Intergenerational Coresidential Patterns by Young Adult’s and Their Mother’s Mental Health and Substance Use

Abstract

Intergenerational coresidence is at a 30-year high. Studies find that economic, familial, and demographic factors shape the likelihood of this arrangement. We use NLSY79 and NLSY79YA data (2000–2014; N = 3,092) to examine how the mental health and substance use of both adult children and their mothers matter for coresidential biographies, estimating risks of moving out of and returning to their mothers’ households. Adult children who drink, smoke, or have more depressive symptoms, or whose mothers drink or smoke, are more likely to leave their mother’s household; adult children with more depressive symptoms and who smoke are more likely to return. Our findings show that children’s and mothers’ health are key determinants of coresidential patterns, suggesting that it is not just family arrangements that impact health but health that impacts family arrangements. As intergenerational coresidence increases, researchers should continue to look beyond economic, familial, and demographic determinants of coresidence to health dynamics.

In 2012, 36 percent of adults aged 18–31 years lived in their mother’s household—a 30-year high (Fry, 2013). Coresiding adult children include those who have never left their mother’s household as they aged into adulthood and “boomerang kids” who returned to their mother’s household after living independently (Dey & Pierret, 2014). Researchers link increased coresidence rates to recent economic downturns, delayed time to completing education, lower employment rates and wages, and delays or foregoing of long-term intimate relationships (Furstenberg, 2010; Houle & Warner, 2017; Stone, Berrington, & Falkingham, 2014). Covariates of exiting mothers’ households—generally seen as a key step in adult development (Furstenberg, 2010)—are often different than those for re-entering mothers’ households after leaving (e.g., boomeranging; Copp et al., 2017; South & Lei, 2014; Stone et al., 2014; Warner & Houle, 2018).

In this study, we hypothesize that both adult children’s and their mother’s substance use and mental health are key in explaining intergenerational coresidential patterns, both exiting from and returning to the mother’s household. Adult children’s mental health and substance use have shifted in recent decades; among young adults, depressive symptoms and amounts of alcohol consumed have increased, and smoking has decreased (Center for Disease Control, 2018; Gaydosh et al., 2019; Kerr et al. 2009; White et al., 2006). Similar patterns around substance use and mental health are seen today among midlife women—the mothers of the current cohort of young adults—compared to midlife women in previous years (Abram & Mehta 2019; Han et al., 2017). The importance of adult children’s mental health and substance use is likely interdependent with their mother’s health and substance use, making it important to consider mothers alongside their adult children. We estimate differences in adult children moving out of the mother’s home and returning, beginning at age 18, by alcohol use, smoking behavior, and depressive symptoms using NLSY79 and NLSY79YA data from 2000 to 2014 (N=3,092).

Background

Intergenerational coresidence in young adulthood is driven by two related shifts: delaying moving out of the mother’s household and returning after departing (Dey & Pierret, 2014). Among those born between 1980 and 1984, 90 percent left their parents’ household by age 27, but 55 percent moved back at least once after initially moving out (Dey & Pierret, 2014). Moving out is viewed as a normative transition to adulthood alongside completing school, beginning a career, and getting married (Furstenberg, 2010). Yet, returning to the mother’s household—despite being an increasingly common experience for young adults—is not typically viewed as a normative life transition and is often viewed as a negative development (Warner & Houle, 2018). Caputo (2019) found that after young adults moved back to their mother’s household, they experienced worse mental health, and she postulated that moving back may be distressing because it represents “a loss of a valued marker of adulthood that has already been achieved” (p. 16).

Previous studies identify clear predictors for exiting and returning (South & Lei, 2014). People with full-time employment and women are more likely to leave their mother’s household, but studies generally do not find employment or gender differences in who returns (Dey & Pierret, 2014; Warner & Houle, 2018). Those who have more education and who are non-Hispanic White are more likely to move out and move back than those with less education and those who are Black and/or Hispanic (Dey & Pierret, 2014). Notably, fewer studies examine the role of mothers’ characteristics in shaping coresidential biographies, although, given mothers’ primacy in this intergenerational tie, mothers’ characteristics likely have some impact. Previous research on mothers largely focuses on socioeconomic variables and finds children with more educated mothers move out at higher rates but children with mothers with greater incomes return home at higher rates (Dey & Pierret, 2014; South & Lei, 2014). In one of the few studies to consider the health of the mother, South and Lei (2014) find that young adults are more likely to return when their mother is in poor physical health, but their mother’s health does not impact moving out. We use these past studies, explicated further below, to build hypotheses. We summarize each of these hypotheses in Table 1.

Table 1:

Summary of Hypotheses

Hypothesis	Variables of Interest	Expectation Regarding Leaving Mother’s Household	Expectation Regarding Returning to Mother’s Household
Care Need	Adult child’s substance use and depressive symptoms	Negatively associated with leaving	Positively associated with returning
Social Control and Stigma	Adult child’s substance use and depressive symptoms	Positively associated with leaving	Negatively associated with returning
Limited Opportunity	Adult child’s substance use and depressive symptoms	Positively associated with leaving	Positively associated with returning
Maternal Conflict and Lack of Support	Mother’s substance use and depressive symptoms	Positively associated with leaving	Negatively associated with returning
Caregiving	Mother’s substance use and depressive symptoms	Negatively associated with leaving	Positively associated with returning
Interactional	Adult child’s and mother’s substance use and depressive symptoms	Extent to which adult children’s substance use and depressive symptoms associated with leaving or returning partially dependent on their mother’s substance use and depressive symptoms

Children’s Mental Health, Substance Use, and Coresidence

We focus on young adult children’s and their mother’s mental health and substance use in both leaving and returning biographies. Depressive symptoms and substance use are understood as respectively “internalizing” and “externalizing” behaviors that may shape family dynamics and patterns, especially during transitional periods such as aging from adolescence to adulthood (Rosenfield et al., 2005). Some past studies have addressed the role of depressive symptoms and alcohol use, finding that children’s depressive symptoms are associated with more coresidential instability (Copp et al., 2017; Sandberg-Thoma et al., 2015) and that children’s problem drinking is associated with returns to the mother’s households (Sandberg-Thoma et al., 2015). Yet, we theorize that alcohol use may matter more generally beyond just problem drinking and that both alcohol use and depression should be considered alongside other indicators of substance use, like smoking. The impact of substance use and mental health on coresidential patterns may be particularly salient among the current cohort of young adults given the prevalence of depressive symptoms, smoking, and alcohol as well as the shifting cultural meanings for each of these health areas. For example, smoking rates have plummeted in recent decades but there has been an increased stratification in who smokes—especially related to education—coupled with a greater stigma around smoking (Center for Disease Control, 2018; Kelly et al., 2018). We suggest that children’s mental health and substance use play a key role in adult children’s coresidence via three divergent pathways, described below.

Care need hypothesis.

The care need hypothesis suggests that adult children’s depressive symptoms, smoking, and moderate or heavy drinking would be positively associated with staying in and returning to the maternal home. In this hypothesis, children with markers of poor mental health and risky substance use may remain in their mother’s home for a longer period of time and return after leaving because of explicit needs for emotional, instrumental, and financial support from their mother. For example, it may be that individuals with poor mental health and moderate or heavy alcohol use in particular require more intensive instrumental support—such as transportation and help with everyday tasks—and emotional caregiving, and therefore stay or return home to receive this support (Smits, Van Gaalen, & Mulder, 2010).

Social control and stigma hypothesis.

The social control and stigma hypothesis suggests that adult children’s depressive symptoms, smoking, and moderate or heavy drinking would be positively associated with leaving and negatively associated with returning to the maternal home. In adolescence, mothers are the primary regulators of children’s risky health behaviors—especially substance use (Avenevoli & Merikangas, 2003). While these regulatory efforts may positively impact a minor child’s health behavior, once the child enters adulthood they may resent health regulation and desire independence (Reczek et al., 2014; Reczek & Zhang, 2016). Adult children with risky or stigmatized substance use such as heavy drinking or smoking may feel judged by their mother, experience discrimination and punishment, and feel pushed out of their mother’s home.

Limited opportunity hypothesis.

The limited opportunity hypothesis suggests that adult children’s depressive symptoms, smoking, and moderate or heavy drinking would be associated with more housing instability, characterized by both exiting and returning to their mother’s household. Due to the stigmatization of depression, heavy alcohol use, and smoking (Kelly et al., 2018; Link & Phelan, 2014), people with these health statuses may find that their mother does not allow them to remain at home. But at the same time, they may be limited in their ability to gain economic resources to establish and keep their own household. This may be due to limited opportunities to develop non-familial social relationships that encourage the pooling of resources (e.g., roommates, intimate partners). Similarly, adults with moderate or heavy alcohol use, depressive symptoms, and who smoke may have difficulty maintaining positive relationships with their mother as well as difficulty entering and maintaining positive relationships with friends and intimate partners (Reczek et al., 2016). Without financial stability or other support from steady employment or significant others, these young adults may have few options regarding housing, leading to instability and earlier transitions out of their mother’s home as well as back into her home at a later date.

Mother’s Depression and Substance Use and Coresidence

Mother’s mental health and substance use may also be linked to coresidential patterns. We theorize this may occur in two different ways.

Maternal conflict and lack of support hypothesis.

The maternal conflict and lack of support hypothesis suggests that mother’s substance use and depression would be positively associated with children leaving and negatively associated with returning to the maternal home. Mothers often furnish adult children’s independence through providing financial and emotional support (West et al., 2017). Yet, mothers with poor mental health and substance use behaviors are likely to be less financially and emotionally situated to provide aid to their adult children, as their health may inhibit their earning and increase their expenses (e.g., medical debt; van Rijn et al., 2014). These mothers may also be less able to provide instrumental or emotional care to children in need. Mothers’ substance use may further reflect negligence and increased conflict (Dube et al., 2001), which may push children out of the home.

Caregiving hypothesis.

The caregiving hypothesis suggests that mother’s substance use and depression would be positively associated with children staying in and returning to the maternal home. In Seltzer and Friedman’s (2013) study of midlife adults and their older mothers and South and Lei’s (2014) study of young adults and their mothers, mothers’ poor physical health and functioning prompted adult children to return home, likely for adult children to provide care to their mother. We expect these same processes may influence children’s coresidence during young adulthood, although it may be that other measures of mothers’ health beyond physical health and functioning (e.g., mental health, substance use) may be more salient in impacting coresidential patterns at the early adulthood life stage. Adult children whose mother has poor mental health or greater substance use may be less likely to leave their mother’s household and more likely to return so that they can provide instrumental, emotional, and financial care for their mother. In the case of moderate or heavy alcohol use, smoking, and depressive symptoms, children may witness the effects of these health behaviors and attempt to intervene to improve their mother’s well-being by staying in the maternal home longer or returning home after being away.

Mother’s and Children’s Depression and Substance Use and Coresidence

Interactional hypothesis.

We hypothesize that the mother’s health and children’s substance use and mental health act interdependently to shape children’s coresidence patterns, such that the impact of children’s substance use and mental health on leaving or remaining in their mother’s household is dependent in part on their mother’s substance use and mental health. This is the interactional hypothesis. We expect these interactions to be impactful because research shows that mothers’ and children’s mental health and substance use are co-varying, wherein children’s mental health and substance use are positively associated with their mother’s mental health and substance use (Avenevoli & Merikangas, 2003; Capaldi et al., 2016; Mason et al., 2017). Similarity in poor mental health or substance use may exacerbate conflict and reduce coresidence due to the increased likelihood of individual and interpersonal strain. Alternatively, similarity in these behaviors or outcomes could contribute to children and their mothers living together due to decreased stigma—especially around substance use and depressive symptoms-- and improved compatibility. For example, an adult child who smoke may be allowed to live at home if their mother also smokes but asked to live elsewhere if the mother does not smoke.

Methods

In considering the associations between adult children’s and their mother’s mental health and substance use and children’s exit and re-entry into the maternal household, we analyzed young adults ages 18–33 years from the National Longitudinal Study of Youth-79 Young Adults (NLSY79YA) from 2000 to 2014, along with data from their mother from the NLSY79. The NLSY79 is a household probability sample, and in 1979, the original sample included 12,686 men and women born in the United States between 1957 and 1964 (Rothstein, Carr, and Cooksey, 2018). In the original interview in 1979, respondents were 14 to 22 years. Respondents were interviewed yearly from 1979 to 1986, and then on a biennial basis. The NLSY79YA dataset—the focus of our analysis—consists of all biological children born to the 6,283 women in the NLSY79. Mothers provided survey information about their children beginning in 1986, and children were interviewed themselves beginning in 1994 when they are at least 15 years old, with these interviews occurring biennially. A total of 6,011 respondents 15 years and older have been interviewed.

Dependent Variables

For this study, our sample size was 3,092. We restricted our sample to respondents who were in the study for at least four waves (i.e., eight years), beginning at age 18. This means we excluded respondents younger than 26 in 2014 and respondents who turned 18 before 2000. These inclusion criteria allowed us to follow respondent’s movements into and out of their mother’s household for at least eight years. We also excluded respondents who were not living in their mother’s home at age 17 (N=249). We estimated the risk of the first occurrence of two outcomes: adult child leaving their mother’s home and adult child returning to their mother’s home. At every wave, respondents were asked where they were currently residing, and whether for at least a one month period during the last two years they resided elsewhere. This allowed us to capture movement between waves. We conducted robustness checks where we fit models using only movement from wave to wave, ignoring movement between waves that did not result in a consistent new living arrangement across waves, but results were statistically and substantively similar. In these supplementary models, 165 fewer respondents exited their mother’s home during the study period, and 254 fewer respondents re-entered the mother’s home.

First, to create our exit variable, we constructed a measure of approximate age first left the mother’s household as an adult in years. Next, to construct our re-entry variable, using the subsample of respondents who had continuing data after leaving their mother’s home after age 18 (N=1,849), we constructed a measure of the number of years living away before returned home (if returned). Living in mother’s household included living with mother only or living with mother and father. Only 48 respondents lived with their father only and never with their mother. Supplementary analysis found coding adult children living with fathers only as coresiding with mothers or dropping them from the analysis did not change the results.

Independent Variables

In our main analysis, we measured depressive symptoms, using the NLSY’s 7-item CES-D scale. We treated depressive symptoms as a continuous measure, from 0–21. In supplementary analysis, we dichotomized this measure, using the cutoff score of 8, the threshold for depression based on the 7-item scale (Levine, 2013), and these results are discussed below. As a key limitation, the CES-D questions were not asked to all respondents in all waves. In the 2000 and 2002 surveys, the CES-D questions were limited to respondents not interviewed in the previous survey round. Additionally, in 2012, respondents between the ages of 25 and 29 were not asked the CES-D questions, and in 2014, only respondents between the ages of 14 and 24 years were asked. Thus, for respondents missing CES-D data, we imputed their most recent CES-D score. CES-D scores were highly correlated across waves, especially after the age of 25, but this is still a significant limitation of this study design.

We divided drinking behavior into three categories: drinks alcoholic beverages less than once a month (non-drinking), drinks at least once a month but less than seven beverages a week (occasional drinking), and drinks seven or more beverages a week (moderate to heavy drinking). In supplementary analysis, we separated moderate drinking (7–14 beverages a week) from heavy drinking (more than 14 beverages a week) but these categories were statistically similar. Our smoking measure was three categories: current non-smoker, current occasional smoker, and current daily smoker. Current occasional smoker was defined as smoked any cigarettes during the last 30 days, and current daily smoker was defined as smoked every day during the last 30 days.

For mothers’ health, we used the same measures and categories, as reported by the mother in the NLSY79 survey. Depressive symptoms were measured in the 40s and 50s health modules by a CES-D index, asked when mothers turned 40 and 50, and smoking behavior was not asked in 2000–2006 waves. We imputed depressive symptoms and smoking to the nearest wave. As with adult children’s CES-D score, these imputations for mothers’ CES-D and smoking are key limitations of our study, and means that these variable should be not interpreted as dynamic to the same degree as other variables in this study. Child’s and mother’s substance use and mental health variables were time-varying.

Other Covariates

In all models we controlled for the adult child’s gender (woman or man), race/ethnicity (non-Hispanic White, Black, Hispanic, or identified as another race or ethnicity), birth year of child, and mother’s age, as each has been found to be associated with coresidential patterns (South & Lei, 2014). Substance use, depression, and coresidential patterns are each strongly correlated with socioeconomic status, marital status, and parental status (Goldscheider et al., 2014; South & Lei, 2014; Warner & Houle, 2018), so, in our final model, we introduced controls for the child’s educational attainment by age 24 and mother’s educational attainment (less than high school or high school only, some college or Associate’s degree, or college graduate), child’s and mother’s employment status (unemployed, part-time employed, or full-time employed), child’s current school enrollment (enrolled currently in school or not enrolled currently in school), child’s parental status (no children or at least one child), and child’s and mother’s relationship status (single, cohabiting, or married). We considered child’s education attainment by age 24 because current educational attainment was highly correlated with age in this young adult sample which began at age 18. Children’s employment status, school enrollment, parental status, and relationship status and mother’s age, employment status, and relationship status were all time-varying in all models.

Analysis

We estimated hazard ratios and 95 percent confidence intervals with Cox’s proportional hazard regression models using Stata. Cox’s proportional hazard models estimate the incremental risk of an event occurring, with a respondent’s hazard a multiplicative replica of the baseline hazard based on the respondent’s own covariates (Cox, 1972). This type of model is advantageous when the exact parametric form of the relationship of the hazard with time is unknown or not of primary interest (Allison, 1984), as was the case in our analysis. These models make no assumptions about the baseline hazard except that it is the same for all respondents who are at risk at any particular time (Vuchinich, Teachman, & Crosby, 1991). The hazard is defined as

$$h_i(t)=\text{Pr}(T_i=t|T_i\ge t)$$

representing the probability that an event T (leaving mother’s household in first set of models and returning to mother’s household in second set of models) occurs at time t, given it did not occur in prior observations (t – 1). Age was our analytic unit of time in the models for moving out of mother’s home (from age 18 until move out of mother’s home or a censoring event), and time since moving out of mother’s home was the analytic unit of time for our models for returning to mother’s home (from age moved out of mother’s home until moved back in or a censoring event). We included respondents in models as at risk when they were at least 18 in the first set of models, and respondents in models were at risk when they first exited their mother’s household in the second set of models. Thus adult children who remained in their mother’s household throughout the study period were included in the first set of models, and they were not included in our second set of models. Additional analysis based on model fit found that the semiparametric models produced more efficient estimates than parametric hazard models, although parametric models found statistically similar results. Using Schoenfeld residuals, we found that the proportional-hazards assumption was not violated (Schoenfeld, 1982). We handled ties with the Breslow method.

As robustness checks, we also fit models including respondents who have been in the sample for at least ten years (ages 18–28; N=1,677) rather than eight years, models including respondents who have been in the sample at least from ages 22–28 and were residing at their mother’s home at age 21 (N=1,763), and models including respondents who had been in the sample at least from ages 20–26 and were residing at their mother’s home at age 19 (N=1,924). In general, all sets of results were statistically similar to the models we present, despite the smaller sample sizes.

Our first models included the baseline variables (gender, race, birth year, mothers’ age) and children’s and mothers’ depressive symptoms, alcohol use, and smoking. In supplementary analysis, we tested children’s health variables separately from mothers’ health variables, but results were statistically similar and thus not presented. In our second models, we included all sociodemographic variables as well as all the health and baseline variables. In our final models, to test whether the impact of adult children’s health variables on coresidence patterns were conditioned on mother’s health variables, we tested for interactions between mother’s health variables and children’s health variables.

Because our health covariates were time-varying, we tested their interaction with time, but we did not find evidence that health effects changed with time. We do not include these interactions in our presented models, so coefficients represent average effects of a given variable across the observations (Allison, 1984). We also test for interactions between the health variables and the adult child’s gender, race/ethnicity, and educational attainment, based on previous research that indicates that pathways in and out of the mother’s household are conditioned on gender, race/ethnicity, and educational attainment, but these interactions were not significant (South & Lei, 2014; Stone et al., 2014). To deal with missing data, we imputed respondents missing data on depressive symptoms from the most recent wave (as discussed above) and dropped respondents missing on smoking (n=7). Because 286 respondents were missing data on drinking, with most of these respondents born in later years, we created a missing flag for drinking to retain these respondents. We created missing flags for the remaining respondents missing information on other non-health variables, thus retaining these respondents.

Results

Descriptive Statistics

In our sample, 2,814 respondents moved out of their mother’s home during the study period (91.01% of the analytic sample), and the mean age for moving out was 21.08 years. Of those respondents who moved out of their mother’s home, 865 moved back in during the study period. The mean age for moving back in was 23.35 years, and these young adults who moved out and then back during the study period composed 27.98 percent of the total sample. At age 21, only about one third of the sample was living away from their mother, but by age 27, three fourths of the sample was currently living outside their mother’s home.

Table 2 shows characteristics for this sample by coresidence patterns at ages 24 and 25. All differences discussed are significant by a two-tailed test (p < .05). Among adult children who smoked daily, fewer lived in their mother’s home. Adult children’s depressive symptoms and alcohol use were not associated with living in mother’s home. Regarding mother’s substance use and mental health variables, adult children whose mother was a non-drinker more often lived in their mother’s household, and adult children whose mother drank heavily more often lived away from their mother. Other descriptive statistics related to adult children’s and mothers’ sociodemographic and economic variables (e.g., parenthood status, gender, race/ethnicity) were generally in line with other studies (Goldscheider et al., 2014; South & Lei, 2014; Warner & Houle, 2018).

Table 2:

Descriptive Statistics (Means/Proportions and Standard Deviations) at age 24/25 (NLSY79YA and NLSY79 2000–2014; N=3,092)

	Living in Mother’s Home (N=1,052)	Not Living in Mother’s Home (N=2,040)
Adult Child’s Health
CES-D	4.34 (3.80)	4.31 (3.73)
Drinking Behavior (Non-drinker)	0.38	0.35
Occasional Drinker	0.28	0.28
Moderate to Heavy Drinker	0.34	0.37
Smoking Behavior (Not a Current Smoker)	0.64	0.60*
Occasional Smoker	0.14	0.14
Daily Smoker	0.22	0.26*
Mother’s Health
CES-D	3.82 (4.47)	4.01 (4.61)
Drinking Behavior (Non-drinker)	0.69	0.61***
Occasional Drinker	0.17	0.18
Moderate to Heavy Drinker	0.14	0.20***
Smoking Behavior (Not a Current Smoker)	0.77	0.75
Occasional Smoker	0.05	0.05
Daily Smoker	0.18	0.20
Adult Child’s and Mother’s Other Sociodemographics
Child’s Birth Year	1985.96 (2.68)	1985.46*** (2.82)
Mother’s Age	50.99 (3.24)	49.37*** (3.36)
Woman	0.48	0.55***
White	0.47	0.59***
Black	0.37	0.30***
Hispanic	0.15	0.09***
Other	0.01	0.02
Child’s Educational Attainment by Age 24 (Less than High School/High School Only)	0.37	0.34
Some College or Associate	0.48	0.47
College Graduate	0.15	0.18*
Child’s Employment Status (Unemployed)	0.11	0.08*
Part-time Employed	0.19	0 14***
Full-time Employed	0.71	0.78***
Child Currently Enrolled in School	0.23	0.21
Mother’s Educational Attainment (Less than High School/High School Only)	0.59	0.58
Some College or Associate	0.28	0.27
College Graduate	0.13	0.16
Mother’s Employment Status (Unemployed)	0.23	0.23
Part-time Employed	0.16	0.15
Full-time employed	0.61	0.62
Child Is a Parent	0.29	0.43***
Child (Single)	0.92	0.49***
Cohabiting	0.04	0.28***
Married	0.04	0.23***
Mother (Single)	0.39	0.36
Cohabiting	0.04	0.07**
Married	0.57	0.57

^*p < .05.

^**p < .01.

^***p < .001.

Moving out of Mother’s House

The descriptive statistics presented above only focus on a narrow age range, do not capture movement into or out of the mother’s home including movement between waves, do not account for changes in the predictor variables (e.g., change in employment status, change in alcohol use), and do not adjust for other key variables of interest. We now present the regression analysis, testing our hypotheses (summarized in Table 1). When discussing percentage likelihoods of moving out of the mother’s home, these were conditional on still being in their mother’s home the year before. The baseline models for risk of moving out of the mother’s home are presented in Model 1 in Table 3.

Table 3:

Hazard Ratios for Exit from Mother’s Home, NLSY79 & NLSY79-YA 2000–2014 (N = 3,092)

	Model 1		Model 2
	HR	95% CI	HR	95% CI
Adult Child’s Health
CES-D	1.02**	1.01–1.03	1.01**	1.00–1.03
Drinking Behavior (Non-drinker)
Occasional Drinker	1.13*	1.02–1.25	1.14*	1.03–1.26
Moderate to Heavy Drinker	1.18**	1.06–1.30	1 24***	1.12–1.37
Smoking Behavior (Not a Current Smoker)
Occasional Smoker	1.09	0.97–1.23	1.08	0.95–1.21
Daily Smoker	1.23***	1.11–1.36	1.12*	1.01–1.24
Mother’s Health
CES-D	1.01	1.00–1.01	1.00	1.00–1.01
Drinking Behavior (Non-drinker)
Occasional Drinker	1.07	0.97–1.18	1.03	0.93–1.14
Moderate to Heavy Drinker	1.11*	1.01–1.23	1.09+	0.99–1.20
Smoking Behavior (Not a Current Smoker)
Occasional Smoker	0.91	0.75–1.10	0.95	0.78–1.15
Daily Smoker	1.11*	1.01–1.22	1.11*	1.01–1.22
Child and Mother Demographic, Socioeconomic, and Family Variables
Child’s Birth Year	1.01	0.99–1.03	1.01	0.99–1.03
Mother’s Age	0 97***	0.95–0.98	0.97**	0.95–0.99
Woman	1.15***	1.06–1.24	1.06	0.98–1.14
Child’s Race (White)
Black	0.93	0.85–1.01	0.99	0.90–1.08
Hispanic	0.82**	0.72–0.93	0.84**	0.74–0.96
Other	1.15	0.84–1.58	1.12	0.81–1.53
Child’s Educational Attainment (Less than High School/High School Only)
Some College or Associate			1.12*	1.02–1.24
College Graduate			1.39***	1.21–1.59
Child’s Employment Status (Unemployed)
Part-time Employed			0.93	0.82–1.07
Full-time Employed			1.12+	0.99–1.27
Child Currently Enrolled in School			0.86**	0.78–0.95
Mother’s Educational Attainment (Less than High School/High School Only)
Some College or Associate			1.12*	1.03–1.22
College Graduate			1.19**	1.06–1.34
Mother’s Employment (Unemployed)
Part-time Employed			0.96	0.84–1.09
Full-time Employed			0.93	0.85–1.02
Child Is a Parent			1.13*	1.02–1.24
Child’s Relationship Status (Single)
Cohabiting			2.33***	2.09–2.59
Married			2.30***	1.97–2.66
Mother’s Relationship Status (Single)
Cohabiting			1.01	0.85–1.19
Married			0.87**	0.79–0.95
Log Likelihood	−18906.25		−18661.57

⁺p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

CI = confidence interval; HR = hazard ratio

The results in Table 3 provide support for the idea that both children’s and mother’s substance use and children’s mental health matters for whether and when children move out of their mother’s house. Every additional point on the CES-D score reported by children increased the likelihood of moving out of their mother’s home by 2 percent. This is a substantively modest but statistically significant result. If considering CES-D score as a dichotomous variable, we find that meeting the threshold of a score of 8 or more (Levine, 2013) is associated with a 14 percent greater likelihood of moving out (p < .01). Regarding drinking, all levels of adult children’s drinking were associated with increased likelihood of moving out compared to children drinking less than one alcoholic beverage a month (i.e., non-drinker). For smoking, adult children who smoked daily were 23 percent more likely to move out than adult children who did not currently smoke, but adult children who smoked occasionally had a statistically similar likelihood of moving out as those who did not smoke. Each of these results provides partial support for the social control and stigma hypothesis and the limited opportunity hypothesis, which both predict higher rates of moving out when adult children have depressive symptoms, smoking, or heavy drinking. Regarding mother’s mental health and substance use, adult children whose mother smoked daily were 11 percent more likely to move out than adult children whose mothers did not smoke at all. Additionally, adult children whose mother drank 7 or more drinks a week were 11 percent more likely to move out than those whose mother drank less than one drink a month. Both of these results provide partial support for the maternal conflict and lack of support hypothesis.

In Model 2, we controlled for our key socioeconomic and family variables to test for possible confounders. With these controls added, there was little change between the health variables and the hazard of moving out of mother’s house with two exceptions: the increased hazard associated with adult children’s smoking was reduced and the increased hazard associated with mother’s drinking was reduced to marginal significance (p < .10).

Moving Back into Mother’s House

Using the subsample of the respondents who moved out and were in the study for at least one additional wave, we fit models to test the hazard of moving back into their mother’s house (Table 4). Adult children’s depressive symptoms and smoking, but not alcohol use, were associated with the greater likelihood of moving back into their mother’s home, in Model 1. Similar to moving out, each additional point on the CES-D score reported by children increased the risk of the adult child moving back into their mother’s home by 2 percent. Considering this as a dichotomous variable, those with a CES-D score of 8 or greater had a 16 percent higher likelihood of moving back in with their mother, although this was only marginally significant (p < .10). Adult children who smoked occasionally had a 40 percent higher likelihood of moving back in with their mothers than adult children who did not smoke, and 23 percent higher if smoked daily. However, once controlling for socioeconomic and demographic variables (Model 2), depressive symptoms were no longer significantly associated with hazard of moving out, and daily smoking was only marginally significantly associated with moving out. Notably, however, the confidence intervals for the CES-D coefficient in Model 2 in Table 4 were the same as in Model 2 in Table 4, suggesting that this finding may just reflect differences in sample sizes in the two models. In other words, the larger sample for estimating the hazard of leaving mothers’ home may provide more robust estimates than the smaller sample for returning to mother’s home, and this limitation should be taken into account when comparing these results. At the same time, examining this as a dichotomous variable, those with a CES-D score of 8 or greater do not have a statistically different risk of returning to their mother’s home than those with a lower CES-D score (p = .194) in the full model. Considered alongside the moving out results, these findings indicate greater general support for the limited opportunity hypothesis. Mother’s depressive symptoms, drinking, and daily smoking were not related to the hazard of adult children moving back into the mother’s home in Model 1 or Model 2.

Table 4:

Hazard Ratios for Re-entry Into Mother’s Home, NLSY79 & NLSY79-YA 2000–2014 (N = 1,849)

	Model 1		Model 2
	HR	95% CI	HR	95% CI
Adult Child’s Health
CES-D	1.02**	1.01–1.04	1.01	1.00–1.03
Drinking Behavior (Non-drinker)
Occasional Drinker	0.91	0.76–1.09	0.93	0.78–1.11
Moderate to Heavy Drinker	0.92	0.77–1.09	0.87	0.73–1.04
Smoking Behavior (Not a Current Smoker)
Occasional Smoker	1.40**	1.14–1.71	1.34**	1.09–1.64
Daily Smoker	1.23*	1.04–1.46	1.18+	0.92–1.44
Mother’s Health
CES-D	1.00	0.98–1.01	1.00	0.98–1.01
Drinking Behavior (Non-drinker)
Occasional Drinker	1.02	0.85–1.23	1.07	0.89–1.28
Moderate to Heavy Drinker	0.95	0.80–1.14	0.96	0.80–1.15
Smoking Behavior (Not a Current Smoker)
Occasional Smoker	0.94	0.69–1.28	0.89	0.65–1.21
Daily Smoker	0.94	0.79–1.11	0.90	0.76–1.07
Child and Mother Demographic, Socioeconomic, and Family Variables
Child’s Birth Year	1.02	0.99–1.05	1.01	0.98–1.04
Mother’s Age	0.98*	0.95–1.00	1.00	0.97–1.02
Woman	1.03	0.90–1.19	1.11	0.96–1.29
Child’s Race (White)
Black	1.35***	1.16–1.58	1.12	0.94–1.32
Hispanic	1.53***	1.24–1.90	1.43**	1.15–1.77
Other	1.18	0.71–1.94	1.12	0.67–1.86
Child’s Educational Attainment (Less than High School/High School Only)
Some College or Associate			0.83*	0.71–0.97
College Graduate			0.64**	0.48–0.85
Child’s Employment Status (Unemployed)
Part-time Employed			0.97	0.76–1.24
Full-time Employed			0.90	0.73–1.11
Child Currently Enrolled in School			1.03	0.84–1.28
Mother’s Educational Attainment (Less than High School/High School Only)
Some College or Associate			0.98	0.83–1.15
College Graduate			0.90	0.70–1.17
Mother’s Employment (Unemployed)
Part-time Employed			0.83	0.65–1.06
Full-time Employed			0.93	0.79–1.09
Child Is a Parent			1.02	0.88–1.20
Child’s Relationship Status (Single)
Cohabiting			0 43***	0.35–0.53
Married			0.38***	0.30–0.48
Mother’s Relationship Status
Cohabiting			0.94	0.71–1.24
Married			1.09	0.93–1.27
Log Likelihood	−6156.10		−6074.1

⁺p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

CI = confidence interval; HR = hazard ratio

Interactions between Adult Child’s and Mother’s Health

Table 5 shows the hazard of exiting the mother’s house (Model 1) and moving back into the mother’s house (Model 2), including interactions between adult child’s and mother’s mental health and substance use. As shown in Model 1, the interaction between mother’s and child’s daily smoking is significantly associated with moving out of the mother’s house. We plot the hazard functions for four groups—both adult child and mother smoke daily, neither adult child nor mother smoke at all, adult child smokes daily but mother does not smoke at all, and mother smokes daily but adult child does not smoke at all—in Figure 1. When the adult child smokes daily and their mother does not, the hazard of moving out is less than if both the child and the mother smoke daily. Similarly, if the adult child does not smoke but the mother smokes daily, the hazard of moving out is more than if both the child and the mother does not smoke. Adult children who smoke daily but whose mother does not smoke have a higher likelihood of exiting than adult children who do not smoke but their mother does. These findings provide partial support to the interactional hypothesis. But the interactions between mother’s and child’s smoking were not significant in re-entry models, and alcohol and mental health interactions were not significant in any models.

Table 5:

Hazard Ratios for Exit from and Re-Entry to Mother’s Home with Interactions between Adult Children’s and Mother’s Mental Health and Substance Use, NLSY79 & NLSY79-YA 2000–2014 (N = 3,092)

	Exit		Re-entry
	N = 3,092		N = 1,849
	HR	95% CI	HR	95% CI
Adult Child’s Health
CES-D	1.02*	1.00–1.03	1.02+	1.00–1.04
Drinking Behavior (Non-drinker)
Occasional Drinker	1.10+	0.98–1.23	0.95	0.79–1.15
Moderate to Heavy Drinker	1.18**	1.06–1.32	0.89	0.74–1.08
Smoking Behavior (Not a Current Smoker)
Occasional Smoker	1.09	0.97–1.23	1 45***	1.18–1.78
Daily Smoker	1 31***	1.16–1.47	1.17	0.96–1.43
Mother’s Health
CES-D	1.01	0.99–1.02	0.99	0.97–1.02
Drinking Behavior (Non-drinker)
Occasional Drinker	1.04	0.93–1.17	1.07	0.87–1.32
Moderate to Heavy Drinker	1.12+	1.00–1.26	0.91	0.73–1.14
Smoking Behavior (Not a Current Smoker)
Occasional Smoker	0.90	0.74–1.11	1.07	0.76–1.49
Daily Smoker	2.27**	1.42–3.63	0.91	0.73–1.14
Interactions
Adult Child CES-D * Mother CES-D	1.00	1.00–1.00	1.00	1.00–1.00
Adult Child Occasional Drinker * Mother Occasional Drinker	1.11	0.89–1.39	0.81	0.54–1.23
Adult Child Moderate to Heavy Drinker * Mother Moderate to Heavy Drinker	0.97	0.80–1.18	1.14	0.80–1.62
Adult Child Occasional Smoker * Mother Occasional Smoker	0.99	0.55–1.75	0.51	0.21–1.24
Adult Child Daily Smoker * Mother Daily Smoker	0.82*	0.67–0.99	1.19	0.85–1.66
Log Likelihood	−18902.93		−6153.42

⁺p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

CI = confidence interval; HR = hazard ratio

All models control for child’s birth year, mother’s age, child’s gender, and child’s race/ethnicity.

Figure 1:

Smoothed Hazard Function for Adult Child Exiting Mother’s Household by Adult Child and Mother’s Smoking Behavior (NLSY79 & NLSY79-YA 2000–2014; N = 2,326)

Discussion

While most studies of family and health focus on how family arrangements shape health outcomes (Umberson, Crosnoe, & Reczek, 2010), we consider how health itself may shape family arrangements—specifically intergenerational coresidence. We examine how both mothers’ and children’s depressive symptoms, alcohol use, and smoking behaviors play a role in children leaving and/or returning to the maternal home. Overall, we find that adult children’s mental health and smoking matters for children moving out and back into the maternal household, while mother’s substance use and adult children’s drinking is linked to adult children leaving the maternal household. Below, we further articulate the study’s main findings as they relate to our hypotheses shown in Table 1, provide theoretical explanations of these patterns, and demonstrate the potential implications for these findings.

Children’s Health and Coresidence Patterns

First, our findings regarding smoking and depressive symptoms generally support the limited opportunity hypothesis, in that children’s depressive symptoms and smoking are related to both leaving their mother’s home and returning to her home after departure. Smoking and depressive symptoms appear to contribute to household instability by pushing adult children out of their mother’s home, perhaps because their health behaviors or mental health is unacceptable to their mother or drives conflict. Additionally, perhaps due to broader stigma regarding smoking and depression (Kelly et al., 2018; Link & Phelan, 2014), these health factors may also push adult children out of alternative housing with cohabiting partners, roommates, or even property managers who prefer nonsmokers or people without depressive symptoms, further limiting young adults’ ability to live independently. Regarding smoking in particular, other studies find that smokers face stigma as smoking rates are at record low levels among young adults (Center for Disease Control, 2018; Kelly et al., 2018). This increased stigma might lead to discrimination in the housing market and thus be a driver of this boomeranging trend—both pushing smokers out of their mother’s household and driving them back in. The limited opportunity hypothesis highlights that economic issues, health behaviors, mental health, and boomeranging in and out of a mother’s households in young adulthood are all inextricably linked, and thus these correlates should continue to be tested (Copp et al., 2017; Sandberg-Thoma et al., 2015). For example, future research should measure how levels of social and cultural capital as well as various forms of discrimination based on health status (e.g., employment discrimination, housing discrimination) mediate the associations between smoking and depression and moving in and out of the mother’s home.

Our findings regarding children’s alcohol use provide partial support for the social control and stigma hypothesis, as alcohol use is associated with both leaving and not returning as predicted by this hypothesis. Adult children who drink more alcohol may experience conflict and regulation around this behavior, possibly due to stigmatization by their mothers, leading them to desire increased independence and to leave the mother’s household earlier and also to not return (Reczek & Zhang, 2016). As an alternative explanation, adult children’s alcohol use may reflect the stress they experience at home, driving them out of the maternal home but not (as suggested by the limited opportunity hypothesis) pushing them back to their mother’s house. To better test this in future studies, we would need to include measures of stress, conflict, and stigma within the maternal household. Notably, there does not seem to be a dose response for alcohol use, as we find that any drinking increases the likelihood of moving out of the mother’s household and remaining out. This indicates that the processes linking alcohol use to moving out are likely multifaceted and not only linked to problem drinking, reflecting that alcohol use is both a health behavior and a somewhat protective social behavior during young adulthood.

Our finding that alcohol use is not related to moving back into the maternal home is in contrast to Sandberg-Thoma and colleagues (2015), who find alcohol problems are not associated with leaving the mother’s home but are associated with earlier returns. This difference in findings likely reflects different operationalizations of alcohol use. Sandberg-Thoma and colleagues use a dichotomous measure of problematic drinking (i.e., in past thirty days whether respondent drank five or more drinks in one sitting or had a drink prior to school or work), whereas our measure is not designed to capture problematic drinking but rather distinguish non-drinkers from occasional drinkers and from moderate to heavy drinkers. The differences in our results suggest alternative meanings and consequences of problematic drinking as compared to moderate to heavy drinking, perhaps especially within the broader young adult context. While problematic drinking may be stigmatized in young adult peer groups and interfere with work responsibilities—thus limiting a young adult’s ability to support one’s self and live independently— moderate drinking, as refleted in our measure, may be an asset in certain young adult contexts even while possibly pushing young adults from their mother’s home (Crosnoe, Kendig, & Benner, 2017). Future research should test this possibility.

Finally, although the care need hypothesis was not supported by our results, this is not to suggest that it is not an important process in coresidential patterns, but rather that our data may be too limited to test this hypothesis. It may be that young adults with higher levels of depressive symptoms are most likely to move back to the maternal home for financial and emotional support, especially if they exited earlier, before ready to live independently. This may also be true of adult children who smoke, as they may experience health issues not captured in the NLSY79YA data that make living on their own difficult (e.g., breathing issues, Jayes et al., 2016). Additionally, the increased cost of smoking may place a financial burden on adult children that limits being able to live independently. Future studies could test this by including measures of instrumental and emotional support needs and provisions, especially by mothers to their adult children by mental health and substance use.

Mother’s Health and Children’s Moving Out

In our analysis, the association of mother’s health with coresidence pattern is more limited compared to adult children’s health variables, only seen for smoking and drinking and only in relation to adult children moving out. Although our caregiving hypothesis suggests that mother’s substance use and depressive symptoms will encourage adult children to stay in the home (Seltzer & Friedman, 2013), we do not find support for this hypothesis. Instead, mother’s riskier health behaviors are associated with adult children leaving sooner, with the impact of mother’s smoking somewhat conditioned on the adult child’s own smoking behaviors. This partially supports the maternal conflict and lack of support hypothesis, which suggests that poor mother’s health may be compromising to adult children and encourage them to move out of the home. To the extent that adult children serve as a resource to mothers in midlife, especially those with poorer health (Seltzer & Friedman, 2013), these mothers with risky substance use behaviors may be especially vulnerable as they transition to later life.

Mother’s smoking moderates the impact of their adult child’s smoking on moving out, lending partial support of the interactional hypothesis. We specifically find that adult children’s smoking is more strongly associated with moving out if the mother also smokes. While we cannot fully explain this finding with our analyses, we suggests that mother’s and children’s smoking habits are likely related and may reflect general strain in both individuals’ lives; similar smoking patterns may also reflect interpersonal conflict between mothers and children that lead to moving out. The findings regarding interactions between mother’s and adult child’s smoking on the hazard of moving out further support our general assertion that studies considering how adult children’s health is related to coresidence patterns need to also consider mother’s health. Further, in discordant households, adult children’s smoking seems to matter more than mother’s smoking in shaping whether or not adult children exit the home, likely because children’s smoking reflects children’s strain in ways that will impact leaving or exiting and because mother’s ultimately control whether or not the adult child is allowed to live in their household.

Limitations

This study’s contributions to our understanding of children’s and mother’s health and coresidence patterns should be considered within the context of study limitations. We chose measures that were available for both mothers and children and therefore do not include more extensive measures of mother’s and child’s health, including illicit drug use, chronic conditions, sleep disorders, or anxiety. We also tested measures of self-rated health and functional impairments, but these were not significant and were not a focus of this study. However, in future analysis, additional measures of children’s and mother’s health may be instructive towards understanding intergenerational coresidential patterns. We were also unable to consider the role of father’s health in this dataset, but because fathers have higher rates of risky substance use behaviors than mothers (Kang, Sung, & Kim, 2010), father’s health is likely influential and should be considered in future studies. Future studies should explore the ways in which economic resources, education, and intimate relationships—and transitions into and out of these statuses—work together with mental health and substance use to shape coresidential patterns (Warner & Houle, 2018). Additionally, even though our analysis did not provide evidence that gender, race/ethnicity, or educational attainment serve as moderators in the association between health indicators and coresidence patterns, future studies should continue to examine the importance of these sociodemographic categories given that health is distributed unequally across them even in young adulthood (Gaydosh et al., 2019).

Conclusion

Increasing rates of mother-adult child coresidence are of key scholarly and policy interest, in part because changes in residential biographies have implications for both generations across the life course. Despite clear theoretical reasoning, previous work fails to explore the full range of health determinants of these coresidential forms. Our findings show that children’s substance use and depressive symptoms are key determinants of coresidential patterns. Our approach shows that adult children’s mental health, and both adult children’s and their mother’s substance use, matters for coresidence patterns — pushing children out and bringing them back in to their mother’s home during young adulthood. Perhaps because young adult children are often assumed to be healthy and the health consequences of substance use occurs later in the life course (e.g., lung cancer, liver disease), their health beyond disabilities or serious addictions is often overlooked in coresidential biographies (Wiemers et al., 2017). Yet the idea of young adults as generally healthy relies on a narrow definition of health, overlooking mental health and substance use. In fact, there is increasing evidence that today’s young adults and their mothers have worse health—and more health inequalities—across multiple outcomes than the most recent previous generations (Abram & Mehta, 2019; Gaydosh, et al., 2019; Kerr et al., 2009; White et al., 2006), making examining the role of young adult health in shaping coresedential arrangements especially important. As intergenerational coresidence patterns continue to shift in the 21^st century, researchers should continue to look beyond economic determinants of coresidence to health dynamics. We further call on future research to explore how other family trends between parents and children, as well as between spouses, siblings, and grandparents, may be impacted by health outcomes.