BLACK-WHITE DIFFERENCES IN OFFSPRING EDUCATIONAL ATTAINMENT AND OLDER PARENTS’ DEMENTIA

Abstract

Emerging research documents the health benefits of having highly educated adult offspring. Yet less is known about whether those advantages vary across racial groups. This study examines how offspring education is tied to parents’ dementia risk for Black and White parents in the United States. Using data from the Health and Retirement Study, findings suggest that children’s education does not account for the Black-White gap in dementia risk. However, results confirm that parental race moderates the relationship between children’s education and dementia risk and that the association between children’s education and parents’ dementia risk is strongest among less-educated parents. Among less-educated parents, higher levels of children’s attainment prevent the risk of dementia onset for Black parents. Conversely, low levels of offspring schooling increase dementia risk among White parents. The study highlights how offspring education shapes the cognitive health of social groups differently and points to new avenues for future research.

A growing body of research documents how parents with well-resourced adult children fare better on a variety of physical, mental, and cognitive health outcomes compared to parents of offspring with fewer socioeconomic resources. This association has been reported across numerous contexts, including China, Mexico, Taiwan, South Africa, Sweden, and the United States. Most of this work focuses on adult children’s education as an established marker of socioeconomic status (Friedman and Mare 2014; Torssander 2013; Wolfe et al. 2018; Zimmer et al. 2007), whereby parental health is positively influenced by the knowledge, social support, and health spillovers that result from children’s education (see De Neve & Kawachi, 2017 for a review).

Whether the salubrious effects of offspring education are shared across individuals of different demographic backgrounds is less clear. At the individual level, evidence points to how the educational gradient in health may not be equal across racial groups (Hummer and Hernandez, 2013). Specifically, the “payoff” to a higher education may be weaker for Black Americans, in part due to the stress and discrimination that stem from living with institutional racism, and general “weathering” that Black Americans face, a process which may be particularly acute for upwardly mobile, or highly-educated Black Americans (Geronimus et al. 2006; Hudson et al. 2013). Prior research finds mixed results. Whereas some studies find a stronger effect of children’s education on the mental health of Black versus White parents (Yahirun, Sheehan, and Mossakowski, 2022), others find that the benefits of offspring education on overall longevity are similar for Black and White mothers (Wolfe et al. 2018).

Yet little to no research has examined whether racial differences in the relationship between offspring education and parental health extends to older adults’ cognitive health. With broader gains in longevity, however, the number of older Americans living with cognitive disorders, including dementia, is growing. The racial disparities in cognitive health suggest severe disadvantages for older Black compared to White adults, with older Black adults 2–3 times more likely to experience symptoms of dementia than White adults in later life (Farina et al. 2020; Hayward et al. 2021). In addition, although the share of older adults with dementia is expected to decrease with the educational “upgrading” of younger birth cohorts, this decline is not uniform for Black and White Americans, with educational differences across racial groups contributing to this disparity (Farina et al. 2020; Hayward et al. 2021). Even though prior research highlights how having highly-educated children benefits older parents’ cognitive health (Ma et al. 2021; Torres et al. 2021; Yahirun, Vasireddy, and Hayward 2020), it remains unclear whether offspring education may help to close – or perhaps exacerbate – the racial gap in dementia and whether the advantages of children’s education are similar for Black versus White parents.

BACKGROUND

The life course origins of racial disparities in cognitive health

In the United States, race stratifies the health outcomes of Americans, with systemic racism acting as a fundamental cause of disease (Phelan and Link 2015). The life course origins of cognitive health, including dementia, point to the importance of considering how systemic and institutionalized racism created vastly divergent socioeconomic conditions in early childhood for older Black compared to White adults today. Black older adults have historically, and continue to be, much more likely to be raised under conditions of material deprivation than their White counterparts – including greater exposure to childhood poverty, lower quality housing, segregated neighborhoods, and poorer nutrition. The vast majority of older Black adults living today also grew up in the South under Jim Crow laws (Glymour and Manly 2008; Krieger et al. 2014; Zhang, Hayward, and Yu 2016), a codified system of statutes and laws that ensured second-class citizenship for Black families through the restriction of social and economic rights. Scholars have described, in great detail, the deleterious health consequences of living through this period (Glymour and Manly 2008; Krieger et al. 2014). However, as prior research shows, accounting for early childhood conditions, including the region of birth, parental schooling, and Southern birth, only partially accounts for the Black-White difference in cognitive impairment (Zhang et al. 2016).

Early childhood conditions affect cognitive health in multiple ways. For instance, exposure to enriching social environments – either through the home or school – is important for developing a stock of cognitive abilities, often referred to as cognitive reserve, that preserves individuals’ functioning in the face of aging and disease later in life (Stern 2002). Formal education is especially critical for cognitive health outcomes at least in part because of the establishment of cognitive abilities early in life. Schooling provides individuals with new knowledge, ways of keeping that knowledge in mind, and responding to new and different tasks in a highly dynamic environment. Students are immersed in a cognitively challenging environment during school. Education also happens when, biologically speaking, brain and cognitive development occur rapidly. A recent meta-analysis reported a robust, dose-response relationship between educational attainment and dementia risk throughout education’s distribution (Xu et al. 2016). One reason why cognitive health outcomes tend to be worse for Black adults stems from their lower average levels of education compared to White adults, a result of long-standing institutional discrimination and historical exclusion from educational institutions based on race (Hayward et al. 2021; Zhang et al. 2016). Whereas educational “upgrading” among Whites, specifically the increasing share of Whites with a college education, may explain lower levels of dementia prevalence among younger compared to older cohorts, it may also explain why Black Americans still experience worse cognitive health, given that Black post-secondary education rates continue to lag behind Whites (Hayward et al., 2021; U.S. Census Bureau 2020).

The cumulative effect of advantageous social conditions is another distinct, but complementary, pathway through which early childhood conditions influence brain health in later life. Here, children raised in enriched social environments go on to attain higher levels of educational attainment overall, which then leads to jobs with more stability, greater autonomy, less physical strain, and higher earnings and wealth, all of which are positively related to cognitive health. However, prior research shows that accounting for mid-life conditions such as an adult’s own education and household wealth only accounts for some, but not all of the Black-White racial gap in cognitive impairment (Zhang et al. 2016). In sum, a life course perspective illuminates the multiple ways in which historic systems of racial inequality stratify the early and mid-life antecedents of the Black-White disparity in cognitive health. And yet, gains in education and greater upward mobility for Black Americans in the latter half of the 20^th century point to the potential role of offspring education as a resource that may be especially important for older Black adults who, through institutionalized racism, have not be able to attain comparable levels of education as Whites (Hertz 2005; Pfeffer and Hertel 2015). However, it is unclear what role, if any, the resources of children, a distinct resource in later life, may play in influencing racial disparities in cognitive health.

Racial disparities in educational attainment and parents’ cognitive health

Although a growing body of evidence documents the importance of offspring education for parents’ cognitive health ( Lee 2018; Ma 2019; Ma et al. 2021; Torres et al. 2021), only a handful of studies have explored the role of race. For example, prior research finds that race moderates the association between having college-educated offspring and mental health, with a steeper gradient in depressive symptoms by children’s education among Black, compared to White older parents (Yahirun, Sheehan, and Mossakowski 2022). However, other studies found that offspring schooling is a robust predictor for both Black and White women’s mortality, among older cohorts of women born between 1923 and 1937 (Wolfe et al. 2018). Yet an assessment of how offspring resources matter for the Black-White gap in cognitive functioning may be particularly relevant given that dementia risk is closely tied to health behaviors and chronic conditions (Livingston et al. 2020), pathways that are likely influenced by children’s resources (Friedman and Mare 2014; Torres et al. 2022).

Furthermore, increasing rates of college completion, greater earnings, and growing rates of homeownership among younger Black cohorts coming of age in the post-civil rights era warrants an examination of the way in which offspring resources matter for parents’ cognitive health. Despite these gains, stratification researchers note that even as the racial gap in upward educational mobility (e.g., achieving higher levels of education vis-à-vis parents) has narrowed from older to younger cohorts of Americans, racial disparities in mobility remain. Specifically, studies underscore the greater “stickiness” of social class status among Black Americans at the bottom end of the socioeconomic distribution compared to Whites. That is, Black Americans born into low-income families are less likely to be upwardly mobility compared to similarly disadvantaged White Americans (Chetty et al. 2020; Hertz 2005; Torche 2017). Likewise, Black individuals raised in high socioeconomic status households also face a greater odds of downward mobility compared to White peers, suggesting a more precarious transfer of educational advantages across generations in Black families (Chetty et al. 2020; Hertz 2005; Torche 2017).

These trends should be considered in light of the ample work indicating that when children experience personal, familial, and financial setbacks, parental health suffers. In particular, parents whose adult children are incarcerated, divorced, unemployed, do not complete schooling on time, or become gravely ill, report worse mental health (Barr et al. 2018; Kalmijn and Graaf 2012; Smith, 2022; Tosi and Albertini 2019). From the perspective of the older generation, parents whose children report personal or financial hardships may funnel resources to disadvantaged children, resources that could be used to protect parents’ own wellbeing and safeguard against future financial or health shocks. A child’s socioeconomic disadvantage may also lead to psycho-social stressors, which in turn negatively impact cognitive health (Gatchel et al. 2019). On the other hand, having access to highly educated children could help parents navigate complex health information. In addition, parents with highly educated children may also benefit from the greater financial resources that children provide, or simply from the security of knowing that children will not present an economic burden to parents in the future. Children who achieve more years of schooling, attain higher incomes, and enter into more prestigious professions may also be a source of parental pride (Yahirun, Sheehan, and Mossakowski 2020).

Thus, offspring education plays an integral role in shaping parental cognitive health ( Lee 2018; Ma 2019; Ma et al. 2021; Torres et al. 2021) and this may be especially true for parents with fewer educational resources themselves. Any number of pathways through which children’s educational attainments shape parental health could be amplified or dampened by parents’ own levels of schooling. For example, the challenges of children’s educational deficits could be even more detrimental to parental health when parents have low levels of education, who will now need to tap personal or financial reserves, compared to parents with more resources. Conversely, a child’s schooling deficits may be less determinative of parental health among parents with ample access to their own capital. Although prior studies find mixed results with regards to how parental education might moderate the link between children’s education and parental health outcomes (Friedman and Mare, 2014; Yahirun, Sheehan, and Hayward 2016), we know even less about whether these moderating effects are similar across racial groups.

PRESENT STUDY

This study builds upon growing research on the potential benefits of children’s resources for later life health (Friedman and Mare 2014; Torssander 2013; Zimmer, Hermalin, and Lin 2002) and recent evidence of a potential difference in the efficacy of children’s education by parental race (Yahirun et al. 2022). Moreover, this study connects to the broader literature on Black-White disparities in cognitive health (Farina et al. 2020; Hayward et al. 2021; Zhang et al. 2016) by explicitly examining whether exposure to offspring education minimizes – or exacerbates - the Black-White gap in parents’ dementia risk and whether the association between children’s educational outcomes and parents’ cognitive health varies for Black versus White parents. Using prior research to guide our hypotheses, we predict that:

Hypothesis 1: The Black-White gap in parents’ dementia risk is attenuated with the inclusion of children’s educational levels.

Hypothesis 2: Parental race moderates the association between children’s education and parental dementia risk such that Black parents are thought to be more sensitive to the educational outcomes of children than White parents.

Hypothesis 3: Parental education moderates the association between children’s education and parental dementia risk and the moderation effects will be greater for Black compared to White parents.

DATA AND METHODS

Analytical Sample

This study draws on the U.S. Health and Retirement Study (HRS), a biennial study of adults over 50 that began in 1992. We use data from the RAND HRS Family Data 2014 (v1), and the RAND HRS Longitudinal File 2018 (v1). The HRS offers a number of advantages for this study, because few datasets with rich socio-demographic and health characteristics taken at the respondent level also provide information on the respondent’s children, regardless of where children live. In the United States, the HRS is the only dataset that provides detailed, parent-child data over a significant amount of time.

Our analytical sample includes data on HRS respondents interviewed in 2000–2014 as the data are only available until 2014 in the RAND HRS Family Data 2014 file. Although basic demographic information on children is updated by a designated family respondent periodically, not every question regarding each child is asked at each wave. Questions on education and marital status were skipped occasionally. However, in 2000 (Wave 5), respondents were asked detailed questions regarding all of their children’s educational levels as well as other characteristics of their children. In addition, we include only the sample waves from 2000 or later because consistent cognitive information for both community-dwelling and nursing home residents first became available in 2000. We limit our sample to respondents over the age of 65 given that dementia incidence prior to age 65 remains low. Similar to previous studies (Friedman and Mare 2014), we include offspring in the sample to adult children aged 25 or older with the assumption that most children will have completed their formal education by age 25. However, offspring aged 24 or younger are allowed to “age into” the sample when they reach age 25.

Our study includes proxy respondents, persons who are unable to complete the survey. We include proxy respondents because respondents who are unable to complete the survey are more likely to be cognitively impaired (Crimmins et al. 2011). Frequently, proxy respondents are excluded from population-level studies of cognitive health, leading to bias (David Weir, Jessica Faul, & Langa, 2011). Approximately 8.4% of our sample consists of proxy interviews.

We constructed our analytic sample from all respondent-observations in the RAND HRS Longitudinal File 2018 (v1) file that fell within our observation window of 2000–2014 when the respondent was over the age of 65, had at least one child aged 25 or older, and excluded respondents with any “bad links” to children (defined as those in which the parent-child relationship is evaluated by RAND as inconsistent across survey years), and those respondents with any missing values on the measures of cognitive impairment (n = 3,862 person-observations).

There was generally little missing data on either the respondent or offspring characteristics (4.08% of person-observations contained missing information). To address this, we first excluded person-observations with missing cognition scores, which were used to categorize persons with and without dementia (see below). Next, independent variables that were missing information (i.e., transfers from children) were imputed on the modal category. The final analytic sample includes 11,768 unweighted respondents at baseline. For the event history sample that examines the transition from no dementia to dementia, death, or attrition (see below), we removed respondents with dementia at baseline, thus leaving us with 10,120 unweighted respondents, comprising 46,451 person-observations.

Measures

Dementia:

The HRS provides several questions that assess cognitive health status. In this study, we distinguish between individuals with and without dementia. The classification scheme uses the Langa-Weir approach, based on the concordance of HRS cognitive functioning scores and diagnosis of dementia in a subset of individuals who completed a neuropsychological assessment in the Aging, Demographics, and Memory Study (ADAMS) (Crimmins et al. 2011). The HRS allows for an assessment of HRS participants with dementia among self-respondents and proxy respondents, although the measures used to assess dementia differ somewhat between the two groups. To measure cognitive health for respondents who provided self-reports, we use a summary score that includes items assessing immediate word and delayed word recall of ten words, five trials of serial 7’s, and backward counting. The score ranges from 0 to 27. Respondents are classified as having no dementia based on scores ranging from 7–27; using the traditional cutoff point or are classified as having dementia if they score 0–6 on the scale. To measure cognitive health for respondents who relied on proxy respondents, we use three measures. The first measure is based on the proxy respondent’s reports of challenges to five instrumental activities of daily living, including managing money, taking medication, preparing hot meals, using phones, and grocery shopping (Crimmins et al. 2018). Next, the proxy was also asked to provide an evaluation of the respondent’s general memory (0 excellent, 1 very good, 2 good, 3 fair, 4 poor). The last measure included the interviewer’s assessment of respondent difficulty in completing the interview due to perceived cognitive limitations (none 0, some 1, prevents completion 2). These three measures are then summed into a scale (0–11) where they are then classified into the two categories of no dementia (0–5) and having dementia (6–11). This classification scheme has been applied in recent studies of dementia using the HRS (Crimmins et al. 2018; Farina et al. 2020; Hayward et al. 2021).

Education:

Our main measure of interest is parental exposure to offspring education. Educational expansion at the societal level means that the value of certain educational credentials may decrease as a greater share of the population acquires them (Breen et al. 2009). To address this, we use measures of both parents and children’s education relative to the median levels of education in the parent’s or child’s birth cohort accordingly (measured in 10-year intervals). Children and their parents either achieve below the median, at the median, or above the median levels of education compared to their similarly-aged peers. We assess the educational attainment of the respondent’s most-educated child, although we also test the robustness of this measure, which is described in the Discussion section.

Race:

We use respondents’ reports of racial and ethnic self-identification in the HRS. We distinguish between respondents who self-identify as Black and who do not identify as Hispanic (Non-Hispanic Black, abbreviated as Black in this study) and those who self-identify as White and who also do not identify as Hispanic (Non-Hispanic White, abbreviated as White here). We do not include Hispanic respondents in our study given the paper’s focus on understanding the Black-White gap in dementia and the historical and systemic forces that created this disparity. However, we recognize that future work should take into account the growing number of Hispanic and Asian older adults, many of whom are foreign born and whose offspring may be especially influential for parental health (Ram et al. 2022).

Key Controls:

We control for key sociodemographic factors shown to be associated with cognitive health, including gender, marital status, and number of children (Saenz, Díaz-Venegas, and Crimmins 2019; Zhang et al. 2016). We also control for respondent early life factors to have a clearer assessment of our key hypotheses. In this study, we account for parental education, childhood health, and Southern birth in our models. Finally, the last set of control measures includes information on respondent’s other children given that the support of older parents in later life should be framed within a broader familial context (Utz, Berg, and Butner 2017).

Analytical Plan

We use discrete-time event history models to predict the hazard, or “risk” that the respondent (parent) experiences dementia within an observation interval, given that the respondent does not have dementia at the beginning of the interval. Discrete-time event history models are appropriate for the data used here for several reasons (Allison, 1982). First, the data are right censored, such that many respondents do not experience symptoms of dementia during the last year that they are observed in the sample. Second, the HRS observation interval is approximately two years in length given that the data are only collected every other year, thus suggesting that discrete-time, rather than continuous-time models are appropriate. Age in our models captures the changes in the risk associated with aging.

Our sample is restricted to parents who report no dementia at baseline. We use multinomial logistic regression models to assess the risk of dementia onset in any given observation interval while also incorporating possible competing risks such as death and attrition from the sample (Allison 1982). The risk of dementia starts at the age at baseline (age 66 or older), and respondents are followed until they have symptoms of dementia, or they are censored through the last interview date or death. All events are treated as absorbing states. Variables are allowed to vary over time, including our main measure of offspring education, respondent age, marital status, and number of children, to name a few. Time-varying variables are measured at the beginning of the observation interval and are not contemporaneous with the outcome. We account for clustering at the household level in our models. This method has been used extensively in prior work (Yahirun, Vasireddy, et al. 2020; Zhang et al. 2016). Model results are first presented using beta coefficients from the logistic regression models in Tables 2 and 3, but we then present risk probabilities using the Stata -margins-command for easier interpretation of the final models in Figures 1 and 2.

Table 2.

Results from Event History Logistic Regression Models Predicting Dementia Onset, Adults aged over 65 (N = 46,451 person-observations)

	M1	M2	M3
Respondenťs characteristics
Black (/White)	0.897*** [0.073]	0.779*** [0.079]	0.777*** [0.080]
Respondenťs Education (/Median)
Above Median	−0.456*** [0.072]	−0.440*** [0.073]	−0.363*** [0.075]
Below Median	0.765*** [0.067]	0.727*** [0.067]	0.672*** [0.067]
Female	0.087 [0.063]	0.076 [0.063]	0.069 [0.063]
Age	0.093*** [0.004]	0.093*** [0.004]	0.096*** [0.004]
Marital Status (/Married)
Separated/Divorced	0.125 [0.113]	0.122 [0.113]	0.071 [0.114]
Widowed	0.153* [0.070]	0.146* [0.069]	0.118 [0.070]
Never married	−0.178 [0.348]	−0.147 [0.346]	−0.222 [0.347]
Number of Kids	0.009 [0.014]	0.011 [0.014]	0.015 [0.014]
Proxy status	0.708*** [0.087]	0.690*** [0.087]	0.679*** [0.088]
Respondent early life conditions
Parenťs education (<High school)
High school		−0.083 [0.091]	−0.08 [0.091]
Some College and above		−0.107 [0.119]	−0.088 [0.119]
Poor childhood health (/good)		0.204 [0.104]	0.203 [0.104]
Southern birth (/non-Southern)		0.227*** [0.063]	0.211*** [0.063]
Characteristics of offspring
Child's Education (/Median)
Above Median			−0.027 [0.075]
Below Median			0.289*** [0.069]
Received transfers from children (/None)			0.034 [0.145]
1+ child cores/lives w/in 10miles (/None)			0.102 [0.061]
Race X Kid's Education
Black X Above Median
Black X Below Median
Constant	−11.545***	−11.529***	−11.903***
	[0.342]	[0.363]	[0.374]
N	46,451	46,451	46,451

Source: HRS, 2000–14

Table 3.

Effects of Schooling of Respondent, Respondenťs Parent, and Respondenťs Children on Dementia Onset, by Race

	Black respondents only				White respondents only
	Outcome: Dementia Onset				Outcome: Dementia Onset
	Model 1	Model 2	Model 3	Model 4	Model 1		Model 2			Model 3	Model 4
Respondenťs characteristics
Respondenťs Education (/Median)
Above Median	−0.746*** [0.206]	−0.773*** [0.204]	−0.680** [0.210]	−0.848* [0.361]	−0.420*** [0.077]		−0.400*** [0.078]			−0.322*** [0.080]	−0.465** [0.144]
Below Median	0.845*** [0.146]	0.843*** [0.146]	0.828*** [0.147]	1.025*** [0.269]	0.732*** [0.075]		0.689*** [0.076]			0.627*** [0.077]	0.444** [0.154]
Female	0.125 [0.139]	0.112 [0.139]	0.112 [0.139]	0.109 [0.142]	0.077 [0.071]		0.066 [0.071]			0.057 [0.071]	0.057 [0.071]
Age	0.063*** [0.009]	0.065*** [0.009]	0.066*** [0.009]	0.067*** [0.009]	0.101*** [0.005]		0.101*** [0.005]			0.104*** [0.005]	0.104*** [0.005]
Marital Status (/Married)
Separated/Divorced	0.048 [0.200]	0.067 [0.201]	0.029 [0.204]	0.028 [0.206]	0.138 [0.139]		0.133 [0.138]			0.075 [0.139]	0.072 [0.139]
Widowed	0.111 [0.150]	0.123 [0.151]	0.093 [0.153]	0.084 [0.154]	0.141 [0.079]		0.133 [0.079]			0.103 [0.079]	0.104 [0.079]
Never married	−0.358 [0.397]	−0.305 [0.392]	−0.352 [0.394]	−0.367 [0.392]	0.238 [0.779]		0.271 [0.775]			0.23 [0.771]	0.232 [0.771]
Number of Kids	0.035 [0.025]	0.034 [0.025]	0.037 [0.025]	0.037 [0.026]	−0.007 [0.017]		−0.003 [0.017]			0.001 [0.017]	0.000 [0.017]
Proxy status	0.387* [0.184]	0.380* [0.184]	0.366* [0.186]	0.347 [0.186]	0.809*** [0.096]		0.789*** [0.096]			0.776*** [0.096]	0.775*** [0.097]
Respondent early life conditions
Parenťs education (<High school)
High school		0.218 [0.216]	0.2 [0.217]	0.2 [0.215]			−0.12 [0.101]			−0.112 [0.101]	−0.112 [0.101]
Some College and above		0.469 [0.302]	0.483 [0.302]	0.503 [0.308]			−0.158 [0.130]			−0.134 [0.130]	−0.134 [0.131]
Poor childhood health (/good)		0.108 [0.218]	0.097 [0.218]	0.087 [0.221]			0.222 [0.121]			0.223 [0.120]	0.224 [0.120]
Southern birth (/non-Southern)		0.446* [0.182]	0.435* [0.182]	0.459* [0.184]			0.205** [0.069]			0.188** [0.069]	0.188** [0.069]
Characteristics of offspring
Child's Education (/Median)
Above Median			−0.314 [0.170]	0.299 [0.303]						0.036 [0.084]	−0.081 [0.139]
Below Median			0.028 [0.138]	−0.05 [0.282]						0.364*** [0.079]	0.227 [0.122]
Received transfers from children (/None)			−0.084 [0.258]	−0.085 [0.258]						0.107 [0.174]	0.109 [0.174]
1+ child cores/lives w/in 10miles (/None)			0.024 [0.146]	0.029 [0.145]						0.114 [0.067]	0.113 [0.068]
Interaction between child X parent education
Respondenťs Education X Kid's Education (/Median)
Above Median X Above Median				−0.356 [0.508]								0.194 [0.191]
Above Median X Below Median				0.608 [0.507]								0.193 [0.206]
Below Median X Above Median				−1.047** [0.390]								0.188 [0.226]
Below Median X Below Median				−0.038 [0.334]								0.263 [0.182]
Constant	−7.55*** [0.677]	−8.12*** [0.709]	−8.17*** [0.729]	−8.31*** [0.759]		−11.25*** [0.369]		−11.26*** [0.402]	−11.73*** [0.414]			−11.65*** [0.418]
N	5,204	5,204	5,204	5,204		41,247		41,247	41,247			41,247

Source: HRS, 2000–2014

Figure 1:

Predicted Probability of Dementia Risk, Black Parents aged over 65

Source: HRS, 2000–14

Figure 2:

Predicted Probability of Dementia Risk, White Parents aged over 65

Source: HRS, 2000–14

RESULTS

Tables 1A and 1B provide an overview of our analytical sample, stratified by respondent race. At first glance, the racial disparity in dementia prevalence is evident. In Table 1A, 11.7% of Black respondents have dementia at baseline, compared to 3.5% of White respondents in Table 1B. Women constitute a greater share of the Black sample, compared to Whites, reflecting differences in early mortality among Black men. With respect to marital status, fewer Black respondents are married/partnered, compared to Whites, and more than one-third of Black parents are widowed, compared to less than one-fourth of White parents in the sample. Racial disparities in the educational attainments of parents and children reflect a long history of systemic racism towards African Americans. Half of all older Black parents in the sample have lower than the median years of schooling compared to their similarly aged peers. The educational norm for White parents, comparatively, is to have completed at least the median level of schooling, if not higher, compared to peers born in the same 10-year birth cohort. A stark illustration of racial disparities in educational advantage is that whereas 40% of White parents have more than the median years of schooling, only 23% of Black parents achieve more than the median. Assessing educational disparities among offspring shows a narrowing of racial gaps in education from older to more recent birth cohorts. Specifically, the gap between Black and White offspring with median or higher levels of education is smaller than the same gap between Black and White parents. Still, most of this is due to the smaller share of White offspring who achieve above median schooling compared to their parents (30.7% among children compared to 40.2% among parents), whereas the share of Black offspring with above median levels of schooling did not depart much from parents (19.6% among children compared to 23.0% among parents). A closer look at adult children’s education vis-à-vis parents (Appendix Table A1) shows that among children who achieved more than the average levels of schooling, Black offspring were more likely to have parents with the lowest levels of schooling (31.1%) compared to White parents (9.6%). However, also evident is the “stickiness” of social origins for highly-educated White families and less-educated Black families. For example, 60.3% of White children with above median years of schooling had a parent with above average education, whereas 59.5% of Black children with below median schooling had a parent with a below median education. Variation and “stickiness” in educational attainment across generations and racial groups underscores how an examination of the effects of children’s education on parental cognitive health is warranted.

Table 1A:

Descriptive Statistics at Baseline, Black Adults aged over 65

Percent/Mean		SE		Percent
Respondenťs characteristics			Offspring Education
Dementia at baseline	11.7		< Median	51.6
Age	70.9	0.17	Median	28.8
Birth cohort			> Median	19.6
1920 or earlier	15.6
1921–30	21.9
1931–40	46.2		Characteristics of offspring
1941 or later	16.3		Respondent received transfers	5.3
Women	62.6		Respondent proximity to children
Marital status			No child cores/lives w/in 10miles	25.4
Married/Partnered	46.2		1+ child cores/lives w/in 10miles	74.7
Separated/Divorced	16.8		Offspring birth cohort
Widowed	34.4		1945 or earlier	14.8
Never married	2.6		1946–55	23.9
Number of kids	4.1	0.07	1956–65	34.0
Education			1966 or later	27.4
< Median	50.1
Median	26.9
> Median	23.0
Respondent early life conditions
Parental Education
< High school	82.9
High school	12.6
Some college and above	4.6
Poor childhood health	7.9
Southern birth	81.3
Proxy status	12.3
N (persons)	1,597

Source: HRS, 2000–14

Note: Distributions weighted using person-level weights

Table 1B:

Descriptive Statistics at Baseline, White Adults aged over 65

Percent/Mean		SE		Percent
Respondenťs characteristics			Educational Trajectories
Dementia at baseline	3.5		< Median	38.8
Age	71.8	0.07	Median	30.5
Birth cohort			> Median	30.7
1920 and before	19.2
1921–30	29.4
1931–40	37.2		Characteristics of offspring
1941 and over	14.2		Respondent received transfers	1.7
Women	56.8		Respondent proximity to children
Marital status			No child cores/lives w/in 10miles	40.6
Married/Partnered	68.3		1+ child cores/lives w/in 10miles	59.4
Separated/Divorced	7.4		Offspring birth cohort
Widowed	24.2		1945 or earlier	17.4
Never married	0.1		1946–55	29.6
Number of kids	3.3		1956–65	31.5
Education			1966 or later	21.5
< Median	23.0
Median	36.8
> Median	40.2
Respondent early life conditions
Parental Education
< High school	64.2
High school	21.8
Some college and above	14.1
Poor childhood health	5.5
Southern birth	28.1
Proxy status	8.4
N (persons)	10,171

Source: HRS, 2000–14

Note: Distributions weighted using person-level weights

Table 2 presents results from the event history analysis predicting the risk of dementia for the full sample of Black and White respondents among those who did not have dementia at baseline. The model results for other outcomes, i.e., the competing risks of death or attrition, are provided in the Appendix (Appendix Table A2). Model 1 includes respondent race and education, and controls for other characteristics, including gender, age, marital status, number of children and whether the respondent completed a proxy interview. As is already established, Black respondents in the sample have a higher risk of dementia than White respondents (b=0.897, p-value < .001). When exponentiated, the odds of dementia are nearly 2.5 times higher for Black parents in the sample compared to White parents. With respect to education, respondents who achieved more than the median years of schooling were nearly 40% less likely to experience dementia onset at any given survey year compared to those at the median (b=−0.456, p-value < .001), whereas those who received less than the median level of schooling for their peer group were more than twice as likely to experience dementia onset (b=0.765, p-value < .001). In Model 2, early life conditions are accounted for, including the respondent’s parents own education as well as childhood health and southern birth. Although the racial gap in dementia risk shrinks with the inclusion of these early life conditions, it remains significant, with Black parents still reporting dementia risk at more than twice that of White parents (b=0.779, p-value < .001). Finally, in Model 3, we see that the educational attainments of offspring are significantly associated with parents’ dementia risk. Compared to children with median levels of schooling, a child’s schooling deficit harms parents, such that having a child with less than the average educational attainment increases parental dementia onset by 33% (b=0.289, p-value < .001). Note, however, that the racial gap in dementia risk remains significant and only slightly attenuated from Model 2 (b=.777, p-value < .001).

In separate analyses (not shown here), we tested an interaction for respondent race and family educational trajectories and found that the interaction was statistically significant (X²=198.55, p-value <.001). However, Table 3 presents the full model results from Table 2 separately for Black and White parents to focus on how parental and children’s education, as well as other social and demographic characteristics, potentially affect the risk of dementia differently for both groups of parents. In Model 1, we note that education, age, and proxy interview status are significant predictors of dementia risk across racial groups. For Black and White parents, higher levels of schooling protect against the risk of dementia onset, whereas having lower than the median level of education increases the risk of dementia. In Model 2, the inclusion of early life conditions does not alter the association between the significant sociodemographic predictors and dementia risk as found in Model 1 for either Black or White parents. Model 3 includes children’s education and other characteristics of the respondent’s offspring into the model. Apparent is that the detrimental effects of children’s educational deficits are only present for White, but not Black parents. In particular, White parents whose children attained less than the median years of schooling experience a 43% increase in the risk of dementia onset (b=0.364, p-value < .001). However, there is no comparable association for Black parents. Not shown in the models, though, is the contrast in dementia risk between having children with above median versus below median levels of schooling. For Black parents, the difference between having a child with above median education from having a child with less than the median years of schooling approaches statistical significance (results not shown, p-value <.10), but the difference is not statistically significant for White parents (p-value>.10). Finally, Model 4 adds the interaction between parents’ and offspring education. Wald tests (not shown here), show that the interaction itself is statistically significant (X²=132.72, p-value <.001 for Black parents, X²=263.82, p-value <.001 for White parents). However, very few of the coefficients in the model are significant, with the exception of the interaction between below and above median levels of schooling for Black respondents. Here, the coefficient indicates the protective effects of having highly-educated offspring for less-educated parents, compared to parents with median levels of schooling whose children also attain the median. To better assess the model results, Figures 1 and 2 present the risk of dementia onset by parental and offspring education, across racial groups.

Figure 1 demonstrates the predicted risk of dementia for Black parents by parent’s and child’s education based on the final model from Table 3. The figure shows that parents with below median levels of schooling have the highest risk of dementia. Considering the effects of children’s education, we see that offspring schooling does not shift the risk of dementia for parents at or above the median level of schooling themselves. However, noteworthy is that for less-educated parents, offspring education matters. Having a child with above-median levels of schooling substantially shifts the risk of dementia downwards in comparison to parents whose children achieve the median or fewer years of schooling, from a probability of 13% to 7%. This suggests that having children with more education offers protection for the least-educated Black parents. Figure 2 presents a parallel figure for Whites. Similar to Black parents, parents with below median schooling experience the highest risk of dementia onset, whereas parents with above median schooling have the lowest risk. Unlike Black parents, however, high educational attainment among children does not lower the risk of dementia for the least-educated parents. In this instance, however, having a child who achieves lower than the average level of schooling increases the risk of dementia relative to parents whose child achieves the median years of schooling.

A few notes should be made when comparing Black to White parents in the figures. First, examining across racial groups, regardless of respondent or child educational attainments, Black parents in the sample experience a continuously elevated risk of dementia than White peers with similar educational backgrounds and offspring schooling. Second, the racial gap in dementia is largest for parents whose own education remains below the median for their age cohorts. Conversely, the racial gap is the smallest among parents with educational advantages. Third, assessing within-racial group differences reveals a flatter gradient between dementia risk and educational attainment for White parents, compared to Black parents, findings that parallel recent research on older adults more generally (Farina et al. 2020). Whereas the risk of dementia varies between 2% and 6% for the most to least educated White parents, the risk interval is much broader for Black parents, between 4% and 13% for the most- and least-educated. When adding in children’s education, the gradient does not change much for White parents, but is altered substantially for Black parents, suggesting the stronger effect of children’s resources for Black Americans.

DISCUSSION

In this study, we build upon prior work to consider how offspring education shapes the risk of parental dementia (Lee 2018; Ma 2019; Ma et al. 2021b; Torres et al. 2021; Yahirun, Vasireddy, et al. 2020). Given substantial racial differences in rates of education, social mobility and racial/ethnic differences in the prevalence of dementia (Garcia et al. 2019), our study asks how offspring education contributes to dementia risk across racial groups and whether this varies by parent’s own education. Returning to our original hypotheses, we find that the inclusion of children’s educational attainment does indeed attenuate the Black-White disparity in dementia risk, albeit only slightly. Even after accounting for children’s education, Black parents in the sample were still significantly more likely to experience a higher risk of dementia at any given observation period compared to White parents.

Second, our results indicate that race plays a moderating role in the association between children’s educational attainments and parental dementia risk. Yet our race-stratified analysis detected some unanticipated results. For White parents, having children who experienced educational deficits slightly increased the risk of dementia compared to those whose children had achieved median levels of schooling, whereas this detrimental effect was not found for Black parents. Here our findings echo prior studies suggesting that Black families may be more resilient to the potential impacts of negative life events than White families. For example, research from the family demography literature points to the less harmful consequences of divorce and being raised in a single parent household among Black, compared to White children (Cross 2020; Fomby and Cherlin 2007). Whereas White parents are less likely to have children with below median levels of schooling, having children who experience educational deficits may leave White parents more susceptible to health problems in part because normative schooling is expected.

Our final hypothesis considered the potential interaction between children’s education and parental education and whether these effects varied across parental race. The results show partial effects of moderation by parental education. Noteworthy was that Black parents with low levels of schooling whose offspring achieved above median levels of education have a much lower dementia risk than less-educated Black parents whose children achieved average or below average levels of schooling. Our outcomes thus suggest that educational advantages may be particularly protective for Black parents’ cognitive health. It is possible that Black parents may be more efficacious at drawing on the resources of their offspring to improve their health outcomes, and/or that Black offspring are more likely to provide instrumental support to parents (Fingerman et al. 2011; Sarkisian and Gerstel 2004; Taylor et al. 2022). Yet there is no parallel difference for White parents. Rather, for White parents, we found that the negative effects of children’s educational deficits (achieving less than the median years of schooling) compared to having children with average levels of schooling, was in fact concentrated among the least-educated parents themselves. Here too, we can draw on prior work from the family literature, which suggests that “downward” transfers of financial resources tend to be more common among White, compared to Black families (Fingerman et al. 2011; Swartz 2009).

Our results reaffirm the value of assessing the role of children’s education on cognitive health risks by parental race and parent’s own educational attainments. That higher-than-average levels of education among children act as a protective factor for Black parents, but conversely, that lower-than-average schooling is a health detriment to White parents, are significant findings that should help inform future studies of racial disparities in dementia. Additionally, our attention to the role of parental education suggests that the health advantages and disadvantages of offspring schooling are largely concentrated among the least-educated parents themselves.

Limitations to our study should be noted. First, this study looks at associations only and does not attempt to address the challenges posed by causality that these types of analyses present. That is, parents whose children achieve greater than or less than median levels of schooling may be predisposed towards better or worse cognitive health, respectively. However, prior work has examined this question using methods that do address causality, including the use of historical school reform data and these studies report that parents exposed to more years of schooling among children experienced better verbal memory scores and overall cognition (Ma 2019; Torres et al. 2022). Other studies also use propensity score matching (Dennison and Lee 2021) to tackle challenges related to causality. We see the challenges of addressing causal questions as a promising avenue of future work and one that would further frame the importance of considering how children’s educational attainment – or deficits - work to influence parental health. Second, the findings are limited by our measures of educational attainment. In the results presented here we include measures for the most-educated child only, but supplementary analyses show that findings remained robust when measuring education based on the least-educated child. Third, we do not offer a way to test the mechanisms that could help to explain some of our findings. If the advantages and disadvantages of children’s education are concentrated among the least-educated parents, do parents with highly-educated children receive greater instrumental or financial support from kids? Or does a child’s educational deficit lead parents to divert resources to offspring, thereby paving the way for poor health as parents age? Future research should assess these pathways and how they vary across racial groups. Finally, we constrain our analysis to Black and White families in the United States, which does not reflect the growing racial and ethnic diversity of American families. Future work should consider the growing number of immigrants who arrived after the dramatic change to immigration law in 1965. The aging of this cohort of immigrants is noteworthy given the racial diversity of these individuals and the substantial educational variation among their children.

Despite these constraints, this study underscores the importance of linking the education of younger generations to the study of racial and ethnic health disparities among older adults. In particular, the analyses presented here emphasize considering how children’s educational attainments - and deficits - work in concert to delay - or speed up - cognitive health decline in later life. Overlooking offspring education is a missed opportunity that could shed light on how family resources affect dementia onset differently across racial groups. Future work on racial and ethnic health disparities in particular should consider the lens of offspring mobility – not just attainment - to understand how health outcomes that are in part explained by generations of socioeconomic disadvantage may improve when educational access for one person can improve the health outcomes of family members.

APPENDIX

Appendix Table A1:

Child's education by parental education

	Child's education
	< Median	Median	>Median
Panel A: All parents
Parental education
< Median	39.05	19.73	10.91
Median	39.43	35.93	29.7
>Median	21.51	44.33	59.39
Total	100%	100%	100%
Panel B: Black parents
Parental education	< Median	Median	>Median
< Median	59.52	45.96	31.12
Median	26.53	28.26	23.49
>Median	13.95	25.78	45.39
Total	100%	100%	100%
Panel C: White parents
Parental education	< Median	Median	>Median
< Median	36.46	17.37	9.57
Median	41.07	36.62	30.12
>Median	22.47	46.01	60.32
Total	100%	100%	100%

Source: HRS, 2000–14

Note: Percentages weighted using person-level weights

Appendix Table A2.

Results from Event History Logistic Regression Models Predicting Dementia Onset, Adults aged over 65 (with Death and Attrition as outcomes) (N=46,451 person-observations

	Outcome: Death	Outcome: Attrition
	M4	M4
Respondenťs characteristics
Black (/White)	−0.13 [0.078]	0.252*** [0.038]
Respondenťs Education (/Median)
Above Median	−0.037 [0.052]	−0.025 [0.023]
Below Median	0.172** [0.054]	−0.041 [0.030]
Female	−0.409*** [0.046]	−0.011 [0.019]
Age	0.098*** [0.003]	0.058*** [0.002]
Marital Status (/Married)
Separated/Divorced	0.413*** [0.090]	0.205*** [0.043]
Widowed	0.303*** [0.054]	0.149*** [0.031]
Never married	0.424 [0.321]	0.458** [0.144]
Number of Kids	−0.005 [0.011]	−0.01 [0.006]
Proxy status	1.381*** [0.057]	0.663*** [0.050]
Respondent early life conditions
Parenťs education (<High school)
High school	−0.205** [0.067]	0.609*** [0.025]
Some College and above	−0.232** [0.083]	0.524*** [0.029]
Poor childhood health (/good)	0.03 [0.093]	0.017 [0.043]
Southern birth (/non-Southern)	0.066 [0.048]	0.161*** [0.024]
Characteristics of offspring
Child's Education (/Median)
Above Median	−0.122* [0.055]	0.024 [0.028]
Below Median	0.179*** [0.052]	0.089** [0.028]
Received transfers from children (/None)	0.338*** [0.101]	−0.118 [0.089]
1+ child cores/lives w/in 10miles (/None)	0.083 [0.047]	0.033 [0.029]
Constant	−6.797*** [0.170]	−6.797*** [0.170]
N	46,451	46,451

Source: HRS, 2000–14