https://orcid.org/0000-0002-5321-3363
Abstract
Many U.S. parents share a household with an adult child in later life. However, the reasons parents and adult children coreside may vary over time and across family race/ethnicity, shaping relationships with parents' mental health. Using the Health and Retirement Study, this study investigates the determinants and mental health correlates of coresidence with adult children from 1998 to 2018 among White, Black, and Hispanic parents under age 65 and aged 65+. Findings show that the predictors of coresidence shifted with increasing odds that parents lived with an adult child, and several varied by parents' age group and race/ethnicity. Compared with White parents, Black and Hispanic parents were more likely to live with adult children, especially at older ages, and to indicate that their children helped them with household finances or functional limitations. Living with adult children was associated with higher depressive symptoms among White parents, and mental health was negatively related to living with adult children who were not working or were helping parents with functional limitations. The findings highlight increasing diversity among adult child–coresident parents and underscore persistent differences in the predictors and meaning of coresidence with adult children across race/ethnicity.
Introduction
Although the incidence of U.S. households with two or more generations of adults declined throughout the mid-twentieth century, the proportion of parents in midlife and later who reside with adult children has again been rising (Cohn and Passel 2018; Eickmeyer and Brown 2019; Ruggles 2007). However, characteristics of parent–adult child households and the reasons why family members decide to share a home have shifted in the preceding decades. Amid contemporary social and economic hardships, struggling adult children appear increasingly likely never to leave or to “boomerang” back to the parental nest (Fry et al. 2020; Kahn et al. 2013). Concordantly, recent research has found negative impacts of coresidence with adult children on aspects of parents' well-being, including mental health (Caputo 2019; Davis et al. 2018; Maroto 2017; Tosi and Grundy 2018). These studies provide a counterpoint to others showing some benefits of living with adult children, particularly for older parents (Aranda 2015; Chen and Short 2008; Courtin and Avendano 2016; Grundy and Murphy 2018). Despite this shorter term evidence, it is unknown whether and how the relationship between living with adult children and parents' mental health has evolved in recent decades alongside shifts in intergenerational coresidence contexts.
In addition, parent–adult child coresidence patterns are not static across race/ethnicity. Intergenerational coresidence is more prevalent in Black and Hispanic than White families, perhaps reflecting different needs, structures, and preferences (Cohn and Passel 2018; Kamo 2000; Keene and Batson 2010; Swartz 2009). Support and resource exchanges may also be less downstream and more reciprocal among intergenerational racial/ethnic minority households than among White households (Kahn et al. 2017; Kamo 2000; Reyes 2018; Swartz 2009). Thus, living with an adult child may have different implications for White, Black, and Hispanic parents' mental health over time. Furthermore, the balance of family needs and the meaning of living with adult children shifts as parents enter older age (Aquilino 1990; Ward et al. 1992). However, we know little about racial/ethnic differences in relationships between parent–adult child coresidence and mental health at any age or period.
Adopting the perspective that intergenerational coresidence decisions are products of family members' needs, resources, and preferences, this study addresses intertwined questions about the determinants and mental health implications of coresidence with adult children over time in the United States for parents of different races or ethnicities. Using the Health and Retirement Study (HRS), we explore whether the predictors of parent–adult child coresidence varied from 1998 to 2018 and across race/ethnicity for parents under age 65 and age 65+. We also assess the relationship between living with younger or older adult children and White, Black, and Hispanic parents' mental health during this period. The findings provide new insights about racial/ethnic differences in intergenerational coresidence and add to research highlighting the significance of ongoing changes in family demographic contexts for adults' well-being as they age.
Background
Coresidence With Adult Children and Parents' Well-being
The proportion of the U.S. population residing in households with multiple generations of adults reached a low of 12% in 1980, owing to factors including a strong industrial economy and improved state support for older persons (Cohn and Passel 2018; Goldscheider and Lawton 1998; McGarry and Schoeni 2000; Ruggles 2007). However, multigenerational living arrangements have again been on the rise, particularly amid crises such as the Great Recession and the COVID-19 pandemic (Fry et al. 2020). By 2016, 20% of Americans again lived in a multigenerational household (Cohn and Passel 2018). Although the most significant gains have been among young adults, intergenerational coresidence is increasing among adult children and parents of all ages (Cohn and Passel 2018; Eickmeyer and Brown 2019; Rappaport 2015). Nearly a quarter of adults aged 55–64 and 21% of those aged 65 and older lived in a multigenerational household in 2016 (Cohn and Passel 2018).
One explanation for the regrowth in multigenerational households is that lengthening life spans have created more opportunities for adult family members to live together (Bengtson 2001; Swartz 2009). However, much research exploring this trend focuses on adult children's hardships. Younger adults have been especially hard-hit by contemporary economic crises (Bell and Blanchflower 2011; Falk et al. 2021; Furstenberg 2010; Settersten et al. 2015). Confronted with decreasing wages, increasing education expectations, and the untenable costs of starting a family, many younger persons are delaying or forgoing marriage and parenthood (Furstenberg 2010; Kahn et al. 2017; Settersten and Ray 2010; Vespa 2017). The parental nest may provide a practical safety net amid these circumstances (Maroto 2017; Newman 2012; Sassler et al. 2008). Indeed, studies have shown that coresident adult children are less likely to be employed, married, or parents than their residentially independent peers (Caputo 2020; Sandberg-Thoma et al. 2015) and that they have become increasingly financially dependent on their parents over time (Kahn et al. 2013). Although research tends to focus on those in their 30s or younger, more older adult children are also living with their parents (Rappaport 2015). Most parents aged 50 and older report that adult children, regardless of age, are the main beneficiaries of transitions to coresidence (Caputo 2019; Choi 2003).
Scholars have drawn on evidence that adult children's needs are driving increases in parent–adult child coresidence to help explain negative associations with various dimensions of parents' well-being. For example, research shows that parents' assets and savings declined during periods that they lived with adult children (Maroto 2017, 2019). Studies have also found that coresidence is negatively associated with aspects of parent–child and marital relationship quality (Davis et al. 2018; Fang et al. 2021; Newman 2012; Umberson 1992; Ward and Spitze 2007). Further, some parents appear to internalize views that their coresident children have “failed to launch,” reporting embarrassment and stigma as culturally valued independencies are set aside (Aquilino and Supple 1991; Newman 2012). These material, emotional, and social stressors may harm parents' mental health. Several recent studies found that parents who transition to coresidence with adult children report a decline in mental health (Hu et al. 2020; Tosi 2020; Tosi and Grundy 2018). A previous analysis of the HRS found that parents who joined households with an adult child of any age around the Great Recession experienced an increase in depressive symptoms, especially if the child was not employed (Caputo 2019).
However, other studies highlight that living with adult children can carry significant benefits despite children's growing dependencies. Adult children may offer parents financial and practical support, help them navigate health problems, provide a sense of meaning in life, and reduce loneliness, all of which improve well-being (Lye 1996; Seltzer and Bianchi 2013; Swartz 2009; Umberson 1992). Coresident children may have especially strong obligations and frequent opportunities to extend these supports, which are likely of increasing importance and benefit as adults age (Compton and Pollak 2015). Underscoring that many parents also have needs that sharing a household can alleviate, research has shown that adults with lower incomes and health problems are more likely to live with children (Aquilino 1990; Caputo 2019; Seltzer and Bianchi 2013; Seltzer and Friedman 2014), and contemporary growth in coresidence has been linked to financial setbacks for older as well as younger adults (Bell and Blanchflower 2011; Eickmeyer and Brown 2019). Emphasizing the benefits of sharing a home, several recent studies based on Asian or European samples have found that parents residing with adult children have better mental health than those who do not, particularly when parents are older and living alone is the alternative (Aranda 2015; Chen and Short 2008; Courtin and Avendano 2016; Grundy and Murphy 2018).
Racial/Ethnic Differences in Parent–Adult Child Coresidence
Although research suggests that ongoing shifts in the contexts surrounding parent–adult child coresidence may have shaped its relationship to parents' well-being, the reasons why contemporary American families choose to share a home are as diverse as families are. Race/ethnicity is one of the most significant sources of heterogeneity in the frequency and meaning of intergenerational coresidence. Multigenerational households are consistently more prevalent among racial/ethnic minority persons than White persons, including those from Black, Hispanic, and Asian backgrounds (Cohn and Passel 2018; Fry et al. 2020; Keene and Batson 2010; Swartz 2009). These gaps appear to be narrowing as White families show steeper growth in intergenerational coresidence, but they remain significant (Cohn and Passel 2018; Fry et al. 2020; Kahn et al. 2017). In 2016, 16% of the White population lived in a multigenerational household, compared with 26%, 27%, and 29% of the Black, Hispanic, and Asian populations, respectively (Cohn and Passel 2018).
Explanations for these patterns often focus on material inequalities. As is well established, racial/ethnic minority households have lower average incomes, less wealth, and fewer assets than White households (Bhutta et al. 2020; Kamo 2000; Kochhar and Cilluffo 2017). Sharing a home to create economies of scale is an accessible and practical strategy to reduce collective financial hardship among family members with few cash resources (Sarkisian and Gerstel 2004; Swartz 2009). Consistently, several studies have linked lower socioeconomic status to higher rates of intergenerational coresidence among racial/ethnic minority families than White families (Angel and Tienda 1982; Cepa and Kao 2019; Kamo 2000; Lei and South 2016; Reyes 2022). Inequalities in economic capital may also contribute to differences in the structure of coresident relationships. Research suggests that overall, resource flows are more reciprocal among racial/ethnic minority multigenerational households than White ones, at least partly because of greater shared needs (Cepa and Kao 2019; Kamo 2000; Reyes 2018, 2022). Although transfers greatly favor the younger generation, racial/ethnic minority coresident children are less likely to be financially dependent on their parents than their White peers, and minority parents may receive more resources from their coresident adult children (Kahn et al. 2013, 2017; Reyes 2022).
Family structure is another source of racial/ethnic differences in the parent–adult child coresidence, particularly between White and Black families. Marriage rates are lower among Black than White adults, especially young adults (Horowitz et al. 2019; Hummer and Hamilton 2010). However, Black adults tend to become parents at younger ages and are more likely to be single parents (Kamo 2000; Sweeney and Raley 2014). Being unmarried, especially as a parent, is a strong predictor of coresidence with parents (Kamo 2000; Lei and South 2016). Some research has found that lower marriage and higher single-parenthood rates among Black young adults at least partly explain their greater odds of parental coresidence (Britton 2013; Kahn et al. 2017; Lei and South 2016). Single parents may live with parents to buffer the economic strains of child-rearing and receive practical help, such as childcare (Kahn et al. 2017; Kamo 2000; Swartz 2009). Thus, single-parent coresident children may also place greater demands on their parents.
Further, preferences and beliefs surrounding household sharing vary across race/ethnicity. Researchers have identified greater familism—the tendency to value family over individual needs—among racial/ethnic minority persons than White persons in the United States, especially among Hispanic and Asian adults (Britton 2013; Kamo 2000; Swartz 2009). Studies have also suggested that relative to their White peers, Black adults have more frequent contact with family members and provide more support to them (Ajrouch et al. 2001; Sarkisian and Gerstel 2004; Taylor et al. 2013). Strong connections with family members may increase the appeal of intergenerational coresidence. Concordantly, research has found that minority persons have more positive attitudes about sharing a home than their White counterparts (Cepa and Kao 2019). Further, older Black and Hispanic adults are more likely than their White peers to believe family members should provide housing for one another when needed (Burr and Mutchler 1999). Higher emotional connectivity to parents helps explain young racial/ethnic minority adults' greater odds of living with them (Lei and South 2016).
Research Gaps and Questions
Evolving economic and family contexts have accompanied the contemporary regrowth in multigenerational households. Research has linked increases in parent–adult child coresidence to adult children's growing financial dependence on parents (Kahn et al. 2013; Reyes 2022). However, systematic investigations of whether other parent and child characteristics have changed as predictors of coresidence remain absent. Coresidence predictors are likely to shift along with the balance of family needs as parents retire and enter older age, and we know little about the determinants of parent–adult child coresidence over time across race/ethnicity. Previous research has found that White coresident children's greater dependence on parents was stable up to 2010 (Kahn et al. 2017; Reyes 2022). Racial/ethnic differences in reasons and preferences for sharing a home may result in other variations in coresidence predictors. Drawing on the view that each generation's support needs, resources, and preferences shape coresidential living arrangements, we ask two questions:
1. Did the predictors of the likelihood that parents younger than 65 or 65+ lived with an adult child change from 1998 to 2018?
2. Did the predictors of parent–adult child coresidence during this period vary across White, Black, and Hispanic adults?
Although two U.S.-based studies found negative relationships between living with adult children and parents' mental health up to 2012 (Caputo 2019; Pudrovska 2009), they focused on relatively short windows. Further, it is unknown whether parental age and racial/ethnic variations in intergenerational coresidence shape the relationship between coresidence and parents' well-being. Researchers have hypothesized that negative impacts of coresidence with adult children on parents arise from strains that dependent adult children bring to the household and because it is arranged out of necessity but is ultimately undesired. However, these hypotheses may reflect a focus on younger coresident parents and a bias toward the dynamics of White households. The literature suggests that older and racial/ethnic minority adults may have a greater need to coreside with an adult child and view it more positively. Thus, these groups may ultimately reap greater rewards from these living arrangements than their younger and White counterparts. To gain insight into how living with adult children has impacted aging parents' mental health in the past two decades, we investigate two additional questions:
3. How was coresidence with adult children related to depressive symptoms among parents younger than 65 and 65+ from 1998 to 2018?
4. Did the relationships between coresidence with adult children and parental depressive symptoms differ across White, Black, and Hispanic parents?
Methods
Data and Sample
Data for these analyses come from the 1998–2018 waves of the Health and Retirement Study (HRS; Juster and Suzman 1995), which is supported by the National Institute on Aging and the Social Security Administration and housed at the University of Michigan. The HRS is an ongoing biennial panel survey of U.S. adults aged 50 and older and their partners. The sampling design includes a national probability sample and oversamples of Black and Hispanic adults. The study began in 1992 with a sample of 12,654 adults born between 1931 and 1941. Six cohorts have been added since. A new subsample of 3,000 respondents residing in census areas with high concentrations of Black and Hispanic households was added in 2010 to supplement the existing minority oversamples. Sampled persons' spouses and partners are interviewed regardless of their birth year. The cleaned and merged RAND HRS Family Data Respondent–Kid file is our baseline source of information on respondents' adult children (Bugliari et al. 2018). We draw data on coresidence transitions and on adult children in 2016 and 2018 directly from raw HRS household member–child files. Respondent–child dyad-level data were collapsed to the respondent level, producing counts of children with different characteristics, which we merged to the individual-level longitudinal RAND HRS data file (Bugliari et al. 2021).
Our analyses start in 1998 and follow respondents until 2018, utilizing 11 waves of data. The base sample includes all respondents who joined the study by 2010 and had at least one child or stepchild older than 18 at two or more waves (N = 27,551). We excluded respondents with missing data on non-time-varying characteristics (n = 267), missing data at all waves for any time-varying covariate (n = 306), and missing data on depressive symptoms for at least two waves of data (n = 1,480). We also excluded respondents in the heterogeneous “other” race/ethnicity group (n = 694) because their numbers were too small to support meaningful comparisons. The resulting analysis sample of 24,804 includes 17,349 non-Hispanic White, 4,441 non-Hispanic Black, and 3,014 Hispanic respondents.
Variables
Coresident Adult Children
Respondents reported the residential status and ages of all their and their partner's children at each wave. We use this information to create a dichotomous variable identifying those with at least one child older than 18 in their household. We also create variables counting coresident adult children.
Characteristics of Adult Children
We created separate variables identifying respondents with at least one single, nonparent, or nonworking adult child or coresident adult child at each wave based on reports of children's residential and partnership statuses, number of children, and work hours. We also distinguish respondents with coresident adult children who contributed to household finances or helped them with one or more activities of daily living (ADLs) at each wave. All variables for child characteristics are dichotomous. Analyses control for the time-varying total number of children older than 18 and the average age of all adult children.
Beneficiaries of Transitions to Coresidence
Respondents reporting a new household member were asked whether the residential move was primarily intended to benefit them, the new household member, or both parties, or occurred for some other reason. When the new household member was an adult child, we created time-varying dummy variables differentiating these answers.
Race/Ethnicity
Respondents indicating that they were Mexican American/Chicano, Puerto Rican, or Cuban American were coded as Hispanic. Those who did not fall into this group and identified as White/Caucasian in a follow-up question were coded as White, and those identifying as Black/African American were coded as Black. Individuals considering themselves primarily American Indian, Asian, or another race were excluded, resulting in three mutually exclusive categories for race/ethnicity: Hispanic, non-Hispanic White, and non-Hispanic Black.
Depressive Symptoms
At each wave, respondents were asked eight questions derived from the Center for Epidemiologic Studies–Depression scale (Radloff 1977), including whether they felt depressed, felt lonely, enjoyed life, and could not get going during the week before the interview. Positive items were reverse-coded, and all items were summed, producing scores ranging from 0 to 8.
Sociodemographic and Control Variables
Analyses control for respondents' age in 1998, gender, nativity (foreign-born vs. U.S.-born), and years of education (0–17+). We also distinguish periods when parents in the sample were under 65 or 65+. Time-varying scores on a five-item scale of ADL limitations are included in all models because functional health needs are an important determinant of intergenerational coresidence. The items asked whether respondents had difficulty bathing, dressing, eating, getting into or out of bed, or walking across a room because of a health or memory problem. Items were summed, producing scores of 0–5. Additional time-varying covariates include household income (in quartiles), employment status, marital/partnership status, and urban residence.
Analytic Approach
We begin with descriptive statistics and bivariate analyses assessing mean differences between those with and without a coresident adult child at the beginning (1998) and end (2018) of the observation period. We also show racial/ethnic differences among those with a coresident adult child in 1998 and 2018. In supplemental analyses, we show differences by the reported beneficiary of residential moves among those gaining an adult child household member in 1998–2000, 2004–2006, and 2010–2012.
Next, we estimate variations in the likelihood of living with an adult child from 1998 to 2018 among all parents, parents under 65, and parents age 65+ across various predictors, including race/ethnicity, with odds ratios (ORs) from mixed effects logistic regressions. These multilevel models nest observations at the different survey waves within individual respondents. The models include fixed effects, which indicate the overall mean effect of a variable at any given wave, and person-specific random effects or covariance parameters, which adjust for the correlation of residuals at different periods within individuals and indicate between-individual variance in intercepts and slopes. We interact all covariates and time to explore whether coresidence predictors shifted during the two decades. We also calculate ORs within race/ethnicity to capture differences in the determinants of coresidence among White, Black, and Hispanic parents.
The last analyses regress depressive symptoms from 1998 to 2018 on coresidence with adult children and race/ethnicity among all parents and according to parent age group using linear growth models. These models are also multilevel, nesting observations at the 11 waves within individual respondents. Interaction terms for race/ethnicity and coresidence reveal whether the mental health correlates of living with an adult child varied for White, Black, and Hispanic respondents during this period. We again split the sample by race/ethnicity to facilitate the interpretation of these patterns. In supplemental analyses based on the sample present and with nonmissing data on coresident children in 1998, we regress lagged measures of coresidence and depressive symptoms in 1998 on depressive symptoms from 2000 to 2018 to help clarify the direction of these relationships.
Results
Descriptive Patterns by Coresidence With Adult Children and Race/Ethnicity
Table 1 shows descriptive statistics for the total sample and in 1998 and 2018 by the presence of coresident adult children. Close to half the sample lived with a child older than 18 at one or more study waves. Fewer White than Black and Hispanic parents lived with an adult child at either time point, and parents sharing a household with an adult child were, on average, younger than those who did not. Relative to other parents, coresiding parents were more likely to be foreign-born or female, averaged fewer years of education, belonged to a lower income quartile, and were more likely to be working or to live in an urban area but less likely to be married. They also had more and younger adult children, on average, and were more likely to have at least one adult child who was single, a nonparent, or nonworking. Coresident parents had higher average depressive symptom scores and ADL limitation scores than independently living peers. Descriptive statistics using survey weights from the year respondents entered the study (summarized in Table A1, online appendix) show nearly identical patterns.
Differences between White, Black, and Hispanic respondents with at least one coresident adult child in 1998 and 2018 are shown in Table 2. Because minority respondents were oversampled in 2010, they are, on average, younger than the White sample in 2018. Among those with coresident children, more Black and especially Hispanic parents than White parents were foreign-born. Relative to White coresident parents, a higher proportion of Black parents were female, and a lower proportion were married. Relative to their White peers, both Black and Hispanic coresident parents had fewer average years of education, belonged to lower income quartiles, were less likely to be working (at baseline in 1998), were more likely to reside in an urban area, and had more children. Black and Hispanic coresident parents also had higher average depressive symptoms and ADL limitation scores than their White counterparts. These patterns are consistent with racial/ethnic differences in the full sample (shown for 1998 in Table A2, online appendix).
Turning to coresident adult children's characteristics, coresident minority parents had a higher average number of adult children living with them than their White counterparts, and their coresident children were younger in 2018. Most coresident children were single, but more Black than White coresident parents lived with a nonpartnered adult child at both time points. More White than minority coresident parents lived with a nonparent child, and fewer White than Black parents lived with a nonworking adult child. More Black and Hispanic parents reported that a coresident child contributed to household finances or helped them with ADL limitations at both time points.
Although the number of respondents reporting a new adult child household member in any given year is small, we tabulate these parents' reports of who primarily benefitted from the residential transition by race/ethnicity. The subsamples are respondents who joined the HRS sample in 1998, 2004, or 2010 and gained an adult child household member by their second interview two years later. As shown in Table A3 (online appendix), adult children were the most frequently reported beneficiary within each racial/ethnic group, and very few parents named themselves as primary beneficiaries. The proportion reporting a child beneficiary was highest in 2000 and declined by 2012, whereas the proportion indicating that moves benefitted both parties steadily grew. White parents were the most likely to indicate that a coresident move benefitted their adult child and the least likely to say it benefitted both parties. By 2012, approximately one quarter of White parents, 42% of Black parents, and 54% of Hispanic parents with a new coresident adult child reported that the move was mutually beneficial.
Odds of Living With Adult Children Over Time and Across Race/Ethnicity
Table 3 displays ORs from mixed-effects logistic regressions of coresidence with adult children from 1998 to 2018 on race/ethnicity and other predictors. Note that the average number of waves at which each individual was observed ranges from six for the model without age restrictions to approximately four in the model for periods when parents were under 65. In the non-age-restricted sample in the first model, the OR for time (1.357) indicates that the odds of a parent living with at least one adult child increased by 35.7% in any given survey wave from 1998 to 2018, net of control variables. Black and Hispanic parents were, respectively, approximately two (OR = 1.953) and five (OR = 4.922) times as likely to live with an adult child as their White counterparts. The odds of living with an adult child increased by approximately 13% with each year of age at any given wave. Because age is measured at baseline rather than used as the time scale, this term may capture birth year effects as well as chronological aging effects. Women were nearly three times as likely as men to live with an adult child. Odds of living with an adult child were higher for foreign-born, working, urban, and health-limited parents and lower for those with more years of education and who were partnered. Parents with more adult children were more likely to live with one, but coresidence odds decreased as the average age of adult children increased. Parents with at least one single adult child were more than 15 times as likely to live with an adult child at any given wave as those whose adult children were all partnered (OR = 15.206). Those with at least one nonparent or nonworking adult child had approximately 90% greater odds of coresidence than peers whose children were all parents or working (ORs = 1.915 and 1.891, respectively).
Table 3 also shows ORs for coresidence with adult children when respondents were under 65 and 65+. Patterns are very similar. Notably, however, minority respondents' greater odds of living with adult children were especially high at ages 65+. Being married had a stronger inverse relationship to living with an adult child for parents ages 65+ than for younger peers, and ADL limitations predicted higher odds of coresidence for older parents but not for those under 65.
Figure 1 shows the predicted probability of living with an adult child from 1998 to 2018 among White, Black, and Hispanic parents under age 65 (panel a) and aged 65+ (panel b) based on models from Table 2. Although the probability of coresidence increased within each group, parents under age 65 had higher predicted odds of living with an adult child and showed steeper gains in coresidence than those aged 65+. Hispanic parents stably had the highest probability of living with at least one adult child, especially when under age 65; White parents stably had the lowest probability. The gap in the predicted odds of coresiding with an adult child between White and minority parents was greater at ages 65+ than under 65.
We next explored whether the predictors of coresidence with an adult child varied over the study period with interaction terms for each covariate and time. Significant interaction effects for the whole sample and among respondents under age 65 and aged 65+ are shown in Table 4. For the non-age-restricted sample, the positive effect of being a year older on the odds of coresiding with an adult child increased with each successive survey wave. Being female, having more adult children, and having an adult child who was not working became increasingly important predictors of coresidence. Having older adult children grew as a predictor of living apart from them. At the same time, having more years of education, being employed, and having at least one adult child who was either single or a nonparent declined as predictors of living with an adult child. In addition, being married or partnered declined in predicting parents' lower likelihood of sharing a home. The predictive power of other characteristics, including race/ethnicity, did not change during this period.
These relationships were very similar among parents under age 65 and aged 65+, although the effect sizes are smaller in the age-limited samples. One notable difference is that Black parents' greater odds of living with an adult child decreased somewhat over time among adults aged 65+.
In Table 5, we split the sample by race/ethnicity to explore differences in the predictors of coresidence among White, Black, and Hispanic parents. To highlight meaningful differences, we present significant interactions between each covariate and race/ethnicity in Table A6 (online appendix). The increased odds of living with an adult child associated with aging were greater among Black parents and smaller among Hispanic parents relative to White parents. Being female was a stronger predictor of living with an adult child at each wave among Black parents than among White respondents, and being foreign-born predicted higher coresidence odds among Black and Hispanic parents but not White parents. Education predicted a greater decrease in the odds of coresidence among White parents than among their minority peers. White parents in the lowest income quartile had the greatest odds of coresiding with an adult child. However, among Black parents, the odds of coresidence increased with income. Being married had a stronger inverse relationship with coresidence among White parents than other parents, and ADL limitations predicted greater odds of coresidence among White parents only. Also, depressive symptom scores predicted greater odds of coresidence with an adult child at each wave among White parents but were inversely associated with coresidence among Black parents. Having a single adult child was a stronger predictor among White parents than minority parents, although it was highly significant among all three racial/ethnic groups. Hispanic parents were more likely than White parents to coreside with an adult child when at least one of their adult children was a nonparent, and having a nonworking adult child was a somewhat stronger predictor of increased odds of living with an adult child for White than other respondents.
A dichotomous variable for under 65 reveals increased odds of living with adult children for younger White parents but not for their minority peers. Interactions between race/ethnicity and predictors for those over age 65 and aged 65+ in Table A6 show nuanced differences in effect sizes but are highly consistent with the non-age-restricted results.
We next explored racial/ethnic differences in over-time shifts in the predictors of coresidence with adult children. Table 6 displays odds ratios for the significant differences between White and minority respondents, based on triple interactions for race/ethnicity, time, and predictors. Overall, these analyses indicate greater shifts in the predictors of coresidence among White parents than among their Black or Hispanic peers. Specifically, White parents showed greater decreases in the power of years of education, partnership, and having single or nonparent children to predict coresidence and greater increases in coresidence odds linked to older age, being female, and having nonworking children.
Coresidence With Adult Children and Depressive Symptoms
Table 7 presents fixed-effects and covariance parameters from multilevel linear growth models regressing depressive symptom scores from 1998 to 2018 on time-varying coresidence with younger and older adult children, race/ethnicity, and other covariates. Model 1 shows that living with an adult child was not a significant predictor of depressive symptoms at any given wave in the non-age-restricted samples. Depressive symptom scores decreased over time and with age. Black, Hispanic, and female parents had higher depressive symptom scores than their White and male counterparts. Higher education and income, being employed, and being partnered predicted lower depressive symptom scores. ADL limitations were strong predictors of depressive symptoms, and having older adult children was associated with higher scores (net of respondents' age). The random effect for time was significant, indicating significant variations in depressive symptoms over time between individuals.
Model 2 shows that coresiding with an adult child who was single, nonworking, or helped with ADL limitations predicted higher depressive symptom scores. Parents living with an adult child who contributed to household finances had somewhat lower depressive symptom scores than parents who lived apart from all their adult children at any given wave. To put the magnitude of these relationships into context, the association between living with a nonworking child and depressive symptoms (0.086) was roughly equivalent to the effect of each year of lower education (0.081). Living with a child who helped with ADL limitations (0.188) was a stronger predictor of depressive symptom scores at any given wave than being female (0.156). Effects of living with single adult children (0.054) and adult children who contributed to the household (−0.045) were weaker. The main coefficient for coresidence with adult children in this model remains nonsignificant and can be interpreted as the relationship between depressive symptoms and living with a child without the characteristics included (i.e., who was partnered, working, a parent, and who did not help with household finances or ADL limitations).
The significant and negative interaction between coresidence with adult children and time in Model 3 suggests that positive associations between depressive symptoms and living with an adult child decreased over the observation period. Further, the main effect for coresidence with an adult child is significant and positive in this model. In the context of the other coresidence covariates, this term indicates that baseline coresidence with an adult child who was partnered, a parent, and working but did not assist with finances or limitations predicted depressive symptom scores that were 0.074 higher than those of parents not living with an adult child. Models interacting other coresident child characteristics and time (Table A5, online appendix) similarly show that relationships with depressive symptoms decreased over the study period and suggest that living with most types of adult children (except for nonparent children) at baseline predicted increased depressive symptoms. Last, significant and negative interactions in Model 4 imply that the baseline relationship between living with adult children and depressive symptoms was stronger for White parents than for their Black and Hispanic peers.
Models 5–12 of Table 7 show relationships between depressive symptoms and coresidence with adult children for respondents under age 65 and aged 65+. Overall, the results are similar. However, the main effect for coresidence with adult children is significant only in Model 8, including time and race/ethnicity interactions among adults under age 65. Living with single children is not a significant predictor of depressive symptoms in the split-sample analyses, but the relationship between living with nonparent adult children and depressive symptoms among parents under age 65 is significant. Living with an adult child who contributes to household finances predicts slightly lower depressive symptoms among parents aged 65+ but not for younger parents. In addition, the relationship between having a coresident adult child and depressive symptoms does not significantly differ between White and Hispanic parents aged 65+.
To explore differences in the relationship between coresidence with adult children and depressive symptoms across White, Black, and Hispanic parents further, we split the sample by race/ethnicity in Table 8. The first models in each group show that before we account for coresident children's family and work status or their provision of help to parents, living with an adult child predicts higher depressive symptom scores among White parents but not among their Black or Hispanic peers. The effect size is relatively small but significant at p < .001. The increase in depressive symptoms associated with living with an adult child for White parents (0.057) is more than half the size of the effect of having one less year of education (0.098).
Models 2, 5, and 8 in Table 8 show that living with nonworking adult children or those that help with ADL limitations predicted higher depressive symptoms among White and Black parents but not among Hispanic parents. Accounting for these coresident child characteristics also explains relationships between living with an adult child and depressive symptoms for White parents. Further, the negative relationship between living with an adult child and depressive symptoms is slightly significant for Black and Hispanic parents: those who lived with a nonhelper, partnered, parent, and employed adult child in these groups have slightly lower depressive symptom scores than peers who did not live with any adult children. Last, significant and negative interactions between coresidence with an adult child and depressive symptoms among the White and Hispanic samples in Models 3 and 9 suggest that relationships between coresidence and depression decreased over time for these two groups of parents.
Figure 2 shows predicted depressive symptom scores by coresidence with an adult child from 1998 to 2018 among White (panel a), Black (panel b), and Hispanic (panel c) parents, based on Models 1, 4, and 7 of Table 8. Predicted scores were initially highest among Hispanic parents and then among Black parents, but scores decreased somewhat during this period for all three groups and converged toward similar levels. Differences between those with and without coresident adult children are distinguishable only among White persons, with coresident parents showing stably higher scores.
Age-restricted analyses for each racial/ethnic group are summarized in Table A6 (online appendix). Differences of note include that nonparent coresident children predicted lower and nonworking coresident children predicted higher depressive symptoms among White parents under age 65 but not among those aged 65+. In models before we account for coresident children's status and their provision of help, Black parents aged 65+ with coresident adult children reported lower depressive symptom scores than noncoresident peers.
Table A7 (online appendix) shows key findings from models regressing depressive symptoms from 2000 to 2018 on covariates, including lagged values of depressive symptoms and coresidence with adult children in 1998. These models include only respondents with nonmissing data on coresidence variables and depressive symptoms in 1998, significantly limiting the sample. Also, because the survey waves are two years apart, changes in coresidence arrangements between those years and shorter term effects of coresidence on depressive symptoms may be missed. Even so, the results offer valuable insights into the potential ordering of these relationships. The direction of most lagged coefficients aligns with those in the nonlagged analyses for the full sample. For example, in the non-age-restricted analyses in Table A7, coresident children in 1998 had a nonsignificant but negative link to depressive symptoms, consistent with Model 1 of Table 7. Coresidence with adult children who helped with ADL limitations was positively linked to later depressive symptoms, as it was when measured contemporaneously in Model 2 of Table 7.
However, a few differences are notable. When we control for lagged depressive symptoms, living with adult children in 1998 does not predict poorer subsequent mental health among White parents two years later. This finding supports the possibility that White parents with poorer mental health select into coresidence with adult children. Also, lagged coresidence with a nonparent child predicts higher later depressive symptoms in Table 7, including for those under 65 and White parents. It is possible that having lived with an adult child who is a parent had positive impacts on subsequent mental health in 1998 but that younger parents who were more depressed were more likely to select into living with children who are parents over the full observation window. In other words, this difference may reflect the shifting meaning of coresiding with an adult child who is (or is not) a parent across cohorts and over time. Similarly, negative contemporaneous relationships between living with a child who helped with household finances among the nonrestricted and 65+ samples in Table 7 are not present in the lagged models, suggesting that older parents with better mental health may be more likely to live with children who help with finances.
Discussion
Contemporary growth in multigenerational households in the United States is often attributed to the social and economic hardships of adult children, who have increasingly sought refuge in the parental nest. Scholars have drawn on this trend to explain findings that living with adult children has negative impacts on various aspects of parents' well-being, including mental health. However, it is unclear whether and how the precursors of parent–child coresidence and its relationship with parents' mental health have evolved in recent years owing to limited longitudinal analyses. Further, these discussions often overlook parental age-based and long-standing racial/ethnic differences in need and norms surrounding sharing a home. In addressing these gaps, we investigated linked questions about the predictors and mental health implications of coresidence with adult children over the past two decades among White, Black, and Hispanic parents under age 65 and aged 65+.
Consistent with other studies, we found growth in the proportion of parents under age 65 and aged 65+ who lived with at least one adult child. We also found that many parent and adult child characteristics changed in their propensity to predict coresidence during this period in ways that were overall consistent for older and younger parents. Some patterns resonated with the view that growth in intergenerational coresidence reflects the increasing primacy of child needs. For example, having a nonworking adult child became a stronger predictor of parental coresidence with adult children over time. However, other findings imply greater complexity than the child-need narrative suggests. Having a single or nonparent adult child weakened in predicting coresidence, meaning that greater proportions of coresident children had attained family-related markers of adulthood. At the same time, parental education, marital status, and employment status became less salient predictors of coresidence. Supplemental descriptive analyses provide further insight into questions of need. These analyses showed that adult children were the most frequently named beneficiaries of coresident moves, and parents rarely indicated that a move primarily benefitted themselves. However, the proportion reporting a child beneficiary was highest in 2000, whereas those indicating that the move benefitted both parties grew significantly by 2012. It is possible that parents systematically underestimate their own benefit in these situations, and we do not have input from adult children about whose needs such coresidential moves serve. Still, we think these findings, along with the mixed patterns in the coresidence predictors over time, highlight increasing diversity among the types of parents and children who live together, underscoring that growth in child need is only one part of the explanation for increasing parent–adult child coresidence. Changes in collective economic resources and shifts in cultural beliefs about coresidence (not captured in our study) are also likely influential.
We also assessed racial/ethnic differences in coresidence with adult children and its predictors during the two-decade observation window. A few notable patterns emerged. Black and Hispanic parents were stably more likely to live with an adult child than White peers, especially at ages 65+. Although income was inversely related to the likelihood that White parents lived with at least one adult child, Black parents in the highest income quartile had the greatest odds of living with an adult child. Further, poorer mental and physical health predicted higher odds of coresidence among White parents only, and having single adult children and having nonworking adult children were especially salient coresidence predictors among White parents. Higher depressive symptom scores predicted somewhat lower odds that Black parents lived with an adult child.
White parents also exhibited an overall greater change in coresidence predictors than their Black and Hispanic peers. These findings underscore that parent–adult child coresidence has consistently had different precedents and taken different forms for White, Black, and Hispanic families, with few differences by parental age group. They also suggest that needs may more strongly predict coresidence among White than minority families. Still, Black and Hispanic parents are more likely than White parents to describe moves involving adult children as mutually beneficial. This finding may reflect that coresidence moves are overall more voluntary and viewed more positively among minority than White parents. Factors other than absolute economic and health needs, such as preferences for family closeness and norms surrounding the provision of practical support, may more prominently influence decisions to coreside in racial/ethnic minority families than White families.
Further examining the meaning of coresidence with adult children for parents over and under age 65 during the last two decades, we explored mental health as an outcome. Living with a nonworking child or one that assisted with health limitations predicted higher depressive symptom scores among parents in both age groups. The relationship between living with an adult child who helped with ADL limitations and depression is likely to reflect the mental health toll of declining physical health. Although nonworking coresident children may negatively impact parents' mental health through their financial dependence, nonworking children may help parents in other ways, such as by caregiving. Future research may further explore these possibilities with analyses focusing on links between coresidential child characteristics, physical health, and financial well-being. Residing with nonparent adult children was inversely related to depressive symptoms among younger parents, meaning that living with children who also had children predicted higher depression scores. Parents aged 65+ who lived with a child that helped with household finances had lower depressive symptom scores. The magnitude of the significant relationships between living with adult children and depressive symptoms in any given year are often quite small, but we believe that these relationships are intuitive and meaningful. Supplemental analyses using the 1998 sample and including lagged baseline depressive symptoms and coresidence variables in 1998 support the possibility of a causal link between poorer mental health and living with adult children who help with ADL limitations. These analyses also suggest that older parents with better mental health may select into living with children who help with finances. We interpret these findings as showing that the balance of costs and benefits to parents is key to understanding why parents and adult children coreside and how living with adult children is related to parents' mental well-being. Indicators that coresidence occurred within the context of child economic hardships or parental health problems highlight circumstances in which these living arrangements are likely to be stressful. Coresident arrangements that are reciprocal and voluntary, such as when family members opt to create economies of scale, may be more likely among parents who already had better mental health resources.
Mental health correlates of living with adult children also varied across race/ethnicity. Living with an adult child predicted higher depressive symptom scores among White parents, which lagged analyses suggest may be attributable to White parents with poorer mental health selecting into coresidence with adult children. Living with adult children who were nonworking or helped with limitations predicted higher depressive symptom scores among White and Black parents but not Hispanic parents. Also, coresident Black parents aged 65+ had better mental health than their independently living peers, and Hispanic and Black parents experienced slightly better mental health when coresiding with adult children who had met adulthood markers than when living apart from all their adult children. Thus, the mental health correlates of living with an adult child appear to be contingent on family race/ethnicity and child characteristics and overall are more negative for White parents than for Black and Hispanic parents.
Notable limitations of our study include the inability to include other racial/ethnic groups, particularly Asian parents. Asian Americans are the most likely racial/ethnic group to live in multigenerational households, and studies reveal particularly strong norms of filial responsibility among persons from some Asian backgrounds. Coresidence may thus have very different implications for Asian parents' mental health. We also did not explore three-generation households or consider parents' caregiving for coresident grandchildren in this analysis. Although caring for grandchildren can be costly and stressful, it may also offer unique opportunities for emotional fulfillment and engagement, enhancing health and cognitive functioning. Also, although we stratify the sample by age, aging effects are indistinct from cohort effects because we use survey wave rather than age as the timescale for these analyses. Last, we do not account for the duration of coresidence episodes. We hope that future research will begin to fill some of these gaps and extend these questions to other health outcomes, such as cognitive functioning.
Still, our analyses offer distinct contributions to research on intergenerational coresidence. First, they suggest that contemporary growth in parent–child coresidence is more complex than the child-need narrative implies. We found evidence of growing diversity among coresident parents in both age groups, which we think indicates that the circumstances under which family members opt to coreside are broadening as social and economic contexts shift and coresidence becomes a common and accepted practical resource-sharing strategy. Second, the findings underscore distinct and persistent racial/ethnic differences in parent–adult child coresidence. Resonating with work finding more positive attitudes toward coresiding with other adult kin and greater reciprocity in these relationships among racial/ethnic minority persons than White persons, our results suggest that coresidence with adult children may be driven less by collective need and more by mutual benefits for Black and Hispanic parents than for White parents. Third, our analyses reveal that coresiding with an adult child has modest implications for parents' mental health that appear to hinge on the balance of relative costs and benefits. Further, these effects are similar for parents under 65 and those aged 65+. Hence, we found little support for the hypothesis that older parents gain more from living with their adult children than younger peers. Last, we found that the mental health correlates of living with adult children are most consistently negative for White parents, likely due to both causation and selection mechanisms. By contrast, living with certain types of adult children may be positively linked to Black and Hispanic parents' mental health. These findings support the perspective that multigenerational coresidence may be overall more normative and desired among Black and Hispanic parents than among White parents.
Broadly, our study highlights that intergenerational living arrangements are predicted by interactions among family members' needs, resources, and preferences, which vary across race/ethnicity. It also underscores that living arrangements are not static but reflect historical contexts and period effects. The COVID-19 pandemic has made it even more critical to understand how families respond to hardship and social change with strategies that include adapting their living arrangements. As parent–adult child households continue to become more common, attitudes toward sharing a home and its implications for well-being will also evolve.
Acknowledgments
This work was supported by funding from the National Institute on Aging (T32AG000243 and R03AG072235). The Health and Retirement Study is sponsored by the National Institute on Aging (grant NIA U01AG009740) and is conducted by the Institute for Social Research at the University of Michigan.
