Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Aug 5.
Published in final edited form as: Res Aging. 2020 Jun 24;43(1):14–24. doi: 10.1177/0164027520927137

College Selectivity and Later-Life Memory Function: Evidence From the Wisconsin Longitudinal Study

Sarah Garcia 1, Sara M Moorman 2
PMCID: PMC9353619  NIHMSID: NIHMS1826299  PMID: 32578499

Abstract

Research has shown a consistent association between college completion and laterlife cognition. We extend this work by examining whether college selectivity—the achievement level required to gain admission to a college—is associated with memory functioning more than 50 years later. We analyze data from 10,317 participants in the 1957–2011 Wisconsin Longitudinal Study to examine the relationship between college selectivity and later-life memory. Models control for childhood, midlife socioeconomic status, and later-life health and adjust for selection bias. Selective college attendance was associated with small benefits in memory at age of 72 even after accounting for socioeconomic status in both childhood and midlife and later-life health. The results of this study suggest that college selectivity may be an important component of the education–cognitive functioning relationship that has modest implications for intracohort differences in later-life cognition.

Keywords: cognition, aging, cognitive function, college selectivity, education, memory

Education is a well-established and powerful protective factor against the risk of cognitive decline in later life (Baumgart et al., 2015). Years of education seem to operate in a dose-dependent fashion, whereby each additional year completed decreases risk of cognitive decline (Sando et al., 2008). One study estimated that 19% of Alzheimer’s dementia cases worldwide are attributable to low educational attainment (Barnes & Yaffe, 2011).

A nascent literature is expanding this area of study by examining the relationship between older adults’ cognitive functioning and their experience with educational factors above and beyond attainment. For example, school context—such as more experienced teachers and higher spending per pupil—in Wisconsin high schools in the late 1950s has been associated with better language and executive functioning over 50 years later (Moorman et al., 2019). Likewise, factors including smaller student–teacher ratio and longer school year length in Alabama schools in 1935 have been associated with better global cognitive functioning of adults in their mid-70s (Crowe et al., 2013).

This work will flourish in future decades, as the participants in longitudinal studies of adolescence and young adulthood enter mid and later life. At present, empirical studies must rely on what archival data exist on the educational experiences of current cohorts of older adults. We contribute to this literature using historical data on college selectivity (CS) or the achievement level (e.g., grade point average [GPA], class rank) required to gain admission to a 4-year college in 1960. We use data from 10,317 participants in the Wisconsin Longitudinal Study (WLS) to examine whether the selectivity of the colleges from which participants graduated is associated with their memory ability in 2011 when participants were age 72.

Mechanisms Through Which CS Predicts Cognition in Later Life

CS represents both a student’s own abilities and the abilities of the peers with whom the student attends classes (Ross & Mirowsky, 1999). As such, it may affect cognition in later life through (a) the neurological development it produces during the college years and/or (b) the health-promoting socioeconomic resources that accumulate across adulthood among graduates of selective colleges. First, education requires students to develop communication skills such as reading, writing, inquiring, discussing, researching, and reasoning, as well as analytical skills such as observing, experimenting, summarizing, synthesizing, interpreting, and classifying (Mirowsky & Ross, 1989). Acquiring and practicing these higher order cognitive skills promotes development of brain reserve capacity, defined as individual differences in cognitive processes or neural networks that delay the manifestation of older age neuropathology (Stern, 2009). High-achieving peers may directly aid learning in the classroom through their questions and answers and contributions to the pace of instruction or they may indirectly contribute to learning through students’ motivation to learn (Lazear, 2001). High-achieving roommates and dorm mates affect a college student’s own academic performance positively (Sacerdote, 2011). Moreover, peers may influence behavioral choices that can have long-lasting neurocognitive effects. Fletcher and Frisvold (2011) found that college students and recent graduates who attended selective colleges showed less tobacco and marijuana use but more binge drinking than people who attended less selective colleges.

Second, attending a selective college gives young adults a leg up in socioeconomic attainment after college (Gaddis, 2015). The effect of selective college attendance on earnings is debated, but selective colleges do increase students’ probability of graduating and invest more money in the education of each student, relative to nonselective colleges (Hout, 2012). Other benefits are social, including access to elite social groups of peers and their families, and career networking through exclusive experiences outside the classroom, such as internships (Tholen et al., 2013). In turn, prestigious, high-income careers tend to involve cognitively complex tasks and require skills that promote mental stimulation, which predict higher cognitive functioning and reduced likelihood of dementia (Lachman et al., 2010; Schooler et al., 1999). Additionally, people can leverage socioeconomic resources to obtain health-promoting goods and services such as access to high-quality health insurance and health care providers, safer workplaces, nutritious foods, and neighborhoods that promote physical fitness (Link & Phelan, 1995). All of these factors promote cognitive health (Baumgart et al., 2015).

Empirical Evidence on Educational Attainment, Educational Context, and Adult Cognition

No research to date has considered the role of CS in promoting later-life cognitive functioning. Therefore, we draw on three related literatures to inform our research questions. The first literature establishes a relationship between school characteristics and later-life cognitive function, leading us to hypothesize that CS, too, might be related. These studies rely on the annual reports that primary and secondary public schools filed with regional governments at the city (Mantri et al. 2019), county (Crowe et al., 2013), or state (Moorman et al., 2019) level. These mandated reports included information about school or district characteristics such as teacher pay and the extent of racial segregation. These studies found higher cognition among older adults who attended more advantaged schools in their youth. Given that selective colleges are advantaged institutions (Hout, 2012), we anticipate that their graduates are likewise advantaged in later-life cognition.

The second literature we draw upon examines the extent to which graduates of selective and nonselective colleges diverge on physical health outcomes. Maintaining physical health, particularly cardiovascular health, is important for the prevention of Alzheimer’s disease and other dementias (Baumgart et al., 2015), such that educational factors that influence physical health in midlife might also influence cognitive health at later ages. Fletcher and Frisvold (2011, 2014) examined how CS predicted health and health behaviors and found that selective college attendance was associated with lower tobacco and marijuana use in early adulthood and reduced overweight for individuals in their 60s. Ross and Mirowsky (1999) found that CS had a small positive association with self-reported health and physical functioning among a nationally representative sample of adults age 60 and older. Similarly, in the UK, attendees of higher status universities from the 1970 British Birth Cohort Study had lower body mass index in midlife and healthier lifestyle behaviors than attendees of lower status universities (Bann et al., 2017). Insofar as selective college attendance seems to promote physical health, we hypothesize that it will promote cognitive health, as well.

The third relevant literature assesses how educational attainment may promote later-life cognition. There is evidence for both the direct neurological mechanism and the socioeconomic attainment mechanism outlined above. Several studies have examined the effect of timing of education in the life course, concluding that while education received in childhood appears to exert a lasting neurological effect, later educational experiences such as college may affect adult cognition more through their effects on adult socioeconomic status (SES; Foverskov et al., 2019; Hale, 2017; Kremen et al., 2019; Lyu & Burr, 2016; Nersesian et al, 1985; Richards et al., 2019; Zahodne et al., 2015). Other studies find that neurological effects and socioeconomic effects are of relatively equal importance to later-life function (Greenfield et al., 2020) although this balance may differ across birth cohorts (Hale, 2017). We expect that like total years of educational attainment, CS will influence later-life cognitive ability through one or both of these mechanisms.

Selection Into Selective Colleges

An important consideration is that by definition, selective colleges are not available to all students. Rather, students from socioeconomically advantaged backgrounds have more opportunities to build the strong academic records that make them attractive applicants to selective colleges (Campbell et al., 2002; Gerber & Cheung, 2008). That advantage begins with parenting: Mothers’ cognitive and noncognitive skills are strong predictors of children’s attendance at selective colleges (Doren & Grodsky, 2016). Children whose parents encourage mental stimulation, such as reading with them at home, have higher test scores (Sullivan & Brown, 2013). Later in childhood, high SES families have greater access to highly-resourced primary and secondary schools (Currie & Thomas, 2001), which are associated with selective college attendance (Berkowitz & Hoekstra, 2011).

High SES also privileges certain groups of students above others in college admissions in ways beyond students’ academic achievements. Mechanisms including legacy admissions, family monetary donations, admissions for athletics, and access to college admissions consultants benefit students from high SES families (Golden, 2006; Lartey, 2019). For all of these reasons, an important goal of this study is to document the extent to which childhood SES may attenuate any relationship between CS and later-life cognition. Healthy cognition in later life may be a function of lifelong socioeconomic advantage, rather than the benefits of selective college attendance per se.

Focus of Current Study

Our study aims to extend the literature on education as a risk/protective factor for later-life cognition by examining whether and how CS is related to memory function in later life. We use data from the WLS, which offers high-quality measures of SES across the life course, CS, and memory in later adulthood. After establishing a relationship between CS and memory net of confounders, we work toward assessing mechanisms by introducing adult SES and later-life health into our models.

Educational attainment is an important predictor of Alzheimer’s disease and related dementias (Barnes & Yaffe, 2011; Baumgart et al., 2015), so our findings will illuminate whether part of this association is attributable to factors beyond attainment. If CS is associated with cognitive functioning in later life, institutional selectivity and other educational policies may be target for intervention to reduce Alzheimer’s and related dementias. This point is particularly important given racial/ethnic disparities in both rates of dementia and access to prestigious colleges and universities (Reardon et al., 2012). Findings may also expose new pathways for examining the potential mechanisms that link CS and other aspects of educational context to cognitive functioning.

Method

Data

The WLS is a random sample of high school graduates in 1957 from Wisconsin who were 72 years old when last interviewed in 2011. Data were collected via telephone, mail, and in-person surveys in 1957, 1964, 1975, 1993, 2004, and 2011. The initial survey of 10,317 respondents represented White, non-Hispanic American men, and women of that cohort who completed at least a high school education (Herd et al., 2014). The retention rate excluding those who died before 1993 was 87%, for 2004 was 81%, and for 2011 was 72%. For this study, we use the Heckman correction (see Analytic Strategy and Missing Data subsection, for description) to estimate the probability that respondents remained in the sample for the entirety of the study period, allowing us to interpret the results as though the outcome data were observed for all respondents (N = 10,317).

Predictor and Outcome Measures

Educational attainment.

Respondents were asked about their years of education in 1975 and 1993. All participants graduated high school, such that the lowest educational attainment was 12 years. We used the 1993 response when available and the 1975 response if the 1993 response was missing. Among those with more than a high school education, response categories included some college, bachelor’s degree, and education beyond a bachelor’s degree (master’s degree, professional degree, or PhD). The “some college” category included respondents with a 1-year college certificate, a 2-year teaching certificate, associate’s degree, registered nursing credential, 1 or more years post-2-year degree (but no higher degree), and 1 or more years of college but no degree or certificate. Among college completers, the range of years when the bachelor’s degree was obtained was 1961–1994 with a mean of 1,965 (SD = 6.40).

CS.

The 1975 and 1993 survey questionnaires included questions about the name of the college where the respondent earned a bachelor’s degree or its equivalent (if they graduated from college). For college information, if the 1993 survey response was unavailable, we used the 1975 survey response to determine the college from which the respondent graduated.

The WLS matched college names to Barron’s Profile of American Colleges, which rates schools based on median SAT scores, high school rank, and high school GPA of the freshman class (Fine, 1969). College names were matched to the 1969 edition of this resource because it provides the earliest measures of CS. Among WLS respondents who completed college, the majority (84%) did so prior to 1970 (or by age 26). Measurement error as a result of the remaining time difference is likely minimal due to the stability of college rankings over time (Gnolek et al., 2014).

The Barron’s Profile categorized colleges into six levels of selectivity (see Fine, 1969, for more information). Very few individuals in the sample graduated from schools at the extremes (e.g., less than 6% attended “most” or “highly” selective colleges, and about 7% attended “least” or “not” selective colleges). Because the six-category measure is not normally distributed, we aggregated the top and bottom three categories into a dichotomous measure. “Selective” colleges constituted 29% of the sample of 4-year degree college graduates. These colleges had selection criteria including a high school GPA of B− or higher, and students who were in the top half of their high school class (Fletcher & Frisvold, 2014). The class median SAT scores were 900 (of the 1,600) or higher. “Nonselective” colleges constituted 71% of the sample of 4-year degree college graduates. (In secondary analyses, below, we examined the effect of different operationalizations of selectivity.)

We combined the educational attainment and CS measures into a five-category education variable: high school diploma, some college or 2-year degree, bachelor’s degree from nonselective college, bachelor’s degree from selective college, and graduate degree (regardless of selectivity of bachelor’s degree institution).

Memory functioning in later life.

At age 72 (in 2011), a random 80% of participants completed three memory tests. Factor analyses (available upon request) indicated these three tests were specific to memory. Participants were read a list of 10 nouns and asked to repeat as many as they could remember, in any order (Brandt et al., 1988). After a delay of approximately 9 min, participants were asked—without warning—to repeat as many of the 10 words as they could remember. The score for both measures is the number of words correctly recalled. Additionally, participants reordered several series of digits from smallest to largest, following a modified protocol of the Wechsler Adult Intelligence Scale (WAIS-III) digit backward subtest (Wechsler, 1997). The final score ranged from 0 to 12. We standardized scores on the three tests by subtracting the mean and dividing by the standard deviation, averaged them, and finally standardized the mean score.

Other Focal Measures

Adolescent SES covariates.

The WLS includes a factor-weighted SES score that is a combination of father’s education, mother’s education, father’s occupational prestige score on the 1961 Duncan Socioeconomic Index (SEI; Hauser & Warren, 1997), and average annual income of the parents during the period from 1957 to 1960. We incorporated the SES index as opposed to individual measures to reduce missing data. Results using individual covariates (available upon request) were similar to using the SES score. Parents’ educational attainment was reported by participants, and occupational prestige and income were drawn from state tax records. We calculated the natural log of this measure to transform the skewed distribution. Other measures not included in the SES score included number of siblings, whether participants reported living with both parents up until 1957, and whether participants attended secondary school in a rural area (i.e., fewer than 10,000 residents).

Demographic covariates.

We adjusted for gender whether the respondent was a traditional or nontraditional college student defined by age when college degree was obtained (age 25 or younger vs. older than 25).

Midlife SES.

In 1993, respondents were asked about their income from wages, salaries, commissions, and tips before taxes and other deductions in the last 12 months. We selected 1993, when participants were aged 54 because it preceded the measurement of cognition and presumably the onset of any age-related cognitive decline. We calculated the log to remove skewness and incorporated this variable as a measure of midlife SES. We also included the SEI score as a measure of occupational prestige in midlife (Duncan, 1961). In 1993, participants reported their occupation for their first or only job in their current or last employer job spell. We divided by 100 to increase interpretability of this measure. The scores ranged from −4.68 to 13.69 where a higher score indicates a more prestigious occupation.

Later-life health.

We controlled for self-reported health in later life because cognitive health is associated with physical and mental health in older age. We included a dichotomous measure of respondent’s self-reported health at age 72 (e.g., fair and poor vs. good, very good, or excellent).

Selection covariates.

Given that attrition from the study before age 72 might be correlated with cognitive functioning, we used a Heckman (1977) correction—a statistical technique used to estimate the probability that an individual dropped out of the sample—to control for selection bias. The procedure that (described in the Analytic Strategy and Missing Data subsection) requires covariates that predict selection into the sample, including one or more instrumental variables that predict selection but not the outcome, memory. We included the following selection covariates: whether the respondent was a member of a church in 1975 (instrumental variable), SES during adolescence as indicated above using the factor-weighted SES score, gender, standardized test score during junior year of high school gathered from the Wisconsin State Testing Service Records and incorporated into the WLS data, and self-reported health in 1993.

Analytic Strategy and Missing Data

We first generated descriptive statistics to compare the characteristics of graduates across education categories. We then used a multivariate regression approach to estimate the effect of education category on memory at age of 72. We used robust standard errors to account for the correlations among respondents who attended the same college and a Heckman correction in each statistical model to control for selection bias in the outcome (Heckman, 1977; Stolzenberg & Relles, 1997; Winship & Mare, 1992).

According to the model, each participant, i, produced memory scores at age 72, gi, determined by observed socioeconomic and demographic factors, zi, and unobserved factors, εi. The memory score gi was observed only if the participant remained in the sample and was in the sample of participants who completed the battery of cognitive tests in 2011.

gi=zi+εi.

Because higher memory performance in later life is likely correlated with the probability of remaining in the sample to complete the cognitive tests, estimates of gi will be biased unless participants who left the sample are accounted for in the regression analysis. To correct for this bias, we used a two-step Heckman correction (Heckman, 1977), which assumes that outcome data are missing not at random.

The Heckman correction estimated a variable where D indicates the probability of completing cognitive tests in later life, Z is a vector of individual and social explanatory variables, y is a vector of unknown parameters, and Φ is the cumulative distribution function of the standard normal distribution (as the Heckman correction involves a normality assumption for the error terms).

prob(D=1|Z)=Φ(Zy).

The Heckman correction estimated a probit model—the canonical specification for this relationship—to determine the probability of completing the cognitive tests at age of 72 based on the selection covariates described above. The selection model requires at least one instrumental variable that is a predictor of attrition but is uncorrelated with memory (Heckman, 1977). We used church membership in 1975 for this purpose because membership predicts whether respondents remained in the sample (see selection model results in Table 2), but not memory functioning in 2011 (p > .05). The error term of the selection probit model is then correlated with the error terms of primary regression models predicting memory at age of 72. As a result of the Heckman correction, the results of these regression models can be interpreted as representative of the entire WLS sample (N = 10,317).

Table 2.

Linear Regressions Indicating Associations Between Education Category and Memory Functioning at Age 72.

Model 1 Model 2 Model 3
B (SE) B (SE) B (SE)
Prediction model results
 Education category (nonselective college graduate is reference)
  High school −0.15 (0.13) −0.07 (0.06) −0.06 (0.l2)
  Some college −0.05 (0.08) −0.01 (0.07) −0.00 (0.07)
  Selective college graduates 0.14* (0.07) 0.13* (0.07) 0.l5* (0.07)
  Postgraduate education 0.19* (0.08) 0.16* (0.08) 0.l7* (0.08)
 Adolescent SES
  Adolescent SES score (standardized scale is 0–4.6) 0.04** (0.02) 0.03 (0.02) 0.03 (0.02)
  Number of siblings −0.00 (0.00) −0.00 (0.00) −0.00 (0.00)
  Lived with both parents during high school 0.00 (0.02) −0.00 (0.02) −0.00 (0.02)
  Rural residence during adolescence (urban is reference) −0.11*** (0.02) −0.10*** (0.01) −0.l0*** (0.02)
 Demographics
  Female 0.39*** (0.06) 0.41*** (0.06) 0.40*** (0.06)
  Obtained bachelor’s degree at age <25 0.09 (0.06) 0.09 (0.06) 0.08 (0.06)
 Midlife SES
  Income at age 53 (scale is 0–300,000) 0.01 (0.01) 0.03 (0.02)
  Occupational prestige score (scale is −463 to 1,424) 0.04*** (0.0l) 0.03*** (0.0l)
Later-life health
 Fair or poor self-reported health at age 72 (good, very good, or excellent is reference) −0.28*** (0.02)
Selection model results
 Adolescent SES score (standardized scale is 0–4.6) −0.04*** (0.01) −0.04*** (0.01) −0.04*** (0.0l)
 IQ score during high school (standard scale is −2.8to 2.9) 0.18*** (0.0l) 0.17*** (0.0l) 0.l7*** (0.0l)
 Female 0.09 (0.05) 0.09 (0.05) 0.09 (0.05)
 Educational attainment (0 = high school, 3 = graduate education) 0.03 (0.07) 0.03 (0.07) 0.03 (0.07)
 Fair or poor health in 1993 (good, very good, or excellent is reference) −0.17*** (0.03) −0.17*** (0.03) −0.l4*** (0.03)
 Cognition score in 1993 0.02*** (0.0l) 0.02*** (0.00) 0.02*** (0.0l)
 Church member in 1975 (instrumental variable) 0.09*** (0.02) 0.09*** (0.02) 0.09*** (0.02)
Intercept −0.38 (0.20) −0.38 (0.2l) −0.33 (0.20)
Transformed ρ −1.26*** (0.04) −1.24** (0.03) − l.2l*** (0.03)
Open in a new tab

Note. N = 10,317. SES = socioeconomic status.

Asterisks denote statistical significance where *p < .05. **p < .01. ***p < .001.

Regression Model 1 included education category as the main predictor and potential confounders including gender, childhood health, traditional or nontraditional college student status, adolescent SES score, number of siblings, whether the respondent lived with both parents during high school, and rural versus urban residence. We consecutively added mechanisms and confounders in Models 2 and 3. To test whether socioeconomic resources that selective college attendance provides may drive the association between CS and later-life memory function, Model 2 incorporated postcollege, midlife measures of SES including household income and occupational prestige. Finally, to test whether the association between CS and later-life memory function is a result of poor later-life health, Model 3 incorporated self-reported health in later life. In addition to study attrition, there was a subset of respondents who remained in the study at age of 72 and completed the memory measures but were missing data on covariates of interest. We assumed that these data are missing at random and conducted multiple imputation by chained equations that combined estimates across 10 data sets using Rubin’s (1987) rules. Among respondents who remained in the sample and completed the cognitive measures, the variable with the most missing data was midlife household income with about 9% missing.

We also conducted a variety of sensitivity analyses to test the robustness of the results. To test whether our findings are attributable to college itself or simply attributable to precollege cognitive skills and individual abilities, we incorporated a regression analysis breaking out nonselective graduates into two categories: a high achiever category that combined students who were in the 75th or higher percentile or higher for high school rank and standardized test score (N = 264) and an average achiever category that combined students who were lower than the 75th percentile on either high school rank or standardized test score (N = 655). We generated two categories of nonselective college graduates from high school rank and test scores available in the WLS data.

Next, we estimated multilinear regression models using two different classifications for CS to determine whether our measure of CS masked effects at different levels of selectivity. We incorporated a six-category measure of CS rather than the dichotomous indicator. We then separately incorporated a dichotomous indicator of CS but incorporated colleges rated as “selective” into our selective, rather than the nonselective, side of our dichotomy (i.e. not selective/less selective vs. selective/highly selective/most selective).

Finally, we reestimated the models after removing respondents who died before 2011 to determine whether selective attrition was due to mortality. The Heckman correction allows us to interpret the regression model results as representative of the entire WLS sample, but a substantial portion of the WLS sample (about 22%) died before 2011. Including deceased individuals in the sample when they could not possibly have produced a memory score may introduce selection bias (M. Jones et al., 2015). To examine this possibility, we removed respondents who died before 2011 and reestimated the regression models.

Results

Descriptive Results

Education category.

Table 1 reports descriptive statistics for the full sample and then separately by education category. Over half (59.4%) of participants had no further education after high school. Additionally, 15.6% attended some college, 8.9% graduated from a nonselective 4-year college, 3.6% graduated from a selective 4-year college, and 12.5% attended an institution for postgraduate education. Analyses of variance (Mackenbach et al., 2015) with Scheffe post hoc tests revealed significant between-education group differences. Specific to our research questions, post hoc comparison tests indicated that memory functioning at age 72 was significantly higher among selective college graduates than among nonselective college graduates (p < .001).

Table 1.

Means (and Standard Deviations) or Proportions, Wisconsin Longitudinal Study, 2011.

Variable Total Sample (N = 10,317) High Schoola (N = 6,137) Some Collegeb (N = 1,613) Nonselective Collegec (N = 919) Selective Colleged (N = 374) Postgraduate Educatione (N = 1,274) Significant Subgroup Differences
Memory (standard deviations) 0.00 (1.00) −0.17 (0.99) 0.03 (0.99) 0.16 (0.98) 0.38 (0.93) 0.39 (0.91) ab, ac, ad, ae, bd, be, cd, ce
Midlife SES
 Income at age 53 (range: 0–300,000) −0.03 (0.90) −0.23 (0.67) −0.04 (0.83) 0.13 (0.89) 0.50 (1.40) 0.54 (1.30) ab, ac, ad, ae, bc, be, cd, ce, de
 Occupational prestige score (range: −463 to 1,424) 45.76 (24.50) 37.14 (21.99) 48.82 (22.22) 53.99 (24.00) 63.82 (20.19) 67.28 (18.63) ab, ac, ad, ae, bc, be, cd, ce, de
Adolescent SES
 Adolescent SES score (range: 1–97) 2.52 (0.76) 2.30 (0.71) 2.70 (0.72) 2.65 (0.76) 3.20 (0.62) 2.94 (0.71) ab, ac, ad, ae, bc, be, cd, ce, de
 Number of siblings 3.20 (2.39) 2.97 (2.17) 3.56 (2.54) 2.85 (2.17) 2.22 (1.78) 2.57 (1.99) ab, ac, ad, ae, bc, be, cd, ce, de
 Lived with both parents during high school 89.82% 89.11% 89.2% 90.7% 92.69% 91.85% ac, ad, ae, bd, be, cd, de
 Rural residence during high school 51.17% 56.56% 43.6% 49.6% 33.33% 41.76% ab, ac, ad, ae, bc, cd, ce, de
Later-life health
 Poor self-reported health at age 72 13.90% 15.83% 13.5% 13.01% 10.52% 6.91% ab, ac, ad, ae, be, cd, ce, de
Demographics
 Female 51.60% 56.94% 51.59% 46.86% 40.64% 35.61% ab, ac, ad, ae, bc, be, cd, ce, de
 Obtained bachelor’s degree before age 25 49.90% — 50.00% 27.95% 82.46% 68.55% ac, ad, ae, bc, be, cd, ce, de
Open in a new tab

Note. Significant subgroup differences were tested using analysis of variance with a Scheffe post hoc test to adjust for multiple comparisons. Significant subgroup differences (p < .05) are denoted as the following: ab = high school versus some college; ac = high school versus nonselective college; ad = high school versus selective college; ae = high school versus post-graduate education; bc = some college versus nonselective college; bd = some college versus selective college; be = some college versus postgraduate education; cd = nonselective college versus selective college; ce = nonselective college versus postgraduate education; and de = selective college versus postgraduate education. SES = socioeconomic status.

Multivariate Results

Table 1 reports results from the multivariate linear models. Model 1 estimates the associations between the education category and memory functioning at age of 72, including potential confounders (gender, age when the respondent received bachelor’s degree, adolescent SES score, number of siblings, whether the respondent lived with both parents during high school, and rural/urban residence), with the Heckman correction for selection bias. High school completers had similar memory scores as nonselective college graduates (β = −0.15, p > .05). Respondents with some college but no degree had similar memory to nonselective college graduates (β = −0.05, p > .05). Selective college graduates had significantly higher memory (β = 0.14, p < .05) scores. Respondents with postgraduate education had significantly higher memory (β = 0.19, p < .05) than nonselective college graduates. In terms of our research question, these results indicated that CS is associated with higher memory functioning in older age.

Model 2 added midlife income and occupational prestige. Results for education category and memory remain similar to Model 1, including midlife SES slightly attenuated the association between CS and memory functioning (β = 0.13, p < .05). Higher occupational prestige scores were association with memory functioning (β = 0.04, p < .001). There was no relationship between midlife income and education category (β = 0.01, p > .05).

Model 3 incorporates self-report health at age of 72 to Model 1. Respondents with poor self-reported health in older age had significantly lower memory scores (β = −0.28, p < .001), but the results for education category and memory remain similar to prior models, including later-life health slightly increased the association between selective college attendance and memory functioning in later life (β = 0.15, p < .05).

Table 1 also reports the results from the selection model using the Heckman correction. The results can be interpreted as how well the measures predict remaining in the sample and completing the memory tests. IQ score during high school (β = 0.17, p < .001), self-reported health in midlife (β = −0.14, p < 0.001), cognition score in 1993 (β = 0.02, p < .001), adolescent SES (β = −0.04, p < .001), and church membership in 1975 (β = 0.09, p < .001) predict whether respondents remain in the sample and complete the memory tests in later life. Gender (β = 0.09, p > .05) and educational attainment (β = 0.03, p > .05) do not predict sample attrition. The transformed p (or ρ) reports the assumed correlation between the errors in the prediction and selection equations. The statistically significant result (p < .001) indicates a strong assumed correlation, which means that the Heckman correction provides an important adjustment for attrition bias.

Sensitivity Analyses

To test whether our findings are attributable to college itself or simply attributable to precollege cognitive skills and individual abilities, we constructed a new predictor disaggregating nonselective college graduates by their high school achievement and standardized test scores. Ideally, we would test higher ability/achievement graduates of nonselective colleges and lower ability/achievement graduates of selective colleges, but because low-achieving students were very unlikely to attend a selective college, could not analyze the latter group. The results (Online Appendix A) indicated that average and high-achieving graduates of nonselective colleges have similar memory scores in later life. These findings suggest that CS has an association with later-life memory independent of achievement and abilities in high school.

Second, we estimated the multilinear regression models using a six-category measure of CS. The findings (Online Appendix B) indicated that the effects of CS on later-life memory functioning were approximately linear. The exception is for the most selective category, which also has high standard errors relative to the other categories, due to small sample size (n = 58). The small number of participants who attended “most” selective colleges may be outliers, which supports our decision to use a dichotomous indicator.

Third, we estimated the multilinear regression models using a dichotomous indicator of CS where “selective” colleges were included in the selective category. The results (Online Appendix C) show similar results to the original analysis; CS is significantly associated with memory function at age 72, net of SES.

Fourth and finally, we reestimated the models after removing respondents who died before 2011 to determine whether selective attrition was due to mortality. The results (Online Appendix D) are similar to the original analysis; excluding respondents who died prior to 2011 does not reduce the significant association between education category and memory function in older age. This finding demonstrates that selective attrition is attributable to factors other than mortality and suggests that cognitive functioning predicts sample attrition.

Discussion

Educational attainment, or duration of education, has benefits for cognitive health in later life (Jefferson et al., 2011). An emerging literature, which we contribute to here, suggests that beyond education quantity, educational context matters (Dudovitz et al., 2016; Fletcher & Frisvold, 2011, 2014; Garcy & Berliner, 2018; Ross & Mirowsky, 1999). Our study extends prior literature on the relationship between education and cognition by examining whether CS—the achievement level required to gain admission to a college—was associated with memory and at age 72 among a cohort of individuals from Wisconsin and whether midlife SES explained this relationship. Compared to those who graduated from nonselective colleges, those who attended selective colleges had higher memory scores after controlling for selection bias. This relationship remained similar after controlling for SES and health at several points in the life course. These results highlight ways in which disparities among educational experiences may have cumulative and long-lasting effects for later-life health as well as for SES in adulthood.

Implications

Our findings have implications for future scholarship, practice, and policy. Future research might examine the mechanisms through which CS predicts later-life cognitive functioning and examine the means through which inequality creates differential access to educational and socioeconomic opportunities.

Non-SES mechanisms.

Our findings demonstrate the importance of CS for midlife SES, with those who attended more elite colleges having higher income and occupational prestige. However, midlife SES did not significantly mediate the association between CS and memory functioning in later life. This finding suggests that CS affects cognitive functioning in ways that persist into later life and are independent of SES. Future research might examine non-SES mechanisms through which CS predicts later-life memory.

In particular, CS may promote neurological development in young adulthood. This enriched development would produce brain reserve capacity, or “extra” cognitive power that grants people the capacity to continue to show normal cognitive function in later life despite advancing dementia (Stern, 2009). Brain reserve capacity is a latent construct, and researchers often use educational attainment as a proxy for it (R. N. Jones et al., 2011). High-quality educational experiences such as selective college attendance may contribute to it, just as additional years of exposure to education do. Selective colleges might have a more rigorous curriculum than nonselective colleges, for example, or perhaps the influence of high-achieving peers allows students at selective colleges to develop advanced cognitive skills (Lazear, 2001).

Other research has shown that CS predicts adult health behaviors that may affect cognitive functioning in later life (Fletcher & Frisvold, 2011, 2014; Ross & Mirowsky, 1999). For example, Fletcher and Frisvold (2011) found that CS was associated with lower smoking rates among college students and recent graduates, and smoking is a risk factor for cognitive decline in later life (Baumgart et al., 2015). CS may produce different norms surrounding health behaviors that shape lifestyles and habits, and these behaviors may have long-lasting consequences for cognitive health (Bourdieu, 1977).

Reducing inequality in access to selective colleges.

While adolescent SES only slightly attenuated the associations between education category and memory functioning, our descriptive results revealed that higher adolescent SES was associated with selective college attendance. This finding aligns with cumulative advantage theory which contends that SES advantages compound over time and increase exposure to opportunities (Ferraro & Shippee, 2009). Children from advantaged backgrounds are more likely to benefit from concerted cultivation and to receive high-quality primary and secondary schooling and are thus more likely to be admitted to selective colleges as a result of their academic preparation and achievement (Bourdieu & Passeron, 1990; Campbell et al., 2002; Currie & Thomas, 2001; Gerber & Cheung, 2008; Lareau, 2011).

Cumulative inequality theory also suggests that socioeconomic disadvantage accumulates over the life course to generate inequality (Ferraro & Shippee, 2009). Our finding that adolescent SES is associated with CS is particularly important given continued inequality in access to prestigious institutions (Reardon et al., 2012). People of color and students from low socioeconomic backgrounds are underrepresented at elite institutions, and some research suggests that these populations received the greatest economic and health benefit from selective college attendance (Brand & Xie, 2010; Schafer et al., 2013). Improving equality of opportunity to access elite colleges may promote cognitive health in later life and thus reduce the growing social and economic costs of caring for older adults with dementia (Wimo et al., 2017). Further understanding of the relationship between CS and later-life cognitive functioning has the potential to reduce inequality in access to higher education through policy and higher education institutional practices, particularly in the context of increasing educational attainment and longevity.

Limitations

The WLS is well suited to study education and cognitive functioning in older age because of its large sample size, multiple measures of memory functioning in older age, and ability to measure CS. Despite the methodological strengths of the WLS, there are also limitations of the data that limit its external and internal validity. The WLS sample is representative of White non-Hispanic participants who graduated from high schools in Wisconsin in 1957, which limits the ability to generalize the findings to other populations. The inability to study other racial and ethnic groups is particularly challenging given racial and ethnic disparities in dementia incidence (Mayeda et al., 2016) and low rates of access to selective colleges (Reardon et al., 2012).

We were unable to consider other contextual characteristics that may affect selection into prestigious institutions. First, we only know the selectivity of the college that the student graduated from. While we are unable to stratify the analysis by students who transferred from selective to less selective colleges and vice versa, less than 18% of 4-year college graduates reported attending multiple 4-year colleges by age of 25.

Additionally, peer group achievement and learning influences student achievement and learning, and selective college attendees are surrounded by high-achieving peers (Winston & Zimmerman, 2004). Our study was unable to account for peer influences on achievement and learning in primary and secondary schooling, so future scholarship might examine variations in the peer group abilities to determine the ways in which peers affect an individual’s educational trajectories and later-life memory function.

Although the association between CS and memory was statistically robust, the effect was modest in size at less than a quarter of a standard deviation. However, its benefits were comparable in size to the effects of completing a college degree relative to a high school diploma only; educational attainment is one of the strongest predictors of adult cognition (Baumgart et al., 2015). If the brain’s reserve capacity marks the amount of cognitive loss experienced before exhibiting functional problems, then even small effects contribute to cognitively healthy life expectancy (Lièvre et al., 2008; Stern, 2009).

Finally, memory functioning and decline varies across the life course, and due to the age-group of the WLS cohort, we were unable to capture the magnitude or life course decline in memory (Hartshorne & Germine, 2015). Our study was limited to assessments of cognition among an age-group that may be too young to exhibit symptoms of clinical dementia. The WLS cohort was age 72 during the 2011 follow-up, and one study found that Alzheimer’s disease incidence nearly tripled after age 75 (Kukull et al., 2002). Greater intracohort variable in memory functioning—especially in terms of rate of decline—would be more prominent in an older sample. However, the pathophysiological processes underlying dementia manifest over many years or decades (Jack et al., 2013). Our study captures preclinical symptoms of dementia through memory functioning in later life.

Conclusion

Despite these limitations, findings of this study suggest that other educational factors beyond attainment are important for later-life cognitive functioning. Given increasing longevity, increasing educational attainment, and continued disparities in access to prestigious colleges/universities, our work has population-level implications that extend beyond the economic benefits of education. Future research might examine the potential mechanisms through which CS predicts memory in later life, and policy makers and higher educational institutions should consider the role of legislation and institutional practices that may reduce disparities in access to prestigious college/universities.

Supplementary Material

Supplemental material

Acknowledgments

We would like to acknowledge the Minnesota Population Center (Grant Number P2C HD041023), and the second author’s work was supported by the National Institute on Aging at the National Institutes of Health (Grant Number R01 AG057491). The Wisconsin Longitudinal Study was supported by the National Institute on Aging (AG-9775 and AG-21079), with additional support from the Vilas Estate Trust, the National Science Foundation, the Spencer Foundation, and the Graduate School of the University of Wisconsin–Madison.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Minnesota Population Center (Grant Number P2C HD041023) and National Institute on Aging (AG-9775 and AG-21079).

Biographies

Sarah Garcia is a PhD candidate in the Department of Sociology at the University of Minnesota. Her areas of expertise include health disparities and disability. She studies the persistent relationship between education and health outcomes and the relationship between economic inequality and disability.

Sara M. Moorman is an associate professor in the Department of Sociology at Boston College. She is a social gerontologist and expert in quantitative and survey methods. Her research examines end-of-life medical decision making, negative psychological experiences in personal relationships, and early-life predictors of adult cognitive functioning.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental Material

Supplemental material for this article is available online.

References

  1. Bann D, Hamer M, Parsons S, Ploubidis GB, & Sullivan A (2017). Does an elite education benefit health? findings from the 1970 British cohort study. International Journal of Epidemiology, 46(1), 293–302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Barnes DE, & Yaffe K (2011). The projected effect of risk factor reduction on Alzheimer’s disease prevalence. The Lancet Neurology, 10(9), 819–828. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baumgart M, Snyder HM, Carrillo MC, Fazio S, Kim H, & Johns H (2015). Summary of the evidence on modifiable risk factors for cognitive decline and dementia: A population-based perspective. Alzheimer’s & Dementia, 11(6), 718–726. [DOI] [PubMed] [Google Scholar]
  4. Berkowitz D, & Hoekstra M (2011). Does high school quality matter? Evidence from admissions data. Economics of Education Review, 30(2), 280–288. [Google Scholar]
  5. Bourdieu P (1977). Cultural reproduction and social reproduction. In Karabel J & Halsey AH (Eds.), Power and ideology in education (pp. 487–511). Oxford University Press. [Google Scholar]
  6. Bourdieu P, & Passeron J-C (1990). Reproduction in education, society and culture (Vol. 4). Sage. [Google Scholar]
  7. Brand JE, & Xie Y (2010). Who benefits most from college? Evidence for negative selection in heterogeneous economic returns to higher education. American Sociological Review, 75(2), 273–302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Brandt J, Spencer M, & Folstein M (1988). The telephone interview for cognitive status. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 1(2), 111–117. [Google Scholar]
  9. Campbell FA, Ramey CT, Pungello E, Sparling J, & Miller-Johnson S (2002). Early childhood education: Young adult outcomes from the abecedarian project. Applied Developmental Science, 6(1), 42–57. [Google Scholar]
  10. Crowe M, Clay OJ, Martin RC, Howard VJ, Wadley VG, Sawyer P, & Allman RM (2013). Indicators of childhood quality of education in relation to cognitive function in older adulthood. Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences, 68(2), 198–204. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Currie J, & Thomas D (2001). Early test scores, school quality and SES: Long run effects on wage and employment outcomes worker wellbeing in a changing labor market (pp. 103–132). Emerald Group Publishing. [Google Scholar]
  12. Doren C, & Grodsky E (2016). What skills can buy: transmission of advantage through cognitive and noncognitive skills. Sociology of education, 89(4), 321–342. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Dudovitz RN, Nelson BB, Coker TR, Biely C, Li N, Wu LC, & Chung PJ (2016). Long-term health implications of school quality. Social Science & Medicine, 158, 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Duncan OD (1961). A socioeconomic index for all occupations. In Reiss AJ (Ed.), Occupations and social status (pp. 109–138). Free Press. [Google Scholar]
  15. Ferraro KF, & Shippee TP (2009). Aging and cumulative inequality: How does inequality get under the skin? The Gerontologist, 49(3), 333–343. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Fine B (1969). Barron’s profiles of American colleges. Barron’s Educational Series. [Google Scholar]
  17. Fletcher JM, & Frisvold DE (2011). College selectivity and young adult health behaviors. Economics of Education Review, 30(5), 826–837. [Google Scholar]
  18. Fletcher JM, & Frisvold DE (2014). The long run health returns to college quality. Review of Economics of the Household, 12(2), 295–325. [Google Scholar]
  19. Foverskov E, Mortensen EL, Holm A, Pedersen JLM, Osler M, & Lund R (2019). Socioeconomic position across the life course and cognitive ability later in life: The importance of considering early cognitive ability. Journal of Aging and Health, 31(6), 947–966 [DOI] [PubMed] [Google Scholar]
  20. Gaddis SM (2015). Discrimination in the credential society: An audit study of race and college selectivity in the labor market. Social Forces, 93(4), 1451–1479. [Google Scholar]
  21. Garcy AM, & Berliner DC (2018). Context and implications document for: A critical review of the literature on the relationship between school quality and health inequalities. Review of Education, 6(1), 67–68. [Google Scholar]
  22. Gerber TP, & Cheung SY (2008). Horizontal stratification in postsecondary education: Forms, explanations, and implications. Annual Review of Sociology, 34, 299–318. [Google Scholar]
  23. Gnolek SL, Falciano VT, & Kuncl RW (2014). Modeling change and variation in US News & World Report college rankings: What would it really take to be in the Top 20? Research in Higher Education, 55(8), 761–779. [Google Scholar]
  24. Golden D (2006). The price of admission: How America’s ruling class buys its way into elite colleges—And who gets left outside the gates. Crown Publishing Group/Random House. [Google Scholar]
  25. Greenfield EA, Moorman S, & Rieger A (2020). Life course pathways from childhood socioeconomic status to later life cognition: Evidence from the wisconsin longitudinal study. The Journals of Gerontology: Series B. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Hale JM (2017). Cognitive disparities: The impact of the great depression and cumulative inequality on later-life cognitive function. Demography, 54(6), 2125–2158. [DOI] [PubMed] [Google Scholar]
  27. Hartshorne JK, & Germine LT (2015). When does cognitive functioning peak? The asynchronous rise and fall of different cognitive abilities across the life span. Psychological Science, 26(4), 433–443. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Hauser RM, & Warren JR (1997). Socioeconomic indexes for occupations: A review, update, and critique. Sociological Methodology, 27(1), 177–298. [Google Scholar]
  29. Heckman JJ (1977). Sample selection bias as a specification error (with an application to the estimation of labor supply functions). National Bureau of Economic Research Cambridge. [Google Scholar]
  30. Herd P, Carr D, & Roan C (2014). Cohort profile: Wisconsin longitudinal study (WLS). International Journal of Epidemiology, 43(1), 34–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Hout M (2012). Social and economic returns to college education in the United States. Annual Review of Sociology, 38, 379–400. [Google Scholar]
  32. Jack CR Jr., Knopman DS, Jagust WJ, Petersen RC, Weiner MW, Aisen PS, Shaw LM, Vemuri P, Wiste HJ, Weigand SD, Lesnick TG, Pankratz VS, Donohue MC, & Trojanowski JQ (2013). Tracking pathophysiological processes in Alzheimer’s disease: An updated hypothetical model of dynamic biomarkers. The Lancet Neurology, 12(2), 207–216. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Jefferson AL, Gibbons LE, Rentz DM, Carvalho JO, Manly J, Bennett DA, & Jones RN (2011). A life course model of cognitive activities, socioeconomic status, education, reading ability, and cognition. Journal of the American Geriatrics Society, 59(8), 1403–1411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Jones M, Mishra GD, & Dobson A (2015). Analytical results in longitudinal studies depended on target of inference and assumed mechanism of attrition. Journal of Clinical Epidemiology, 68(10), 1165–1175. [DOI] [PubMed] [Google Scholar]
  35. Jones RN, Manly J, Glymour MM, Rentz DM, Jefferson AL, & Stern Y (2011). Conceptual and measurement challenges in research on cognitive reserve. Journal of the International Neuropsychological Society, 17(4), 593–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Kremen WS, Beck A, Elman JA, Gustavson DE, Reynolds CA, Tu XM, Sanderson-Cimino ME, Panizzon MS, Vuoksimaa E, Toomey R, Fennema-Notestine C, Hagler DJ Jr., Fang B, Dale AM, Lyons MJ, & Franz CE (2019). Influence of young adult cognitive ability and additional education on later-life cognition. Proceedings of the National Academy of Sciences, 116(6), 2021–2026. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Kukull WA, Higdon R, Bowen JD, McCormick WC, Teri L, Schellenberg GD, van Belle G, Jolley L, & Larson EB (2002). Dementia and Alzheimer disease incidence: A prospective cohort study. Archives of Neurology, 59(11), 1737–1746. [DOI] [PubMed] [Google Scholar]
  38. Lachman ME, Agrigoroaei S, Murphy C, & Tun PA (2010). Frequent cognitive activity compensates for education differences in episodic memory. The American Journal of Geriatric Psychiatry, 18(1), 4–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Lareau A (2011). Unequal childhoods: Class, race, and family life: University of California Press. [Google Scholar]
  40. Lartey J (2019). The perfectly legal—But immoral—Ways rich kids get into top colleges. The Guardian. https://www.theguardian.com/us-news/2019/mar/13/rich-kids-top-college-admissions [Google Scholar]
  41. Lazear EP (2001). Educational production. The Quarterly Journal of Economics, 116(3), 777–803. [Google Scholar]
  42. Lièvre A, Alley D, & Crimmins EM (2008). Educational differentials in life expectancy with cognitive impairment among the elderly in the United States. Journal of Aging and Health, 20(4), 456–477. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Link BG, & Phelan J (1995). Social conditions as fundamental causes of disease. Journal of Health and Social Behavior, 80–94. [PubMed] [Google Scholar]
  44. Lyu J, & Burr JA (2016). Socioeconomic status across the life course and cognitive function among older adults: An examination of the latency, pathways, and accumulation hypotheses. Journal of Aging and Health, 28(1), 40–67. [DOI] [PubMed] [Google Scholar]
  45. Mackenbach JP, Kulhánová I, Bopp M, Deboosere P, Eikemo TA, Hoffmann R, Kulik MC, Leinsalu M, Martikainen P, Menvielle G, Regidor E, Wojtyniak B,Östergren O, & Lundberg O, & EURO-GBD-SE Consortium. (2015). Variations in the relation between education and cause-specific mortality in 19 European populations: A test of the “fundamental causes” theory of social inequalities in health. Social Science & Medicine, 127, 51–62. [DOI] [PubMed] [Google Scholar]
  46. Mantri S, Nwadiogbu C, Fitts W, & Dahodwala N (2019). Quality of education impacts late-life cognition. International Journal of Geriatric Psychiatry, 34(6), 855–862. [DOI] [PubMed] [Google Scholar]
  47. Mayeda ER, Glymour MM, Quesenberry CP, & Whitmer RA (2016). Inequalities in dementia incidence between six racial and ethnic groups over 14 years. Alzheimer’s & Dementia, 12(3), 216–224. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Mirowsky J, & Ross C (1989). Social causes of psychological distress. Aldine de Gruyter. [Google Scholar]
  49. Moorman SM, Greenfield EA, & Garcia S (2019). School context in adolescence and cognitive functioning 50 years later. Journal of Health and Social Behavior, 60(4), 493–508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Reardon SF, Baker R, & Klasik D (2012). Race, income, and enrollment patterns in highly selective colleges, 1982–2004. Center for education policy analysis, Stanford University. http://cepa.stanford.edu/content/race-income-and-enrollmentpatterns-highly-selective-colleges-1982-2004 [Google Scholar]
  51. Richards M, James S-N, Sizer A, Sharma N, Rawle M, Davis DH, & Kuh D (2019). Identifying the lifetime cognitive and socioeconomic antecedents of cognitive state: Seven decades of follow-up in a British birth cohort study. BMJ Open, 9(4), e024404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Ross CE, & Mirowsky J (1999). Refining the association between education and health: The effects of quantity, credential, and selectivity. Demography, 36(4), 445–460. [PubMed] [Google Scholar]
  53. Rubin DB (1987). Multiple imputation for survey nonresponse.
  54. Sacerdote B (2011). Peer effects in education: How might they work, how big are they and how much do we know thus far? In Hanushek E, Machin S, & Woessman L, (Eds.), Handbook of the economics of education (Vol. 3, pp. 249–277). Elsevier. [Google Scholar]
  55. Sando SB, Melquist S, Cannon A, Hutton M, Sletvold O, Salt-vedt I, White LR, Lydersen S, & Aasly J (2008). Risk-reducing effect of education in Alzheimer’s disease. International Journal of Geriatric Psychiatry: A Journal of the Psychiatry of Late Life and Allied Sciences, 23(11), 1156–1162. [DOI] [PubMed] [Google Scholar]
  56. Schafer MH, Wilkinson LR, & Ferraro KF (2013). Childhood (mis) fortune, educational attainment, and adult health: Contingent benefits of a college degree? Social Forces, 91(3), 1007–1034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Schooler C, Mulatu MS, & Oates G (1999). The continuing effects of substantively complex work on the intellectual functioning of older workers. Psychology and Aging, 14(3), 483. [DOI] [PubMed] [Google Scholar]
  58. Stern Y (2009). Cognitive reserve. Neuropsychologia, 47(10), 2015–2028. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Stolzenberg RM, & Relles DA (1997). Tools for intuition about sample selection bias and its correction. American Sociological Review, 62, 494–507. [Google Scholar]
  60. Sullivan A, & Brown M (2013). Social inequalities in cognitive scores at age 16: The role of reading (CLS Working Papers no. 13/10). https://discovery.ucl.ac.uk/id/eprint/1473708/. [Google Scholar]
  61. Tholen G, Brown P, Power S, & Allouch A (2013). The role of networks and connections in educational elites’ labour market entrance. Research in Social Stratification and Mobility, 34, 142–154. [Google Scholar]
  62. Wechsler D (1997). Wechsler memory scale (3rd ed.). Psychological Corporation. [Google Scholar]
  63. Wimo A, Guerchet M, Ali G-C, Wu Y-T, Prina AM, Winblad B, Jönsson L, Liu Z, & Prince M (2017). The worldwide costs of dementia 2015 and comparisons with 2010. Alzheimer’s & Dementia, 13(1), 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Winship C, & Mare RD (1992). Models for sample selection bias. Annual Review of Sociology, 18(1), 327–350. [Google Scholar]
  65. Winston G, & Zimmerman D (2004). Peer effects in higher education college choices: The economics of where to go, when to go, and how to pay for it (pp. 395–424). University of Chicago Press. [Google Scholar]
  66. Zahodne LB, Stern Y, & Manly JJ (2015). Differing effects of education on cognitive decline in diverse elders with low versus high educational attainment. Neuropsychology, 29(4), 649–657. [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

Supplemental material
NIHMS1826299-supplement-Supplemental_material.pdf (59.9KB, pdf)

RESOURCES