data come from samples (such as the ACS) or full, administratively defined populations (such as IRS or Social Security records). To the extent that socioeconomic information can be obtained for most of the linked family members, this will afford what, for the United States, is an unprecedentedly rich source of social mobility data.

At the same time, however, both in planning the record linkage project and even when the data are eventually available, rigorous analytic work will be required to assess the scale of population coverage and possible biases that such data may contain for the study of particular populations. This work is needed not only to plan and evaluate the technical challenges of record linkage per se but also to gauge the number and coverage of observable family links when parts of families are included in large-scale samples (such as those provided by the ACS or the long form of the 1990 decennial census) and other parts of families do not fall in the samples (or are observable with very limited information). And it is needed to assess the degree to which the kin relationships between individuals who do not live together, which will be covered with increasing richness as record linkage goes forward in time, constitutes a well-defined and interpretable sample of all such relationships.

Supplemental survey data

Supplemental survey data can enhance linked records from census, IRS, and other federal sources because they can include a much richer set of measures on any given individual or family than is provided by such basic data sources as the ACS or even such basic data sources considered together. From the standpoint of kin relations, surveys have great potential value because they can provide information on kin who do not coreside with survey respondents, thereby extending the kin network that can be incorporated into analyses of social mobility. In the longer run, as noted above, a wider kin network may be observable from linked records alone. But surveys make it possible to analyze some of these kin relations directly without waiting for further “maturation” of the linked records. Additionally, survey responses about nonresident kin may provide a way to crossvalidate kin information inferred from record linkage.

Despite the value of survey data on the socioeconomic characteristics of non-coresident and remote kin, the collection of these data presents additional challenges. Although these challenges crop up with almost all kin, I focus on grandparents in the balance of this discussion. Grandparent data may be collected using several possible strategies. An obvious approach is simply to ask respondents about selected socioeconomic characteristics of up to four of their grandparents. Such information is typically not obtained in general purpose surveys that contain other social mobility information in the United States, but some effort has been made on this in other countries, notably Treiman’s studies in Central Europe, South Africa, and China (Szelenyi and Treiman 1994; Treiman, Moano, and Schlemmer 1996; Treiman and Walder 1998). In the latter surveys, rates of nonresponse on items relating to grandparent socioeconomic characteristics are high, although, to my knowledge, no systematic assessment of the reliability and validity of these items has been carried out.

A second approach to the identification of grandparent effects is to obtain data on the socioeconomic characteristics of cousins. Just as correlations of socioeconomic characteristics of full siblings may provide global measures of the effects of parents’ characteristics on those of their offspring, correlations between the socioeconomic characteristics of cousins may provide global measures of the effects of grandparents’ characteristics. Cousins, of course, generally share at most two of the four grandparents possessed by each individual, and the total impact of grandparent characteristics must be computed from the combined associations of an individual’s characteristics with those of his or her paternal and maternal cousins. Cousin data can be obtained through data that maintain permanent family identifiers, such as in the Scandinavian population registers or in some prospective longitudinal surveys (Hällsten 2014). In a cross-sectional survey the only approach is to ask respondents to identify their cousins and to provide socioeconomic information for one or more of them. I am not aware of studies that have taken this approach, yet it appears to have several practical obstacles. First, a full enumeration of cousins is, for some persons, a very long list. Second, persons’ knowledge of cousins is of unknown reliability. Third, there may be considerable variation in whom respondents are willing to define as their cousins. And fourth, even if the study design narrows the focus to a particular cousin (for example, the eldest child of the eldest sibling of respondent’s mother), obtaining information for the correct cousin places a considerable burden on both the interviewer and the respondent.

A third approach is to regard respondents as the middle generation in a three-generation study and to obtain respondents’ reports of both their parents’ characteristics and also the characteristics of each of their offspring who is beyond a designated age (e.g., 18 or 25). This approach provides data for the association between grandparents’ and grandchildren’s characteristics, albeit censored by the incomplete fertility histories of the younger members of the middle generation and the young ages of some of the grandchildren. This design treats a cross-sectional study as both retrospective and also prospective, an approach that has been used successfully in other explicitly longitudinal studies (e.g., Panel Study of Income Dynamics, Wisconsin Longitudinal Survey).

Conclusion

A major new effort to collect data on intergenerational social mobility affords the opportunity to take a broader view of kin relations and their implications for social inequality and the intergenerational persistence of socioeconomic advantage and hardship. Rather than assuming a fixed and limited set of kin relations, such as father to son or mother to daughter or the coresident nuclear family, we should examine the possible effects of other kin and ask how they affect socioeconomic inequality and mobility. A permanent program of record linkage among the nation’s largest sources of social and economic data, supplemented with focused social surveys, will make it possible to address these key challenges to social science and social policy.

Biography

Robert D. Mare is a distinguished professor of sociology at UCLA and a senior research fellow at the Institute for Analytical Sociology, Linköping University. His research focuses on the connection between demographic processes and social inequalities, including inequality in educational opportunities, social mobility, youth unemployment, differential mortality, residential segregation and mobility, marriage markets, and family structure and poverty.