Skip to main content
NCBI home page
The Journals of Gerontology Series B: Psychological Sciences and Social Sciences logoLink to The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
. 2021 Jun 17;76(9):1857–1869. doi: 10.1093/geronb/gbab109

Adult Children’s Educational Attainment and Parent Health in Mid- and Later-Life

Christopher R Dennison 1, Kristen Schultz Lee 1,✉
Editor: Deborah S Carr
PMCID: PMC8557833  PMID: 34139008

Abstract

Objectives

Intergenerational models of adult health contend that children’s educational attainments influence the health and well-being of their parents. However, it is unclear how much of this association is confounded by background characteristics that predict both children’s educational attainments and parents’ subsequent health, particularly in the United States.

Methods

Data from the National Longitudinal Study of Adolescent to Adult Health Parent Study are used to examine how having no children who completed college influences parents’ self-rated health and depressive symptoms. We rely on propensity score methods to more squarely assess this relationship net of potential confounding bias and to test for heterogeneity in the consequences associated with having no children who completed college.

Results

Having no children who completed college is negatively associated with parents’ self-rated health and positively associated with depressive symptoms. After statistically balancing differences in background characteristics between groups, these associations remain, though the magnitude of the coefficients is attenuated. Supplemental matching analyses suggest that while the association between children’s education and self-rated health might be spurious, the association with depressive symptoms is more robust. Additionally, among parents with the highest propensity for having no children who complete college, the consequences on depressive symptoms are greatest.

Discussion

This study pays particular attention to selection-related concerns surrounding the association between offspring educational attainment and parent well-being in the United States. These findings are important given the call for investments in children’s educational opportunities as promoting both the well-being of adult children and their parents.

Keywords: Cumulative inequality, Intergenerational health models, Propensity score methods

Higher educational attainment has become an increasingly important predictor of socioeconomic and health outcomes (Becker, 2009; Mirowsky & Ross, 2015; Stevens et al., 2008). Higher levels of education are associated with higher paying and more autonomous jobs (Becker, 2009; Mirowsky & Ross, 2015; Stevens et al., 2008), stable marriages (McLanahan, 2004), better neighborhoods (Browning & Cagney, 2003), and better access to health care and healthier lifestyles (Mirowsky & Ross, 2015). All of these factors help explain the positive association between education and health.

In the United States in particular, family income inequalities in enrollment and completion of higher education and the relatively high returns to a college degree have resulted in growing educational, income, and health inequalities. Net financial returns to higher education are much higher in the United States than the OECD average (OECD, 2020). However, access to higher education has become increasingly stratified by family income in the United States. Bailey and Dynarski (2011) found that college completion rates increased across U.S. cohorts (born 1961–1964 and 1979–1982) by 4 percentage points for low-income cohorts but by 18 percentage points for high-income cohorts. These growing family income inequalities in completion of higher education, and the consequences for individual health outcomes in the United States, have been extensively documented in the research literature (e.g., Jackson & Holzman, 2020; Zajacova & Lawrence, 2018).

Although the effects of an individual’s educational attainment on their health have been studied extensively, the effects of adult children’s educational attainment on parents’ health outcomes have received relatively less attention. The life-course perspective’s attention to linked lives has led to the more recent development of intergenerational models of adult health, including the social foreground perspective (Wolfe et al., 2018a; Zimmer et al., 2016). In these intergenerational models of adult health, highly educated children provide more resources to, and impose fewer demands on, their parents. This, in turn, is associated with lower mortality risk (Friedman & Mare, 2014; Wolfe et al., 2018b; Zimmer et al., 2016) and better mental health (Yahirun et al., 2020) for parents.

Although the intergenerational health literature makes the important contribution of adding an attention to linked lives to models of adult health, several questions remain unanswered. For instance, is the association between adult children’s educational attainment and parent health purely due to selection factors? Indeed, families with higher socioeconomic status (SES) are more likely to have children with higher levels of educational attainment and parents with better health outcomes. As a result, it remains unclear from past studies how much of this association between offspring’s educational attainment and parent health is attributable to selection factors, particularly in the United States. In addition, does this association vary depending on a parent’s likelihood of having no children attend college?

Our research uses data from the 2015–2017 parent study from the National Longitudinal Study of Adolescent to Adult Health (Add Health) to examine the association between children’s educational attainment and parent health. We pay particular attention to selection-related concerns by utilizing several propensity score methods as a means to statistically balance differences in background characteristics between parents with and without children who completed college. In doing so, we attempt to more squarely assess how children’s educational attainments influence parents’ health and well-being, net of confounding bias. After accounting for selection factors, we then investigate whether the association between children’s educational attainment and parent health varies depending on the parent’s likelihood of having no children complete college. By testing heterogeneous treatment effects (HTE) in the association between children’s educational attainment and parent health, we weigh in on scholarly discussions surrounding cumulative inequality and education.

Background

Theories of Intergenerational Influences on Health

It is well-established that parents’ educational attainment has an effect on children’s SES and consequently their health. According to the long arm of childhood theory (Hayward & Gorman, 2004), parental SES affects the health risks and health behaviors that children are exposed to early in life. These early exposures have implications for offspring’s health in mid- and later-life. Similarly, from a cumulative inequalities perspective, early life disadvantage (due to parents’ low SES) leads to the accumulation of disadvantages over time in multiple life domains, including health (Ferraro et al., 2009).

What has received less attention is the effect of adult children’s educational attainment on their parents’ health. A life-course perspective emphasizes the interdependence of parent and child trajectories over time and points to the possible links between offspring attainment and parent outcomes (see, e.g., Gilligan et al., 2018). The social foreground perspective predicts that adult children who attain higher levels of education have greater resources, which they can use to improve the health and well-being of their parents (Torssander, 2013; Wolfe et al., 2018a; Zimmer et al., 2016). This represents a divergence from human capital models that conceptualize education as a resource benefitting the individual (Becker, 2009). The social foreground perspective also builds on fundamental cause theory (Link & Phelan, 1995) by highlighting how more highly educated individuals have access to more, and better-quality, resources that benefit not only their own health, but also the health of their parents (Zimmer et al., 2016).

Several specific mechanisms have been proposed to link adult children’s educational attainment and parent health. First, intergenerational transfers, both from adult children to their parents and from parents to their adult children, are associated with offspring’s educational attainment and parents’ health outcomes and may causally link the two. The role played by these transfers, however, depends on the type of transfer (i.e., financial, instrumental, emotional) and the direction (i.e., upward or downward). Overall, most research finds that offspring with lower levels of education provide fewer upward transfers to their parents (McGarry & Schoeni, 1995;Smith-Greenaway et al., 2018) and are in greater need of assistance from their parents (McGarry & Schoeni, 1995). In general, this pattern of transfers is negatively associated with parent health (Mao et al., 2020; Thoits, 2011).

Other explanations for the relationship between offspring educational attainment and parent health point to the role of parental worrying, the stigma associated with poor child outcomes, and also the effect of adult children’s efforts to promote a healthy lifestyle for their parents. Adult children with lower levels of education are on average in a more precarious economic situation and therefore may induce greater parental worrying. Worrying leads to reduced sleep and other poor health outcomes (Beck et al., 2001; Seidel et al., 2018). In addition, parents may feel that poor outcomes among their adult children, including lower levels of educational achievement, reflect poorly on their parenting and may be stigmatizing for middle and higher SES parents in particular, negatively affecting parent well-being (Goldman, 2019; Greenfield & Marks, 2006). Finally, more highly educated children may promote healthier lifestyles in their parents both by providing their parents with health-related resources (e.g., healthy foods, access to high-quality medical care) and by modeling healthy behaviors (e.g., not smoking, exercising regularly; Friedman & Mare, 2014; Lee, 2018).

Based on this rich theoretical underpinning for the association between children’s educational attainment and parent health, scholars have argued that greater investments should be made in the educational opportunities available to young people as a way of improving their parents’ health outcomes (Zimmer et al., 2016). Findings from this literature have also been used to weigh in on the long-standing generational equity debate (Preston, 1984). More specifically, scholars have concluded that different generations are not in competition for resources, but rather that resources directed to the younger generation can benefit the older generation as well (Friedman & Mare, 2014). The implications for policy and theory that emerge from the intergenerational health literature hinge on the assumption that the association between adult offspring’s education and parent health is causal.

Selection and Intergenerational Models of Health

Although it is possible that the association between offspring’s educational attainment and parent health is causal, another possibility is that the association between these factors may be due entirely to selection. Parents who are at greatest risk of poor health outcomes are, on average, from lower SES families, less likely to be married, more likely to have preexisting health problems and risky health behaviors, and less likely to have the cultural capital needed for adeptly navigating educational and medical institutions. These factors predispose an individual to poor health and also increase the likelihood of lower levels of educational attainment among their offspring. Some of the factors, particularly measures of SES, have been controlled for in previous analyses, but the regression and hazard models of observational data that make up the majority of past studies do not attempt to estimate how much of the association between offspring education and parent health is attributable to selection factors (see Yahirun et al. 2020 for an exception). By comparing the association between children’s educational attainment and parent health in samples that have and have not been matched based on their propensity to have college-educated children, we can estimate the proportion of the association due to selection factors. It is possible that once the characteristics distinguishing those families with some college-educated adult children from those families with no college-educated adult children are accounted for, no statistically significant association between adult children’s educational attainment and parent health will remain.

Some studies have accounted for selection factors in their analyses of intergenerational health in contexts outside the United States, with some finding that at least part of the association between children’s education and parent health is causal (De Neve & Fink, 2018; Ma, 2019; Torssander, 2013), and others finding no statistically significant causal relationship overall (Lundborg & Majlesi, 2018). Torssander (2013) analyzed parents’ death risk in Sweden using parent sibling fixed effects models. These models account for some selection factors by “holding constant” genetic factors shared among siblings that could account for both their offspring’s educational attainment and their own health outcomes. Other studies used an instrumental variable approach with quasi-experimental data to estimate the causal effects of children’s schooling on parent health (De Neve & Fink, 2018; Lundborg & Majlesi, 2018; Ma, 2019). These studies capitalized on educational reforms to estimate the causal effects of children’s education on parents’ health outcomes. They were able to account for the effects of unobserved selection factors associated with both children’s educational attainment and parent health outcomes. However, none of these studies examined the causal effects of children’s education on parent health in the context of the United States.

Propensity score models are particularly appropriate methods for dealing with selection issues in the United States where there have not been recent broad scale educational reforms which would create a natural experiment of the sort analyzed in other countries (De Neve & Fink, 2018; Lundborg & Majlesi, 2018; Ma, 2019). Moreover, compared with other methods geared toward addressing potential confounding bias (e.g., instrumental variable models; fixed effects regression), propensity score models also allow us to estimate how much of the overall association between children’s educational attainment and parent health is in fact due to observable selection factors. Additionally, if we do find that the effect of offspring educational attainment on parents’ health is robust even after accounting for observable background characteristics, we can test the role of hidden (i.e., unobservable) bias (Rosenbaum, 2002). Indeed, other studies have used propensity score models in the analysis of an individual’s own educational attainment and their health outcomes (Greenfield et al., 2019); however, we are not aware of any study that has used propensity score models to investigate selection issues in intergenerational models of health in the United States. Most U.S. studies include potential confounding variables as controls, but we argue that our analysis makes a unique contribution by including a wider range of confounding variables, including measures of parent health before children complete their schooling, and by estimating how much of the relationship between children’s education and parent health is explained by confounding factors.

Cumulative Inequality and Parent Health

Once selection factors are addressed in analyzing the association between children’s educational attainment and parent health, the question of variation in the magnitude (and even the direction) of the association arises. Life-course scholars have examined adult health disparities and connected them to family socioeconomic disadvantage, with those from lower SES families falling further behind those from higher SES families with age (Ross & Wu, 1996; Shuey & Willson, 2008). Cumulative Disadvantage theory predicts that individuals starting off with fewer socioeconomic resources will be exposed to more adversities and greater risks than those starting off in a relatively more advantaged position (Dannefer, 2003). An implication of this accumulation of disadvantage is that health disparities between lower and higher SES individuals will grow with age. Ferraro et al. (2009) further elaborate that the disadvantages associated with lower SES are transmitted intergenerationally. Although much of the focus of previous research has been on the downward transmission of disadvantage, our focus is on how children’s attainments can also put their parents at greater risk of poor health outcomes. The stratification multilevel method (Xie et al., 2012) allows us to analyze how the association between children’s educational attainment and parent health varies based on the parent’s likelihood of having no children complete college. Using this method, we can weigh in on how children’s educational attainment may have disparate consequences for parent health depending on the parent’s socioeconomic background (i.e., their propensity to have no children complete college).

The Present Study

Data

Data for this study come from the Add Health Parent Study (AHPS). Beginning in 1994–1995 (Wave I), Add Health collected data on over 20,000 adolescents in grades 7–12, as well as data from their parents, peers, and school administrators. Respondents have been reinterviewed four times since then, with the most recent round of interviews occurring in 2016–2018, when approximately 12,000 respondents ages 33–43 were reinterviewed. Central to the present study, Add Health also reinterviewed a probability sample of 2,013 Wave I parents in 2015–2017 when they were ages 50–80 and gathered information about their current health, financial status, and children’s educational outcomes, among many other topics. Parents were eligible for this study if (a) they were not deceased or incarcerated at the time of the sampling and (b) they were the biological parent, adoptive parent, or stepparent of a Wave I Add Health respondent who was not deceased at the time of the sampling (Eischen et al., 2019).

The weighted AHPS sample is a probability sample of parents of a nationally representative sample of adolescents in 1995. Preliminary comparisons with other nationally representative data sets indicate that the AHPS sample is comparable to the 2014 Health and Retirement Study and the 2014 National Health and Nutrition Examination Survey samples, but is slightly more socioeconomically advantaged and healthier (Oi et al., 2018). A majority of the respondents in the parent sample are women because Add Health targeted the female head of the household for the Wave I parent survey based on evidence that mothers are more familiar with children’s activities and characteristics (Harris & Hotz, 2020). Our study uses data from 1,681 of these parent respondents who have no missing data and who have valid sample weights.

Health-Related Outcomes (2015–2017 Follow-Up)

Most studies of the association between children’s education and parent health have used parent mortality as the measure of health (Friedman & Mare, 2014; Torssander, 2013; Wolfe et al., 2018a, 2018b; Zimmer et al., 2016). We selected self-rated health and depressive symptoms as the measures of parent health in our analysis to broaden the measures of parent health investigated in the intergenerational health literature. Self-rated measures of health have been established as valid indicators of objective health status (Maddox & Douglass, 1973) and depressive symptoms are a common measure of mental health (see Yahirun et al. 2020 as an example of one study in the intergenerational health literature that uses depressive symptoms as the outcome). Self-rated health is a continuous measure based on whether respondents believe their health is (1) poor, (2) fair, (3) good, (4) very good, or (5) excellent. The indicators used to measure depressive symptoms are a subset of the Center for Epidemiological Studies-Depression (CES-D) scale (Radloff, 1977) indicating how often (ranging from (1) never or rarely to (4) most or all of the time) during the past 7 days respondents reported they could not shake off the blues; felt depressed; were happy (reverse-coded); felt sad; and felt life was not worth living. We sum the responses to these questions to create a measure for depressive symptoms, where higher values reflect more depressive symptoms (Cronbach’s α = 0.794). Previous research has demonstrated such abbreviated versions of the CES-D scale to be accurate assessments of depressive symptoms among older adults compared to the full scale (Andresen et al., 1994). By using self-rated health and depressive symptoms as our outcome measures of parent health, we are able to estimate the effects of children’s education on parent health at younger parental ages than studies of parent mortality.

Children’s Educational Attainment (2015–2017 Follow-Up)

During the most recent round of parent interviews, parents were asked about the educational attainments of all of their children. Following the justification raised by Yahirun et al. (2020) that there is recent research evidence of the growing importance of a college degree for health outcomes in the United States, we focus on the distinction between completing a college degree and lower levels of education. We measure children’s educational attainments as a dichotomous variable indicating that none of the respondent’s children had completed college (no children completed college), with parents with a least one child who completed college in the reference category.

Background Controls (Wave I Unless Noted Otherwise)

Demographic, family, social–psychological, health, and community characteristics that are associated both with the likelihood of the respondents’ children graduating from college and with parent health are included in the analysis. Demographic controls for parent respondents include a measure indicating that the respondent is female, as well as the respondent’s age (in years). Race and ethnicity are operationalized via five mutually exclusive measures, including being: non-Hispanic White; of Hispanic origin; non-Hispanic Black; non-Hispanic Asian; and non-Hispanic other racial group. Socioeconomic controls include a measure for the parent respondent’s educational attainment (in years). We also control for whether parents reported receiving welfare (welfare receipt), being unemployed, or having trouble paying bills (bill problems).

Parents’ familial characteristics have important consequences for their children’s educational attainment (McLanahan & Percheski, 2008) and their own health (Umberson & Thomeer, 2020). We include three mutually exclusive measures indicating that the parent is single, married, or widowed/divorced/separated. The number of past marriage-like relationships is assessed via three mutually exclusive measures indicating that the respondent had no past marriage-like relationship, only one past marriage-like relationship, or more than one past marriage-like relationship. We also control for the gender composition of their children: whether the respondent has no girls, some girls, or all girls.

We include a range of past social–psychological and health-related measures for the parent respondents, including whether they reported ever having five or more drinks on one occasion during the past month (binge drink). We also control for whether respondents smoke. Several dichotomous measures indicating different health problems are included, such as obesity, migraine headaches, allergies, asthma or emphysema, alcoholism, diabetes, and being disabled. A control for past self-rated health is also included and operationalized in a similar manner as our outcome measure. Additionally, two binary measures indicating that the respondent had been diagnosed with depression at some point prior to the Wave I interview and that they are generally unhappy are included. By including measures of self-rated health and depression diagnosis at Wave I, we are able to predict self-rated health and depressive symptoms at the most recent wave, net of past self-rated health and depression.

Parent expectations for their children and their engagement in their child’s school have implications for children’s educational achievement (Gonzalez-Pienda et al., 2002) and potentially for the parents’ health, given the importance of future expectations (Hitlin & Johnson, 2015) and productive engagement (Kail & Carr, 2017) for adult health. A dichotomous measure indicating whether parents were a member of a parent/teacher organization is included as well as three dichotomous measures assessing parents’ level of disappointment if their child does not graduate from college. Parents are classified as either not disappointed, somewhat disappointed, or very disappointed. This measure pertains solely to the Add Health respondent. Thus, while in some cases it is not possible to explicitly measure parents’ educational expectations for all of their children, this measure gives at least some indication about parent’s outlook for their offspring.

Neighborhood quality affects both children’s educational attainment (Alexander et al. 2014) and adult health outcomes (Weden et al., 2008). We include two binary measures for whether there is a trash problem or a drug problem in the neighborhood, as well as dichotomous measures indicating that the respondent would like to move away from their neighborhood and that they chose their neighborhood for the schools. Urban is a binary measure indicating that respondents lived in an urban residence. Neighborhood disadvantage is based on the average of four census tract measures indicating the percentage of: adults unemployed; households receiving public assistance; families below the poverty line; and households headed by a single mother (Cronbach’s α = 0.896).

Analytic Strategy

We rely on propensity score methods as a means to address selection-related concerns surrounding the association between children’s educational attainments and parent health (Guo & Fraser, 2015). Propensity score methods attempt to mitigate observable differences in background characteristics between treated and controlled respondents—in this case, parents with no children who completed college (the treatment group) and parents with at least one child who completed college (the control group). To do so, we first regress the treatment variable on the set of background controls via logistic regression and retain the predicted probabilities (or propensity scores). We considered a range of approaches for estimating our propensity scores, including using higher-order terms as well as incorporating a range of interaction terms. Regardless of our propensity score specification approach, we found our results to be consistent with those presented here. The predicted probabilities can range from 0 to 1 and represent one’s propensity for having no children who completed college. After restricting our sample to respondents who fall within the region of common support—that is, the range of propensity scores where treated and controlled respondents overlap—we utilize kernel matching (kernel: Gaussian; bandwidth: 0.06) to estimate our primary matched sample. This kernel matching algorithm matches each treated respondent with each controlled respondent, and weights are assigned to controlled respondents depending on how close their propensity score is to the treated respondent (for more detail about the propensity score matching method, please see Supplementary Appendix A). Although the role of selection could also be examined with a multivariable regression model that includes controls for observable confounders, propensity score methods accomplish this goal while also increasing our confidence that controlled respondents actually resemble those in the treatment group in terms of their observable background characteristics.

After matching, we assess balance in the background controls by calculating the standardized differences for each respective variable between treated and controlled respondents. A conventional threshold for balance is a standardized difference below |0.20| (Porter & Vogel, 2014), while a more sensitive threshold of |0.10| is also recommended (Austin, 2009). After ensuring that balance is achieved, we then separately regress our measures of self-rated health and depressive symptoms on our indicator that no children completed college in our propensity-weighted sample to examine these associations net of differences in background characteristics between groups. In these models, we also include a control for self-rated health and depression diagnosis at Wave I, respectively, so as to partially control for stable differences in these health-related measures.

We also consider whether the overall pattern of results varies by one’s propensity to have no children who completed college. This is accomplished by utilizing the stratification multilevel method developed by Xie et al. (2012), which is an extension of propensity score methods geared toward assessing treatment effect heterogeneity. Respondents are first sorted into balanced strata based on their propensity score. In our data, four balanced strata were estimated so that we could ensure there was a sufficient number of respondents within each group. Stratum 1—that is, those with the lowest propensity for the treatment (i.e., having no children who completed college)—includes respondents whose propensity score falls between 0.04 and 0.20. The propensity for the treatment increases across the stratum rank, as propensity scores in each stratum range from: 0.20 to 0.40 (stratum 2); 0.40 to 0.60 (stratum 3); and 0.60 to 0.92 (stratum 4). Thus, those in stratum 4 have the greatest propensity for having no children who completed college. Within each stratum, the treatment effect is estimated, and these coefficients represent the level-1 estimates. To test for HTE across the strata, we then use a variance-weighted least squares regression to regress the level-1 estimates on the stratum rank, thus yielding a linear estimate of how the magnitude of the treatment effect changes across strata.

Results

Table 1 reports the means, proportions, and standardized differences of all variables for treated and controlled respondents in both the unmatched and matched samples. In the unmatched sample, there are several notable differences between treated and controlled respondents, as evident by the large standardized differences in covariates. For instance, parents with no children who completed college (i.e., the treatment group) report, on average, fewer years of schooling compared to the control group. Those in the treatment group were also more likely to receive welfare, be unemployed, and experience bill problems compared to parents with children who attained a college degree. They were also more likely to report lower self-rated health and being a smoker. In the matched sample, however, the differences in background characteristics between treated and controlled respondents are noticeably attenuated, and all standardized differences fall below |0.10|.

Table 1.

Weighted Bivariate Statistics and Standardized Differences Between Respondents With and Without Children Who Completed College

Unmatched sample Matched sample
Variable No children completed college Some children completed college |Std. diff.| No children completed college Some children completed college |Std. diff.|
Demographic characteristics
 Female 0.964 0.965 0.003 0.966 0.968 0.013
 Age 39.852 42.041 0.386 39.947 40.160 0.038
 White 0.671 0.778 0.241 0.682 0.678 0.009
 Hispanic 0.133 0.096 0.119 0.135 0.144 0.027
 Black 0.165 0.094 0.212 0.153 0.140 0.037
 Asian 0.012 0.023 0.088 0.012 0.011 0.006
 Other 0.019 0.010 0.082 0.018 0.026 0.058
Socioeconomic background
 Educational attainment (years) 12.609 14.779 0.837 12.687 12.787 0.042
 Welfare receipt 0.133 0.044 0.318 0.113 0.104 0.027
 Unemployed 0.099 0.033 0.270 0.083 0.082 0.004
 Bill problems 0.216 0.130 0.229 0.206 0.226 0.049
Familial characteristics
 Single 0.085 0.024 0.272 0.072 0.070 0.008
 Married 0.669 0.783 0.258 0.681 0.696 0.033
 Widowed/divorced/separated 0.246 0.193 0.128 0.247 0.234 0.031
 No girls 0.216 0.156 0.155 0.210 0.178 0.082
 Some girls 0.635 0.670 0.072 0.637 0.658 0.043
 All girls 0.149 0.174 0.070 0.153 0.164 0.032
 No past relationships 0.043 0.024 0.103 0.041 0.032 0.047
 One past relationship 0.630 0.755 0.274 0.639 0.664 0.051
 More than one past relationship 0.328 0.221 0.241 0.320 0.304 0.034
Social–psychological and health-related measures
 Binge drink 0.160 0.116 0.129 0.154 0.171 0.046
 Smoke 0.413 0.187 0.508 0.403 0.393 0.021
 Obesity 0.186 0.186 0.001 0.190 0.200 0.026
 Migraine headaches 0.300 0.236 0.146 0.291 0.292 0.002
 Asthma 0.117 0.069 0.163 0.106 0.111 0.015
 Alcoholism 0.041 0.014 0.164 0.033 0.037 0.019
 Diabetes 0.037 0.030 0.043 0.036 0.034 0.010
 Disabled 0.053 0.039 0.068 0.053 0.055 0.005
 Self-rated health 3.427 3.829 0.406 3.431 3.450 0.019
 Depression 0.133 0.088 0.145 0.129 0.131 0.006
 Unhappy 0.042 0.025 0.095 0.039 0.041 0.013
Community characteristics and involvement
 Member of a parent/teacher organization 0.192 0.442 0.559 0.197 0.207 0.025
 Chose their neighborhood for the schools 0.426 0.550 0.250 0.430 0.437 0.015
 Trash problem 0.544 0.432 0.226 0.540 0.545 0.009
 Drug problem 0.475 0.332 0.295 0.476 0.469 0.014
 Like to move away from their neighborhood 0.516 0.411 0.212 0.509 0.494 0.031
 Urban 0.564 0.539 0.051 0.559 0.546 0.027
 Neighborhood disadvantage 10.452 7.014 0.519 10.131 9.870 0.036
Disappointment regarding child’s education
 Not disappointed 0.234 0.102 0.358 0.227 0.199 0.068
 Somewhat disappointed 0.418 0.417 0.002 0.426 0.439 0.027
 Very disappointed 0.349 0.481 0.272 0.348 0.363 0.031
Sample size 647 1,034 629 961
Open in a new tab

Notes: Matched sample excludes 18 treated respondents and 73 controlled respondents who fall out of the region of common support. Matched sample generated via kernel matching (kernel: Gaussian; bandwidth: 0.06).

With the differences in background characteristics statistically balanced in the matched sample, we examine the associations between having no children who completed college and subsequent self-rated health and depressive symptoms. Table 2 reports the coefficients from ordinary least squares regression models using both the unmatched and matched samples. In the unmatched sample, having no children who completed college is associated with a 0.270 decrease in self-rated health (p < .001). In the matched sample, having no children who completed college remains negatively associated with self-rated health (i.e., −0.136; p < .05); however, the coefficient is reduced in magnitude by approximately 49% [(−0.270–0.136)/−0.270]. Similarly, having no children who completed college is associated with a 0.758 increase in depressive symptoms in the unmatched sample (p < .001). Furthermore, this association remains in the matched sample, although the coefficient is reduced by roughly 41% (0.447; p < .05). To further aid interpretation, Figure 1 displays the predicted self-rated health and depressive symptoms scores for each group from both the unmatched and matched samples.

Table 2.

Ordinary Least Squares Regression Estimates of Not Having Children Who Completed College

Unmatched Matched
Outcome b SE T-value b SE T-value
Self-rated health −0.270 0.060 −4.468*** −0.136 0.069 −1.981*
Depressive symptoms 0.758 0.152 4.983*** 0.447 0.193 2.317*
Open in a new tab

Notes: SE = standard error. Models predicting self-rated health and depressive symptoms include control for self-rated health and depression diagnosis at Wave I, respectively. All models include the Add Health sample weights. Matched sample generated via kernel matching (kernel = Gaussian; bandwidth = 0.06). Coefficients in the matched sample represent the average treatment effect on the treated. Bootstrapped standard errors were estimated for the matched sample across 1,000 replications. Sample size: 1,590 (629 treated; 961 controlled).

*p < .05. ***p < .001 (two-tailed tests).

Figure 1.

Figure 1.

Open in a new tab

Predicted self-rated health and depressive symptoms scores by children’s educational attainment.

As previously described, one advantage of kernel matching is the ability to retain the entire sample of results, thus increasing the generalizability of the findings. At the same time, including all respondents in the matched sample also potentially means that poorer matches are included in the analysis. With this in mind, we replicated our analyses using several different matching algorithms. Details about these analyses can be found in Supplementary Appendix A; here, we briefly summarize the overall pattern of results from these additional analyses. When using alternative matching methods, the effect of having no children who completed college on self-rated health is attenuated to marginal significance (and in some instances, nonsignificance). Thus, this association is potentially spurious. In the case of depressive symptoms, the association is more robust across alternative matching algorithms.

We next turn to examine the heterogeneous effects of having no children who completed college on parent health. Table 3 presents the level-1 and level-2 estimates for having no children who completed college on parents’ self-rated health and depressive symptoms. Recall that the level-1 estimates represent the treatment effect within each stratum, while level-2—the test for heterogeneous effects—represents the change in treatment effects as stratum rank increases (i.e., as the propensity to have no children complete college increases).

Table 3.

Heterogeneous Treatment Effects of Having No Children Who Completed College on Parents’ Self-Rated Health and Depressive Symptoms

Self-rated health Depressive symptoms
b SE b SE
Propensity score strata (level-1)
Stratum 1: (40 treated; 360 controlled)
 Propensity score: (0.04–0.2) −0.244 0.155 −0.043 0.248
Stratum 2: (136 treated; 310 controlled)
 Propensity score: (0.2–0.4) −0.042 0.124 0.099 0.240
Stratum 3: (197 treated; 192 controlled)
 Propensity Score: (0.4–0.6) −0.210 0.110 0.435 0.297
Stratum 4: (256 treated; 99 controlled)
 Propensity score: (0.6–0.92) −0.173 0.116 0.804* 0.348
Level-2 slope
 Stratum −0.007 0.059 0.278* 0.131
Open in a new tab

Notes: SE = standard error. Models predicting self-rated health and depressive symptoms include control for self-rated health and depression diagnosis at Wave I, respectively.

*p < .05 (two-tailed tests).

Beginning with self-rated health, all of the estimates within each stratum are in the expected direction; however, they do not approach statistical significance. Moreover, the level-2 slope for self-rated health is also not significant, thus suggesting that the effect of having no children who completed college on self-rated health does not vary by one’s propensity. As for depressive symptoms, a statistically significant estimate is observed in stratum 4, which suggests that having no children who completed college is associated with a 0.804 increase in depressive symptoms among those with the highest propensity for the treatment. Perhaps most notable, however, is the significant level-2 estimate, which suggests that a unit-increase in stratum rank is associated with a 0.278 increase in the effect of having no children who completed college on parents’ depressive symptoms. As a means to further contextualize these heterogeneous effects, Figure 2 displays both the individual stratum estimates as well as the level-2 slope for depressive symptoms. The figure shows that as propensity score stratum rank increases, so too does the magnitude of the treatment effect. Stated differently, the positive association between having no children who completed college and parents’ depressive symptoms appears to be strongest among those with the greatest propensity for having children with no college degrees.

Figure 2.

Figure 2.

Open in a new tab

Heterogeneous treatment effects of having no children who completed college on depressive symptoms.

Discussion and Conclusion

Growing educational and income disparities in the United States have sparked increased concern regarding the consequences of growing inequalities for other life domains, especially health (Jackson & Holzman, 2020). Intergenerational models of adult health draw attention to the implications of growing educational and income inequalities for the health of individuals and their parents. This attention to offspring’s influence on parent health is consistent with the life-course perspective’s attention to linked lives and the interdependence of parent and child life trajectories (see, e.g., Gilligan et al., 2018). However, past research on the effects of children’s educational attainment on parent health in the United States has not been able to address the question of how much of this association is due to confounding selection factors. In this paper we ask: can we explain the entire association between adult children’s educational attainment and parent health by background characteristics that are associated with both offspring educational attainment and parent health? Secondly, to what extent does this association vary by one’s propensity to have no college-educated children?

We use propensity score methods to address the extent to which the association between children’s educational attainment and parent health is spurious and due to background characteristics that predict both offspring’s educational attainment and parent health. In our kernel-based matched sample (i.e., our most generalizable sample), we find that the association remains statistically significant but is reduced in magnitude by 41% (for parents’ depressive symptoms) to 49% (for parents’ self-rated health). Yet, when using alternative matching algorithms, we find that the effect of children’s educational attainment on parent’s self-rated health is reduced to marginal significance at best. The relationship between children’s education and parent’s reports of depressive symptoms, however, was more robust across matching algorithms. Indeed, this is among the first studies we know of to use propensity score methods to evaluate the role played by selection factors in the association between offspring educational attainment and parent health in the United States. As such, our goal was to provide a comprehensive assessment of the role that observable confounders play in these relationships. Our analysis makes a contribution to the intergenerational health literature by (a) adding a wider range of potential confounders than previous studies, (b) estimating how much of the association is explained by selection factors, and (c) including measures of parent depression and self-rated health before offspring completed their education.

To further analyze potential HTE, we use the stratification multilevel method to investigate to what extent the association between children’s educational attainment and parent health varies depending on the parent’s propensity to have no college-educated children. Although we did not find significant variation across propensity levels in the association between having no children complete college and self-rated health, we did find that the positive association with parent’s depressive symptoms was strongest among those with the greatest propensity for having no children complete college (i.e., the most disadvantaged parents in the sample). This supports cumulative disadvantage and cumulative inequality accounts of the accumulation of disadvantage with age (see, e.g., Dannefer, 2003; Ferraro et al., 2009). This also provides evidence to support the upward intergenerational transmission of disadvantage as a mechanism contributing to the accumulation of disadvantage and widening health disparities with age.

These results lend qualified support to policies proposed in the literature that would increase educational opportunities as a means of improving health outcomes in older generations (Zimmer et al., 2016), particularly if the policy is targeting adult mental health outcomes. The results also support Friedman and Mare’s (2014) contention that concerns over generational equity are largely misplaced. Despite long-standing claims that the interests of different generations are at odds (Preston, 1984), this research shows that different generations are in fact interdependent. Indeed, investments in educational opportunities have the potential to not only increase the individual’s SES and health, but also the mental health of their parents and particularly the mental health of low SES parents.

Although our research contributes an analysis of how much of the association between children’s education and parent health in the United States is attributable to selection factors, it is not without limitations. First, propensity score models account for observed sources of heterogeneity in the population. This means that omitted variables could bias the results. Second, this analysis only analyzes the extent to which selection factors can account for the association between offspring education and parent health. It does not analyze the possible pathways linking these variables, but we think this is an important direction for future research.

In sum, our research weighs in on a persistent concern in the intergenerational health literature: to what extent is the association between offspring education and parent health spurious and to what extent does it vary by propensity to have no college-educated children? Our propensity score model results for parents’ self-rated health do not rule out the possibility that the association is spurious, once background selection factors are accounted for. However, a statistically significant association between children’s education and parents’ depressive symptoms persists even after background selection factors are accounted for. Parents report better mental health when at least one of their children completed college. Moreover, parents with the greatest propensity to have no college-educated children experienced the strongest association between children’s low level of educational attainment and depressive symptoms (but not self-rated health). Our research adds to the growing literature identifying the advantages of a college degree including individual economic mobility (Becker, 2009), marital outcomes (McLanahan, 2004), health (Ross & Wu, 1996), and well-being (Bauldry, 2015). The results lend support to a call for public investments in educational opportunities as a means of not only improving the socioeconomic and health outcomes of one generation, but also the mental health outcomes of their parents and low-SES parents in particular.

Funding

This work was supported by the Spencer Foundation (201800093 to K. S. Lee). This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Information on how to obtain the Add Health data files is available on the Add Health website (https://addhealth.cpc.unc.edu/). No direct support was received from grant P01-HD31921 for this analysis. The Add Health Parent Study/Parents (2015–2017) data collection was funded by a grant from the National Institute on Aging (RO1AG042794) to Duke University, V. Joseph Hotz (PI) and the Carolina Population Center at the University of North Carolina at Chapel Hill, Kathleen Mullan Harris (PI).

Conflict of Interest

None declared.

Author Contributions

C. R. Dennison helped plan the study, performed all statistical analyses, and helped write the paper. K. S. Lee helped plan the study and write the paper.

Supplementary Material

gbab109_suppl_Supplementary_Materials
Click here for additional data file. (199.8KB, pdf)

References

  1. Alexander, K., Entwisle, D., & Olson, L. (2014). The long shadow: Family background, disadvantaged urban youth, and the transition to adulthood. Russell Sage Foundation. [Google Scholar]
  2. Andresen, E. M., Malmgren, J. A., Carter, W. B., & Patrick, D. L. (1994). Screening for depression in well older adults: Evaluation of a short form of the CES-D (Center for Epidemiologic Studies Depression Scale). American Journal of Preventive Medicine, 10(2), 77–84. doi: 10.1016/S0749-3797(18)30622-6 [DOI] [PubMed] [Google Scholar]
  3. Austin, P. C. (2009). Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Communications in Statistics-Simulation and Computation, 28, 1228–1234. doi: 10.1080/03610910902859574 [DOI] [Google Scholar]
  4. Bailey, M. J., & Dynarski, S. M. (2011). Inequality in postsecondary education. In Duncan G. J. & Murnane R. J. (Eds.), Whither opportunity (pp. 117–132). Russell Sage. [Google Scholar]
  5. Bauldry, S. (2015). Variation in the protective effect of higher education against depression. Society and Mental Health, 5(2), 145–161. doi: 10.1177/2156869314564399 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Beck, R., Perkins, T. S., Holder, R., Robbins, M., Gray, M., & Allison, S. H. (2001). The cognitive and emotional phenomenology of depression and anxiety: Are worry and hopelessness the cognitive correlates of NA and PA? Cognitive Therapy and Research, 25(6), 829–838. doi: 10.1023/A:1012983726272 [DOI] [Google Scholar]
  7. Becker, G. S. (2009). Human capital: A theoretical and empirical analysis, with special reference to education. University of Chicago Press. [Google Scholar]
  8. Browning, C. R., Cagney, K. A., & Wen, M. (2003). Explaining variation in health status across space and time: Implications for racial and ethnic disparities in self-rated health. Social Science & Medicine (1982), 57(7), 1221–1235. doi: 10.1016/s0277-9536(02)00502-6 [DOI] [PubMed] [Google Scholar]
  9. Dannefer, D. (2003). Cumulative advantage/disadvantage and the life course: Cross-fertilizing age and social science theory. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 58(6), S327–S337. doi: 10.1093/geronb/58.6.s327 [DOI] [PubMed] [Google Scholar]
  10. De Neve, J.-W., & Fink, G. (2018). Children’s education and parental old age survival—Quasi-experimental evidence on the intergenerational effects of human capital investment. Journal of Health Economics, 58, 76–89. doi: 10.1111/1475-6773.12090 [DOI] [PubMed] [Google Scholar]
  11. Eischen, F. F., Monath, A., Harris, K. M., & Hotz, V. J. (2019). Parents (2015–2017) data documentation. Carolina Population Center. [Google Scholar]
  12. Ferraro, K. F., Shippee, T. P., & Schafer, M. H. (2009). Cumulative inequality theory for research on aging and the life course. In V. L. Bengston, D. Gans, N. M. Putney, & M. Silverstein (Eds.), Handbook of theories of aging (2nd ed., pp. 413–433). Springer Publishing Co. [Google Scholar]
  13. Friedman, E. M., & Mare, R. D. (2014). The schooling of offspring and the survival of parents. Demography, 51(4), 1271–1293. doi: 10.1007/s13524-014-0303-z [DOI] [PubMed] [Google Scholar]
  14. Gilligan, M., Karraker, A., & Jasper, A. (2018). Linked lives and cumulative inequality: A multigenerational family life course framework. Journal of Family Theory & Review, 10(1), 111–125. doi: 10.1111/jftr.12244 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Goldman, A. W. (2019). Linked lives in double jeopardy: Child incarceration and maternal health at midlife. Journal of Health and Social Behavior, 60(4), 398–415. doi: 10.1177/0022146519882328 [DOI] [PubMed] [Google Scholar]
  16. Gonzalez-Pienda, J. A., Nunez, J. C., Gonzalez-Pumariega, S., Alvarez, L., Roces, C., & Garcia, M. (2002). A structural equation model of parental involvement, motivational and aptitudinal characteristics, and academic achievement. The Journal of Experimental Education, 70(3), 257–287. doi: 10.1080/00220970209599509 [DOI] [Google Scholar]
  17. Greenfield, E. A., Akincigil, A., & Moorman, S. M. (2019). Is college completion associated with better cognition in later life for people who are the least, or most, likely to obtain a bachelor’s degree? The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 75(6), 1286–1291. doi: 10.1093/geronb/gbz132 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Greenfield, E. A., & Marks, N. F. (2006). Linked lives: Adult children’s problems and their parents’ psychological and relational well-being. Journal of Marriage and the Family, 68(2), 442–454. doi: 10.1111/j.1741-3737.2006.00263.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Guo, S. & Fraser, M. W. (2015). Propensity score analysis: Statistical methods and applications (2nd ed.). Sage Publications, Inc. [Google Scholar]
  20. Harris, K. M., & Hotz, V. J. (2020). National Longitudinal Study of Adolescent to Adult Health (Add Health) Parent Study: Public use, [United States], 2015–2017 [machine-readable data file and documentation]. Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel Hill. [Google Scholar]
  21. Hayward, M. D., & Gorman, B. K. (2004). The long arm of childhood: The influence of early-life social conditions on men’s mortality. Demography, 41(1), 87–107. doi: 10.1353/dem.2004.0005 [DOI] [PubMed] [Google Scholar]
  22. Hitlin, S., & Johnson, M. (2015). Reconceptualizing agency within the life course: The power of looking ahead. American Journal of Sociology, 120, 1429. doi: 10.1073/pnas.1907258117 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Jackson, M., & Holzman, B. (2020). A century of educational inequality in the United States. Proceedings of the National Academy of Sciences of the United States of America, 117(32), 19108–19115. doi: 10.1073/pnas.1907258117 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kail, B. L., & Carr, D. C. (2017). Successful aging in the context of the disablement process: Working and volunteering as moderators on the association between chronic conditions and subsequent functional limitations. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 72(2), 340–350. doi: 10.1093/geronb/gbw060 [DOI] [PubMed] [Google Scholar]
  25. Lee, C. (2018). Adult children’s education and physiological dysregulation among older parents. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 73(6), 1143–1154. doi: 10.1093/geronb/gbx044 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Link, B. G., & Phelan, J. (1995). Social conditions as fundamental causes of disease. Journal of Health and Social Behavior, 80–94. doi: 10.2307/2626958 [DOI] [PubMed] [Google Scholar]
  27. Lundborg, P., & Majlesi, K. (2018). Intergenerational transmission of human capital: Is it a one-way street? Journal of Health Economics, 57, 206–220. doi: 10.1016/j.jhealeco.2017.12.001 [DOI] [PubMed] [Google Scholar]
  28. Ma, M. (2019). Does children’s education matter for parents’ health and cognition? Evidence from China. Journal of Health Economics, 66, 222–240. doi: 10.1016/j.jhealeco.2019.06.004 [DOI] [PubMed] [Google Scholar]
  29. Maddox, G. L., & Douglass, E. B. (1973). Self-assessment of health: A longitudinal study of elderly subjects. Journal of Health and Social Behavior, 14(1), 87–93. doi: 10.2307/2136940 [DOI] [PubMed] [Google Scholar]
  30. Mao, W., Silverstein, M., Prindle, J. J., & Chi, I. (2020). The reciprocal relationship between instrumental support from children and self-rated health among older adults over time in rural China. Journal of Aging and Health, 32(10), 1528–1537. doi: 10.1177/0898264320943759 [DOI] [PubMed] [Google Scholar]
  31. McGarry, K., & Schoeni, R. F. (1995). Transfer Behavior in the Health and Retirement Study: Measurement and the redistribution of resources within the family. The Journal of Human Resources, 30, S184–S226. doi: 10.1177/0898264320943759 [DOI] [Google Scholar]
  32. McLanahan, S. (2004). Diverging destinies: How children are faring under the second demographic transition. Demography, 41(4), 607–627. doi: 10.1353/dem.2004.0033 [DOI] [PubMed] [Google Scholar]
  33. McLanahan, S., & Percheski, C. (2008). Family structure and the reproduction of inequalities. Annual Review of Sociology, 34(1), 257–276. doi: 10.1146/annurev.soc.34.040507.134549 [DOI] [Google Scholar]
  34. Mirowsky, J., & Ross, C. E. (2015). Education, health, and the default American lifestyle. Journal of Health and Social Behavior, 56(3), 297–306. doi: 10.1177/0022146515594814 [DOI] [PubMed] [Google Scholar]
  35. OECD. (2020). Education at a glance 2020: OECD indicators. Author. doi: 10.1787/69096873-en [DOI] [Google Scholar]
  36. Oi, K., Hotz, J. V., Harris, K., Zhang, L. (2018). The Add Health Parent Study: A preliminary statistical portrait (conference presentation abstract). PAA 2018 Convention, Denver, CO. https://paa.confex.com/paa/2018/meetingapp.cgi/Paper/21184 [Google Scholar]
  37. Porter, L., & Vogel, M. (2014). Residential mobility and delinquency revisited: Causation or selection? Journal of Quantitative Criminology, 30, 187–214. doi: 10.1007/s10940-013-9200-7 [DOI] [Google Scholar]
  38. Preston, S. H. (1984). Children and the elderly: Divergent paths for America’s dependents. Demography, 21(4), 435–457. doi: 10.2307/2060909 [DOI] [PubMed] [Google Scholar]
  39. Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1(3), 385–401. doi: 10.1177/014662167700100306 [DOI] [Google Scholar]
  40. Rosenbaum, P. R. (2002). Observational studies. Springer. [Google Scholar]
  41. Ross, C. E., & Wu, C. L. (1996). Education, age, and the cumulative advantage in health. Journal of Health and Social Behavior, 37(1), 104–120. doi: 10.2307/2137234 [DOI] [PubMed] [Google Scholar]
  42. Seidel, A. J., Yorgason, J. B., Polenick, C. A., Zarit, S. H., & Fingerman, K. L. (2018). Are you sleeping? Dyadic associations of support, stress, and worries regarding adult children on sleep. The Gerontologist, 58(2), 341–352. doi: 10.1093/geront/gnw149 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Shuey, K. M., & Willson, A. E. (2008). Cumulative disadvantage and Black–White disparities in life-course health trajectories. Research on Aging, 30(2), 200–225. doi: 10.1177/0164027507311151 [DOI] [Google Scholar]
  44. Smith-Greenaway, E., Brauner-Otto, S., & Axinn, W. (2018). Offspring education and parental mortality: Evidence from South Asia. Social Science Research, 76, 157–168. doi: 10.1016/j.ssresearch.2018.07.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Stevens, M. L., Armstrong, E. A., & Arum, R. (2008). Sieve, incubator, temple, hub: Empirical and theoretical advances in the sociology of higher education. Annual Review of Sociology, 34(1), 127–151. doi: 10.1146/annurev.soc.34.040507.134737 [DOI] [Google Scholar]
  46. Thoits, P. A. (2011). Mechanisms linking social ties and support to physical and mental health. Journal of Health and Social Behavior, 52(2), 145–161. doi: 10.1177/0022146510395592 [DOI] [PubMed] [Google Scholar]
  47. Torssander, J. (2013). From child to parent? The significance of children’s education for their parents’ longevity. Demography, 50(2), 637–659. doi: 10.1007/s13524-012-0155-3 [DOI] [PubMed] [Google Scholar]
  48. Umberson, D., & Thomeer, M. B. (2020). Family matters: Research on family ties and health, 2010 to 2020. Journal of Marriage and Family, 82(1), 404–419. doi: 10.1111/jomf.12640 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Weden, M. A., Carpiano, R. A., & Robert, S. A. (2008). Subjective and objective neighborhood characteristics and adult health. Social Science & Medicine, 66(6), 1256–1270. doi: 10.1016/j.socscimed.2007.11.041 [DOI] [PubMed] [Google Scholar]
  50. Wolfe, J. D., Bauldry, S., Hardy, M. A., & Pavalko, E. K. (2018a). Multigenerational attainments, race, and mortality risk among silent generation women. Journal of Health and Social Behavior, 59(3), 335–351. doi: 10.1177/0022146518784596 [DOI] [PubMed] [Google Scholar]
  51. Wolfe, J. D., Bauldry, S., Hardy, M. A., & Pavalko, E. K. (2018b). Multigenerational socioeconomic attainments and mortality among older men: An adjacent generations approach. Demographic Research, 39, 719–752. doi: 10.4054/DemRes.2018.39.26 [DOI] [Google Scholar]
  52. Xie, Y., Brand, J. E., & Jann, B. (2012). Estimating heterogeneous treatment effects with observational data. Sociological Methodology, 42(1), 314–347. doi: 10.1177/0081175012452652 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Yahirun, J. J., Sheehan, C. M., & Mossakowski, K. N. (2020). Depression in later life: The role of adult children’s college education for older parents’ mental health in the United States. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 75(2), 389–402. doi: 10.1093/geronb/gby135 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Yahirun, J. J., Vasireddy, S., & Hayward, M. D. (2020). The education of multiple family members and the life-course pathways to cognitive impairment. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 75(7), e113–e128. doi: 10.1093/geronb/gbaa039 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Zajacova, A., & Lawrence, E. M. (2018). The relationship between education and health: Reducing disparities through a contextual approach. Annual Review of Public Health, 39, 273–289. doi: 10.1146/annurev-publhealth-031816-044628 [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Zimmer, Z., Hanson, H. A., & Smith, K. R. (2016). Offspring socioeconomic status and parent mortality within a historical population. Demography, 53(5), 1583–1603. doi: 10.1007/s13524-016-0502-x [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gbab109_suppl_Supplementary_Materials
Click here for additional data file. (199.8KB, pdf)

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press

RESOURCES