Abstract

Three decades ago, scholars discovered that elderly people in 19th-century America usually resided with their children. From the outset, the issue was cast in terms of family transitions over the life course (Chudacoff & Hareven, 1978, 1979; Smith, 1979). The only available data on intergenerational coresidence, however, were census cross sections. As Chudacoff and Hareven (1978) noted, “By necessity, our analysis … is cross sectional by its very nature. We are inferring, therefore, longitudinal patterns from cross sectional information” (p. 218).

Since then, virtually all analyses of the living arrangements of the aged in the 19th and early 20th centuries have relied on analysis of age patterns of coresidence to infer family transitions from cross-sectional evidence (e.g., Elman, 1998; Grattan, 1986; Grattan & Gutmann, 2010; Kramarow, 1995; Ruggles, 2003, 2007). Cross-sectional analysis has proved a powerful approach, stimulating lively debates and enriching our understanding of long-run family change. There are, however, significant limitations to a purely cross-sectional approach. Most important, cross-sectional analysis does not allow for assessment of how transitions in characteristics such as marital status, occupation, and place of residence are associated with particular kinds of family outcomes.

This article assesses transitions in intergenerational coresidence, here defined as residence of an elderly person with an own child. I use new linked samples of the U.S. census to investigate the mechanisms of the high intergenerational coresidence of the 19th century. I focus on transitions from middle age to old age, and I assess the effects of key characteristics including widowhood, retirement, and property. The results provide insight into debates about the 19th-century family system and have implications for the decline of intergenerational coresidence in the 20th century.

Background

When Chudacoff and Hareven (1978) first discovered the high coresidence of the elderly in 19th century Essex County, Rhode Island, they were careful to point out that it did not appear to reflect a simple system of old-age dependency. The elderly, they argued, typically remained in their own homes. Younger generation members usually shared the home, partly because housing shortages constrained their ability to set up independent households of their own. Accordingly, they argued that the relationship of the old and their coresidents was reciprocal rather than one-way support.

Over the following decades, this interpretation lost favor. Most subsequent writers argued that intergenerational coresidence was a form of old-age support (e.g., Costa, 1997; Elman, 1998; Gratton, 1986; Haber & Gratton, 1994; Hammel, 1995; Kramarow, 1995; McGarry & Schoeni, 2000; Schoeni, 1998; Smith, 1986; Wall, 1995). They maintain that the elderly in the 19th century preferred to live alone—in economic terms, that privacy was a normal good. Thus, they contend, most elderly in the 19th century faced challenges owing to widowhood, poverty, frailty, or sickness, and they were forced to rely on the assistance to their children. Hareven (1996) herself later adopted the dependency thesis, writing in 1996 that coresidence occurred “primarily when elderly parents were too frail to maintain a separate residence” (p. 2).

The dependency interpretation of 19th-century intergenerational coresidence had implications for the formation of intergenerational families. Chudacoff and Hareven initially suggested that coresidence occurred mainly when the younger generation remained in their parental home for an extended period after reaching adulthood. According to the dependency paradigm, however, elderly parents and their children moved in together after residing separately. As Hareven (1994) put it, “When elderly parents were unable to maintain themselves … aging parents had an adult child return to live with them, or they moved into a child’s household” (p. 442). Kertzer (1995) described this pattern as the nuclear reincorporation system.

My own interpretation, based on analysis of cross-sectional census data, differs significantly from the prevailing dependency paradigm (Ruggles 2003, 2007). As did Chudacoff and Hareven, I have concluded that most 19th-century intergenerational families probably formed when one child remained in his or her parental home after reaching adulthood. I have also argued that, although such families were usually mutually beneficial, the younger generation was probably more often dependent than was the older.

This article exploits a powerful new source to reexamine these hypotheses. Was 19th-century intergenerational coresidence a system of support for dependent elderly kin, or did it reflect intergenerational interdependence? This debate is critical for theories of family change. Implicit or explicit assumptions about historical change underlie virtually all social theory of the family (Therborn, 2004). Understanding the dynamics of intergenerational families in traditional American society is essential if we hope to unravel the dramatic reconfiguration in the living arrangements of the aged during the 20th century. By identifying the factors associated with key family-status transitions of the aging population in the later 19th century, I hope to shed light on processes of intergenerational family formation and dissolution.

Method

Study Population and Analytic Approach

After 1880, intergenerational coresidence of the elderly began to decline rapidly (Ruggles, 2007). To understand intergenerational coresidence before the onset of this decline, we must turn to the earliest linked IPUMS samples. Accordingly, this analysis uses samples of the 1850, 1860, and 1870 censuses linked to the 1880 LDS database. The censuses from 1850 to 1870 are especially appropriate for investigation of intergenerational dependency, as they included a question on the real property held by each individual, a key indicator of economic power in an era when most of the population resided on farms.

The IPUMS samples for 1850 – 1880 include the entire population of the states and all territories that later became states, with the major exceptions of “Indians not taxed” and the slave population. The study population is a clearly defined subset of the IPUMS census samples. The analysis is limited to persons first observed at ages 30 – 59 in 1850, 1860, or 1870 and observed a second time in 1880 at age 60 or older.

The analysis focuses on persons who were first observed as married household heads and their wives residing with at least one of their own children. This generally excludes persons who never had children; although there were a substantial number of childless in the 19th century, they did not have the opportunity for intergenerational coresidence and therefore are outside the scope of this analysis. The study population is restricted to persons with a spouse present at the first observation because this allows for assessment of the impact of the loss of a spouse on living arrangements. Focusing on persons who were married at the first observation is especially important for assessing the family transitions of women, as such analysis depends on information about husband’s property and occupation. Finally, the study population is limited to persons who were heading households or who were married to a household head, at the first observation. The loss of headship status, I hypothesize, is a key indicator of dependency. To assess whether the transition to old age was associated with a loss of headship status, it is essential to limit the analysis to persons who had achieved household headship by the first observation.

The IPUMS samples for 1850, 1860, and 1870 include 68,786 persons who meet all these criteria: persons aged 30 – 59 at first observation and aged 60 or older in 1880; and persons married, household head or wife of head, and with at least one child present in the household. Mortality was high in the mid-19th century; only about 32,983 of these persons would be expected to survive until the 1880 census (calculated from Hacker, 2010; Haines, 1998). The IPUMS-linked samples provide 6,783 cases that meet all the criteria, for an estimated linkage rate of 20.6%. The linkage rate for women was 16.9%; women who were widowed and subsequently remarried cannot be linked from census to census because their names changed upon remarriage. The linkage rate for men was somewhat higher, at 23.9%. These linkage rates reflect the priority of maintaining representativeness and accuracy rather than maximizing the number of links. Although the number of linked cases is small relative to the size of the IPUMS samples, it is sufficient to provide robust results.

The analysis merges data from three linked censuses (1850 – 1880 – 1860 – 1880, and 1870 – 1880), so the interval between first and second observations is 10, 20, or 30 years. I focus on three possible outcomes: (a) remaining as head of household with a child present in the household, (b) residing without a child, and (c) becoming a parent (or parent-in-law) of a household head. Figure 1 shows the distribution of outcomes by sex and age at final observation.

Open in a separate window

Figure 1

Living Arrangements of Study Population, by Age at Final Observation: U.S. Linked Samples, 1850 – 1880.

For women, the overall percentage of the population residing with children (as either heads or parents) declined until ages 70 – 74 and then increased, peaking at age 80 – 84. Among men, the percentage residing with children reached a low point at ages 75 – 79 and rose slightly for those who survived to 80 or older. The upturn in overall intergenerational coresidence for both sexes at the oldest ages could reflect both nuclear reincorporation and a greater propensity to coreside among cohorts born at the end of the 18th century (Ruggles, 2003).

Pooling the data from the linked samples, I use multinomial logistic regression to assess the factors associated with each transition. Following procedures described in Davern, Ruggles, Swenson, Oakes, and Alexander (2009), I use Taylor series linearization to adjust for the complex sample design and obtain reliable variance estimates.

Results

Table 2 presents complex sample multinomial logit regressions of family status transitions on the independent variables described in Table 1. The model compares transitions from head with children to residence without children and from head with children to parent of head. The results are presented as odds ratios.

The left column of Table 2 describes the factors associated with the departure of children. If the dependence theory of 19th-century coresidence were correct, we would expect to find that the elderly who had gone through adverse transitions—such as widowhood, retirement, or infirmity—would be likely to reside with their children. Moreover, we would expect that the older population with the most resources—the wealthy, literate, and the native whites—would be the most likely to live independently from their children.

None of these predictions holds true. For women, widowhood had no significant effect on continued residence with children. Widowed and remarried men were comparatively likely to lose their children, except for widowed or remarried farmers who remained in the workforce. Among married people, retirement was clearly associated with the departure of children.

The more property held by couples when they were middle aged, the less likely the couples were to reside without children when they became old. Thus, the elderly with the most resources were the ones most likely to remain household heads with children. This is the opposite effect that one would expect if privacy were a normal good, but it is exactly what one would expect if the prospect of agricultural inheritance provided a major incentive for children to remain in the parental home after reaching adulthood.

Chronic illness and disability, race and nativity, and illiteracy had no significant impact on residence with children. Accordingly, these findings provide no support for the dependence model of 19th-century intergenerational coresidence.

Migration to cities had no appreciable impact on intergenerational coresidence, but migration to rural areas—still the dominant destination in this period—was associated with the departure of children. Among nonmigrants, urban residents remained with their children substantially more often than rural people, which suggests that mid-20th-century assumptions about the impact of urban living on the family were mistaken. As Chudacoff and Hareven (1978) argued, the high coresidence of urban nonmigrants may reflect a housing shortage in rapidly growing cities, which made it difficult for young people to leave home and establish independent residences.

The right column examines the odds of transition from household head to parent of the head, and the results are striking. Farm wives who lost their husbands were about 127 times more likely to reside as mothers or mothers-in-law than were farmers and their wives who remained married and working. Women who had been married to nonfarmers and women who remained married to retired men were also extraordinarily likely to lose the household headship and become parents of the head. Among farm men who were both widowed and retired, the transition from head to parent was even more pronounced: They were 177 times as likely as married working farm couples to reside as parent or parent-in-law. Being either retired or widowed dramatically increased the odds of a transition, and the effect of retirement was substantially greater for farmers than for nonfarmers.

The fact that the effect is as strong for men as for women means that the transition in headship is not just a response to the death of the patriarch. Headship transitions depended on widowhood and retirement for both men and women. A shift from household head to parent or parent-in-law of the head was also encouraged by low property holding at the first observation, but none of the other indicators of potential need—illiteracy, chronic illness and disability, race or nativity—has a significant effect on the transition.

One final association is noteworthy. As expected, the number of children present in the household at the first observation was strongly positively associated with coresidence; it makes sense that the more children available, the greater are the odds of living with one of them. Not expected, however, was the inverse association between number of children and residence as a parent. Perhaps having more children to choose from gave the older generation more bargaining power, thus allowing them to retain the household headship longer.

Discussion

Most of the extensive literature on living arrangements of the aged in the 19th century has viewed intergenerational coresidence as a form of old-age support. Investigators have generally assumed that it was normal for the children to leave home when they reached adulthood and for the elderly parents to move in with their children when they became widowed, impoverished, or infirm. This interpretation is anachronistic, projecting modern ideas about old-age support onto an era that was very different. We should be wary of the assumption that older people who live with their adult children are necessarily doing so because of poverty or disability. Even today, there is substantial evidence that intergenerational coresidence often, and perhaps usually, results more from the needs of adult children than from the needs of their elderly parents (Aquilino, 1990; Choi 2003; Crimmins & Ingegneri, 1990; Kotlikoff & Morris, 1990; Smits, Gaalen, & Mulder, 2010; Spear & Avery, 1993; Ward, Logan, & Spitze, 1992; cf. Moehling, 1995).

Under the patriarchal system of the 19th century, there were especially good reasons for the younger generation to remain in the parental household. Old men had disproportionate control over valuable assets—not just farms but also workshops, retail stores, and other businesses. In this context, the younger generation had powerful incentives to remain behind and work on the farm or business with the expectation of eventual inheritance. For residents of high-density, rapidly growing towns and cities, housing shortages provided an additional motivation for the younger generation to remain with their parents.

The linked data allow us to identify factors associated with the departure of children from the parental home. Other than the number of available children, the most important predictor of the empty nest was lack of property. At least for men, being widowed or retired was also associated with residing without children. If the elderly moved in with their children when they were sick, impoverished, and widowed, we would find that those who lived independently from their children were the ones with the fewest challenges. In fact, those facing the fewest challenges were the most likely to reside with children. This is not to say that the elderly never moved in with their children for old-age support. Nuclear reincorporation certainly did occur on occasion, and it was doubtless fairly common. The evidence nevertheless indicates that reincorporation of dependent elderly parents was not the dominant pattern; more often, probably, the younger generation never left home.

Although adverse events did not increase the likelihood of intergenerational coresidence, such events did precipitate changes in the headship of intergenerational families. In particular, transitions from head to parent were closely associated with widowhood or retirement. This was true for both sexes, and the impact of widowhood was as dramatic for men as it was for women: Those who lost a spouse and did not remarry usually lost the household headship as well. It appears likely that people often made the transition from head to parent with no intervening period of separate residence.

It is important to acknowledge limitations of the analysis. The linked data samples are not adequate to completely describe 19th-century family transitions. We do not have continuous residence histories but just two observations for every individual. We cannot, therefore, track every change of status, and we cannot be certain about who moved in with whom. Although the linked data represent a marked improvement over the cross-sectional quantitative sources used in the past, they barely hint at the rich complexity of family relationships reflected in diaries, letters, novels, and the prescriptive literature of the period. Nevertheless, the data do provide important new results that enrich our perspective on 19th-century transitions, and the results presented here have significant implications for our understanding of American families before the great transformations of the 20th century.

The first quantitative description of the high frequency of intergenerational coresidence—Chudacoff and Hareven’s (1978) analysis of Essex County, Rhode Island—seems more and more plausible. Intergenerational coresidence in the 19th century does not seem to reflect a simple system of old-age dependency. As Chudacoff and Hareven (1978) suggested, the elderly probably did, for the most part, remain in their own homes, and the younger generation shared the home. As they argued, the relationship of the old and their coresident children was probably reciprocal: In particular, the young contributed labor essential for the functioning of the household economy, and the old contributed productive assets. Upon widowhood and retirement, household headship generally shifted from the older generation to the young. Only a minority of the 19th-century population, however, ever experienced the empty-nest phase of the family life course that is nearly universal today; especially among property-owning farm households, one child ordinarily remained coresident until the older generation died.

Acknowledgments

References