Co-residence beliefs 1973–2018: Older adults feel differently than younger adults

Abstract

Objective:

This brief study examines support for co-residence (i.e. aging parents living with their adult children), and how age predicts support for this belief considering the rapidly aging U.S. population.

Background:

Co-residence, a form of intergenerational transfer between family members, can help facilitate care for aging parents as well as help older adults age in the community. Support for this type of co-residence was on the rise in the 1970s and 1980s.

Method:

Support for co-residence of older adults living with their adult children is estimated using 36,843 responses from the U.S. General Social Survey from 1973 to 2018. Descriptive analyses, logistic regression, and decomposition analyzes are used to test explanatory factors in trends, focusing on differences for older (age 65 and older) vs. younger (under 65) respondents.

Results:

Older adults are less supportive than younger adults of co-residence even as support has generally increased across time. Decomposition results show that a little over half of the difference between younger and older adults is explained by cohort replacement, with two-fifths of the difference unexplained by social or demographic factors.

Conclusion:

Findings suggest that although cohort replacement has contributed to an attitude shift over time, important age differences in attitudes remain. Older adults are less supportive of co-residence than younger adults.

Introduction

In the United States, many adults believe that families should support older family members in need of help rather than outside institutions like the government or private providers (Patterson, 2020). One way to care for aging family members includes intergenerational co-residence (hereafter, co-residence), where parents live with their adult children (Seltzer, 2019). Historically, rates of co-residence, broadly defined to include older adults living with children and vice versa, declined after World War II, but rates have been slowly increasing since the 1980s (Ruggles, 2007; Taylor et al., 2010). Although co-residence includes many different combinations of family members (e.g. adult children residing in the parent’s home) (Ruggles, 2007), we focus here in this brief report on older adults residing with adult children. In addition to facilitating direct family caregiving (Freedman & Wolff, 2020), co-residence can help keep older adults living in the community and out of institutions (Friedman et al., 2019).

We focus on attitudes toward co-residence because American eldercare behavior is influenced by attitudes about obligations (Cooney & Dykstra, 2011) and support for co-residence of aging parents with their adult children has been increasing over the last few decades (Alwin, 1996). Additionally, there have been large scale changes in the family realm, including lower rates of marriage, increases in divorce, remarriage, and cohabitation, alongside shrinking family sizes due to lower completed parity and rising rates of childlessness -- all of which might also have shifted attitudes about family support of older adults (Seltzer, 2019; Smock & Schwartz, 2020). As life expectancy has increased, the life course has been extended and more of the population is living long enough to develop health limitations that increase the need for help and caregiving. Given that age 65 is a socially significant turning point in the life course, as many older adults retire and become eligible for Medicare in the United States, age may especially influence attitudes toward support for older adults. We extend prior work by using the General Social Survey from 1973 to 2018 to describe long-term changes in support for co-residence and focus on age differences, including explanatory differences by age.

Background

Estimates show that one out of every five American households intergenerationally co-resided in some way in 2016 (Cohn & Passel, 2018), and a majority of two adult generation households are ones in which an adult children resides in the older adult’s home, not the older adult in the child’s home (Ruggles, 2007). However, when accounting for adults ages 65 and older, the proportion shifts to half of intergenerational households being an older adult co-residing in their child’s home (Ruggles, 2007). Co-residence may be important for supporting older adults as they age as it facilitates caregiving and exchanges (Freedman & Wolff, 2020; Keene & Batson, 2010), with shared housing serving as an economic resource to family members (Keene & Batson, 2010; Reyes, 2018).

In the U.S., there are strong normative prescriptions for the family to help aging adults (Patterson, 2020). Attitudes toward supports for older adults reflect perceptions of responsibility and prescriptive norms of behavior (Pew Research Center, 2015) and are often associated with and influential for social policy (Albertini & Pavolini, 2017; Henderson et al., 1995). Attitudes also influence personal behavior, with stronger support of filial obligation, or the idea that adult children should care for parents, being associated with a greater likelihood of caregiving in the U.S. (Cooney & Dykstra, 2011; Silverstein, 2006).

Previous research has found that age is associated with attitudes toward family care responsibilities for older family members, with young adults supporting filial obligation more than older adults (Daatland et al., 2011). In a study of Norwegians, younger adults were also more supportive of various ways to support older adults, including “children should have their parents to live with them when parents can no longer look after themselves” (Daatland et al., 2012). Younger adults may be more supportive of filial obligation and other family-based care for older family members because recent generations of younger adults post World War-II have been more likely to have their parents and grandparents alive for longer periods of their own life course (Daw et al., 2016) and contact with older adults may increase their positive attitudes toward older adults (Drury et al., 2016).

At the same time that families are favored for care, there is a pull of individualism and independence for older adults in the U.S. This drive for independence is especially strong as the introduction of Social Security has provided the financial resources to allow older adults to live independently; for instance in 1940 only 18% of older widows lived alone compared to 62% in 1990 (McGarry & Schoeni, 2000). As older adults have gained economic independence, they have largely chosen to live independently (Ruggles, 2007). Three out of four older adults want to remain in their home and community, although uncertainty about the ability to do so is common (Binette & Vasold, 2018). Although co-residence with family members is one way to remain in the community, given the preference for independence, older adults may be less supportive of co-residence compared to younger adults.

Co-residence Attitudes: Previous Studies and the Current Study

Previous longitudinal studies in the U.S. have found increasing support for co-residence across time and mainly relied on the General Social Survey (GSS). Studies first documented growing support for co-residence of parents in an adult child’s home in the 1970s and 1980s. Both Glenn (1987) and Okraku (1987) found increasing support of co-residence as a “good idea.” For instance, while only 31% of Americans found co-residence to be a “good idea” in 1973, by 1984, 41% agreed with the idea (Glenn, 1987). Singh et al. (1998) extended this work using GSS data from 1973 to 1989 to document both a similar increase in support for co-residence over time and that younger adults are more likely to support co-residence than older adults (Singh et al., 1998). Alwin (1996) further extended these analyses and found growing support for co-residence due to inter-cohort change, that is, each successive cohort of Americans were more supportive of co-residence than previous cohorts using GSS data from 1973 to 1991. Greater support was found primarily among post World War-II birth cohorts.

More recent, but single-year non-GSS studies also found strong support for co-residence. A 2005 poll by Pew found that over half (55%) of Americans see the family as responsible for supporting an older parent by providing co-residence with an adult child, compared to only 32% supporting adult children living in their parent’s home (Taylor et al., 2010). Other studies also found more support for co-residence if an aging parent needs a place to stay compared to when an adult child does (Seltzer et al., 2012). Younger adults (age 40 or younger) were generally more supportive of providing co-residential housing help to aging parents compared to older adults (Coleman & Ganong, 2008).

Despite important social and demographic shifts, including shrinking biological family size but increasing non-biological family ties (Seltzer, 2019; Smock & Schwartz, 2020), no study has examined changes in co-residence attitudes since 1991 in the General Social Survey (GSS). Therefore, we extend previous studies by leveraging recent waves of the long-running GSS in order to study how co-residence attitudes have changed from 1973 to 2018. Our main focus is on age differences in attitudes because we are interested in whether age alters attitudes, even among a cohort of adults who strongly supported co-residence as younger adults.

Data and Methods

The General Social Survey (GSS) is a nationally representative survey of U.S. adults, ages 18 to 89+ (capped by the GSS) conducted since 1972 (https://gss.norc.org/) (Smith et al., 2019). We used 27 years of surveys administered between 1973 and 2018 that include our main attitude measure for co-residence. Starting in 1988 the GSS moved to a ballot system where some questions are only asked of some respondents; we used the respondents within the ballots that received this question. Data was downloaded using the GSS Data Explorer.

Of the 39,664 respondents asked about co-residence from 1973 to 2018, we removed 395 respondents who did not answer the question. We further restricted our sample to 36,843 respondents who had no missing data on covariates. Subjective socioeconomic class had the largest percent missing at 4.5%, partially because it was not asked on the same ballot in 2006; all other items have low levels of missingness across the waves that we used. We adjusted for survey design and use the WTSALL weights for national representation (General Social Survey, 2018). Supplemental analysis comparing unimputed and imputed samples yield similar results; we presented results from the unimputed data in order to use decomposition.

Measures

Our main dependent variable was whether respondents support co-residence between older adults and their adult children (GSS item: AGED). The item asked: “As you know, many older people share a home with their grown children. Do you think this is generally a good idea or a bad idea?” Respondents could reply: a good idea, a bad idea, depends. Because we were concerned with support of this attitude, we create a binary measure of “good idea” vs. “depends/bad idea.” A multinomial logistic analyses showed that “bad idea” and “depends” work in the same direction – that is, as people age, they are more likely to say co-residence is a bad idea or depends (see Appendix Table 1). Analyses by GSS (Smith, 2017), alongside our own supplemental analyses, indicated that this question, although seemingly ambiguous, generally reflects attitudes toward older adults co-residing in their children’s home rather than vice versa (see Appendix Table 2).

Our analyses focused on age differences. In our models, we used an age spline at age 65, because 65 is a socially significant turning point in the life course when it is common to retire and become Medicare eligible. We controlled for birth year (cohort) and yearly national percentage of older adults 65+ in poverty to proxy potential period effects which we assume to be zero (estimates from Li & Dalaker, 2021). We also controlled for a series of individual characteristics correlated with attitudes and co-residence behavior across three categories: demographics, family characteristics, and economic indicators.

For control variables, we prioritized measures that are available across all, or most, years of our analyses (1973 to 2018). For demographics, we controlled for respondents’ gender using a binary measure of female or male. We used multiple GSS indicators together to measure mutually exclusive categories of race and ethnicity (White, non-Hispanic; Black, non-Hispanic; Hispanic; Asian, non-Hispanic; Other, non-Hispanic). Because the GSS did not start inquiring about Hispanic ethnicity until 2000, we followed the GSS’s suggestion for using the variable ETHNIC for pre-2000 Hispanic ethnicity (i.e., including: Spain, Puerto Rico, Mexico, Other Spanish). We used a continuous scale from 0 (never) to 8 (more than once per week) for religious service attendance. We measured the respondent’s regional location by Census region (Northeast, Midwest, South, West).

For family variables we measured whether the respondent is married or not. We also measured a continuous number of siblings and number of children the respondent had.

For economic indicators we measured whether the respondent is employed or not, and a categorical measure of their highest level of degree attainment (less than high school, high school degree, junior college, bachelor, and graduate/professional). We also measured their self-identified socioeconomic class (lower, working, middle, upper); we discuss income as an alternative measure in the sensitivity analyses.

Analytic Plan

We begin with a descriptive analysis illustrating how co-residence patterns have changed from the 1970s to the 2010s but focusing on how attitudes have always been stratified by age. We illustrate these trends using a shaded heat map to show support for co-residence. In trying to understand the age effects over time, we must make some assumptions because of the linear dependency of age, period, and cohort (Bell, 2020). Given that previous research on co-residence attitudes has suggested a cohort replacement effect in the U.S., alongside other research that shows age often stratifies attitudes toward family support of older adults in both the U.S. and other nations, we decide to model cohort and age, while assuming a null period effect (we attempt to relax this assumption by including percent of older adults aged 65 in poverty by survey year to account for possible period effects). Therefore, using a pooled sample we estimate a logistic regression model with an age spline at 65 to understand how age predicts support for co-residence controlling for cohort and other social and demographic characteristics. Finally, in order to decompose the difference between older and younger adults, we used a pooled sample to perform a Kitagawa–Blinder–Oaxaca decomposition using the “Oaxaca” command in Stata.

Results

In Table 1, we present a shaded heat map descriptive table that shows weighted percentages of support for co-residence by survey decade and ten-year age category. A darker shade of grey illustrates more support and lighter shades illustrate less support for co-residence. Looking across the table from left to right in the first row, from younger adults to older adults, patterns illustrate that younger people were more likely to be supportive of co-residence in the 1970s as indicated by darker greys, compared to older adults with lighter greys. Age differences between younger and older adults remained consistent across time with younger adults being more supportive than older adults despite growing support among all respondents over time (from the 1970s to 2010s). Overall, the age pattern was best illustrated by the total across time that showed that from the 1970s to 2010s, over half of young adults aged 18 to 29 (56.2%) supported the idea of older adults co-residing with their adult children, compared to just 28.2% of older adults ages 80 to 89+. The relative consistency of an age effect over time, despite cohort changes, further supported our assumption of no period effect.

Table 1.

Percentage Supporting Co-residence by Survey Decade and 10-year Age Categories

Age Categories
Survey Decade	18 to 29	30 to 39	40 to 49	50 to 59	60 to 69	70 to 79	80 to 89+
1970s	45.4%	37.4%	33.5%	33.6%	26.0%	23.9%	22.1%
1980s	57.6%	53.2%	47.7%	42.9%	30.2%	28.6%	24.5%
1990s	59.9%	53.9%	50.2%	42.0%	33.9%	21.3%	23.6%
2000s	56.2%	56.6%	55.6%	50.1%	40.2%	30.5%	25.1%
2010s	59.8%	57.3%	60.6%	61.6%	50.3%	40.8%	39.0%
Total	56.2%	52.8%	50.8%	47.6%	36.9%	29.3%	28.2%

Data: General Social Survey, 1973 to 2018, Weighted, N=36,843. Darker=more supportive of co-residence.

In Table 2 we present odds ratios from a logistic regression model predicting support for co-residence. An age spline at 65 showed that support for co-residence was generally the same for adults younger than 65, but at age 65, the slope changes slightly with older adults being less supportive of co-residence. We graphed the adjusted predicted age differences in support of co-residence in Figure 1 to show how this attitude shifted with age (vertical line indicates age 65).

Table 2.

Odds Ratios for Supporting Co-residence

	Odds Ratio	(SE)
Age
Age Spline (<65)	1.00	(0.00)
Age Spline (>=65)	0.98***	(0.00)
Control Variables
Birth Year	1.01***	(0.00)
National Percent of Older Adults (65+) in Poverty	0.97	(0.02)
Female	0.92**	(0.02)
Race/Ethnicity (White, non-Hispanic omitted)
Black, non-Hispanic	1.21***	(0.05)
Hispanic	1.31***	(0.07)
Asian, non-Hispanic	2.31***	(0.22)
Other, non-Hispanic	1.17*	(0.08)
Religious service attendance	1.03***	(0.00)
Region (Northeast omitted)
Midwest	0.87***	(0.03)
South	0.84***	(0.03)
West	0.91*	(0.04)
Married	0.84***	(0.02)
Number of Siblings	1.02***	(0.00)
Number of Children	0.99	(0.01)
Employed	1.05	(0.03)
Highest Degree Completed (Less than High School omitted)
High School	0.92*	(0.03)
Junior College	0.97	(0.06)
Bachelor’s	0.99	(0.05)
Graduate/Professional	0.98	(0.06)
Self-identified Class (Lower class omitted)
Working Class	0.93	(0.05)
Middle Class	0.84**	(0.05)
Upper Class	0.77**	(0.07)
Constant	0.00***	(0.00)

Data: General Social Survey, 1973 to 2018, Weighted, N=36,843

Figure 1.

Adjusted Predicted Probability of Supporting Co-residence by Age

In Table 2, other factors that were important in shaping these attitudes include birth year, demographic, family, and economic characteristics. More recent cohorts, represented by birth year, had slightly higher odds of supporting co-residence than prior cohorts. Women were less likely than men to say co-residence was a good idea. Black, non-Hispanic, Hispanic, and Asian, non-Hispanic respondents had greater odds of thinking co-residence was a good idea compared to White, non-Hispanics. Every one-unit increase in the frequency of religious service attendance was associated with a small increase in the odds of supporting co-residence. Compared to the Northeast, respondents in every other region were less likely to say co-residence was a good idea. Married respondents and respondents who identify with the middle and upper class compared to lower class, had lower odds of saying that co-residence was a good idea. Having more siblings was associated with increased odds of supporting co-residence.

Table 3 presents a weighted, pooled Kitagawa–Blinder–Oaxaca decomposition analysis for supporting co-residence. On average, only 31% of adults 65 and older thought that co-residence was a good idea, compared to 51% of adults younger than 65. Our models explained 12 percentage points of the 20-percentage point difference, or about 60.0% of the gap in support for co-residence by age. A majority of the portion that was explained (11 percentage points out of the 12 total) was due to cohort effects, confirming Alwin’s (1996) findings. The remaining 40% could not be explained by our model which suggests a true age effect for support of co-residence.

Table 3.

Decomposition of Support for Co-residence

	Coefficient	Percent of the Total Gap
Mean: 18 to 64	0.51
Mean: 65 and older	0.31
Difference	0.20	100%
Explained Total	0.12	60.0%
Birth Year	0.11
National % 65+ Poverty	−0.00
Demographics	0.01
Family	−0.00
Economics	0.00
Unexplained Total	0.08	40.0%
Birth Year	−4.46
National % 65+ Poverty	−0.13
Demographics	0.05
Family	0.00
Economics	0.01
Constant	4.61

Data: General Social Survey, 1973 to 2018, Weighted, N=36,843. Demographic: female, race/ethnicity, religious participation, region; Family: partnered, number of siblings, number of children; Economic: employed, educational attainment, self-reported socioeconomic class. May not sum to 100 due to rounding.

Sensitivity Analyses

We ran a series of sensitivity analyses to test the robustness of our main models. Previous studies that used the GSS AGED item coded it in a variety of ways (e.g. treated as numerical, removed depends, collapsed depends separately with bad idea or good idea), so we began by testing the item with a multinomial logistic model using the three original categories of good idea, depends, bad idea. Results showed that depends and bad idea function in the same direction by age, that is, as a person ages they are more likely to say co-residence was a bad idea or depends (see Appendix). We therefore collapsed bad idea and depends in our final analyses.

We tested imputed models with income as a control instead of subjective socioeconomic class; main results remain, and income was not associated with co-residence attitudes. We ultimately did not use income because of the high rates of missing. We did not use home ownership status in main models because it was not asked until 1986. Sensitivity analyses show that, in years available, owning a home was associated with greater odds of supporting co-residence. We also analyzed two separate measures of current household size including 1.) number of people in the household and 2.) type of household based on the roster (lives alone, parent’s house, child’s house, couple only). Having more people in the household was associated with more support for co-residence. Compared to those who live alone, all other groups were more likely to see co-residence as a good idea. However, these measures maybe endogenous to attitudes toward co-residence so we ultimately did not control for them.

Conclusion

Co-residence is a way for Americans to help care for or support aging parents, as it facilitates sharing resources and aids in providing caregiving (Freedman & Wolff, 2020; Keene & Batson, 2010; Reyes, 2018). We find that general support for co-residence grew over time and remains high in recent years. However, older adults, no matter the cohort or time period, are less supportive of co-residence than younger adults.

Older adults, ages 65 and older, are more likely to say that co-residence is a bad idea or depends, while younger adults are more likely to say that it is a good idea. Older adults may say co-residence is a bad idea or depends because they value independence and desire to remain in their own home (Binette & Vasold, 2018), whereas younger people may be more favorable because their parents are helping them and familial reciprocity is a strong norm itself in the U.S. (Silverstein, 2006). In addition, partially due to the extended life course of older adults in their lives, young adults may be more likely to receive help from parents in recent years compared to previous cohorts (Henretta et al., 2018), as intergenerational transfers generally flow in a downward fashion, from parents to children throughout the life course (Seltzer & Bianchi, 2013). Co-residence is one form of help young adults receive and recent evidence suggests they are more likely to live with their parents and do so until older ages than in the past (Choi et al., 2019). As young adults are currently benefiting from shared housing, they may view co-residence of older adults more positively and as way to pay them back for past support.

Using decomposition analyses to test what can potentially explain differences in opinions between older and younger adults, we find that our models explain 60% of the age differences and most of this is due to cohort replacement. This finding aligns with Alwin (1996). Similar to other attitudes toward family support of older adults (Daatland et al., 2011; Patterson, 2020), co-residence norms may be especially “pro-familisitic” (Glen, 1987). Recent cohorts of young adults are more likely to experience longer years of shared life with their parents and grandparents than previous cohorts (Daw et al., 2016), which may in turn increase their affinity for older adults overall (Drury et al., 2016).

Although our final models find little contribution from family measures, controlling for cohort replacement may partially adjust for these characteristics as family structures have also changed within these cohorts (Seltzer, 2019; Smock & Schwartz, 2020). The overall cohort changes in the understanding of families and their function may change independent of individual family characteristics as well. Further, our results suggest that different aspects of family structure appear to operate in opposite directions, that is, with marriage decreasing odds of thinking co-residence is a good idea and number of siblings increasing odds. Respondents with more siblings may find it easier to think, even hypothetically, that one of their siblings could take in their parent whereas married respondents may see their spouse as a primary caregiver, not an adult child.

The remainder of our decomposition (40%) is still unexplained, which reflects what we may think of as the pure age effect as well as any potentially untested factors that may account for this change. An important aspect of the age effect may be the shift from being on the hypothetical helping side of co-residence to actually being the older adult who may need the support from co-residence. In part, this change in attitude may reflect a desire to not be a burden to children on the part of older adults or maintain independence for as long as possible. As people age, they may come to value their independence more and see less benefit in co-residence as it becomes closer to reality for them.

This study has some limitations. There is only one item available across time to measure co-residence attitudes and it only captures one type of co-resident household – an older adult residing with adult children. We are confident in our results of this one item, however, because of GSS testing (Smith, 2017) and our own sensitivity analyses. Our analyses are limited in testing causality of attitude change. That is, rising rates of intergenerational co-residence, broadly defined, may influence attitudes or attitudes may influence co-residential behavior. Our modeling does not allow us to tease apart this relationship. Finally, because we prioritized control variables that were available in the same years as the co-residence question in GSS, our current explanatory variables may be limited.

We highlight three main areas of potential future research. First, future research could investigate the effect of policy environments, “safety net” programs, and costs of nursing home access as they may influence intergenerational co-residence and attitudes (Chen, 2017; Orsini, 2010). Second, information on immigrant status is limited due to the years questions were asked in the GSS, but future research could further investigate how country of origin, generational status, and potentially differing normative prescriptions may influence co-residence attitudes (Emery et al., 2019). Third, co-residential decisions and attitudes may depend on circumstances around the decision, including which child’s home, the adult child’s relationship to parent (e.g. whether it is a biological, adoptive, or step parent/child relationship), and parental need (Coleman & Ganong, 2008; Seltzer et al., 2012). Future research could further investigate under which circumstances co-residence is favored.

Given an aging population alongside shifting patterns in families as smaller and more complex, scholars have worried whether and how families may alter their support of older family members. Our results suggest that attitudes regarding co-residence, whereby an older adult hypothetically moves in with their adult children, receives strong support among most Americans and especially among younger people and more recent cohorts. Although attitudes shifted with post World War II generations, as multigenerational co-residence was also on the rise (Ruggles, 2007; Taylor et al., 2010), older adults are consistently less supportive of co-residence than younger adults. Conversations regarding care infrastructure investments should keep in mind the preferences of older adults themselves, who may wish to remain independent and in their own home compared to residing in their adult child’s home. However, many older adults go without home modifications and equipment that would allow them to live safely at home for longer (Lam et al., 2021). Programs and policies aimed at keeping older adults in their home (e.g. by providing mobility adaptions and supports in the home) may yield the most support from both younger and older adults.