Skip to main content
NCBI home page
Karger Author's Choice logoLink to Karger Author's Choice
. 2023 Dec 12;70(3):318–326. doi: 10.1159/000535717

The Association between Education and Cognitive Performance Varies at Different Levels of Cognitive Performance: A Quantile Regression Approach

Johan Rehnberg a,✉, Martijn Huisman b,c,d, Stefan Fors a,e, Anna Marseglia f, Almar Kok b,c
PMCID: PMC10911170  PMID: 38086341

Abstract

Introduction

Educational differences in cognitive performance among older adults are well documented. Studies that explore this association typically estimate a single average effect of education on cognitive performance. We argue that the processes that contribute to the association between education and cognitive performance are unlikely to have equal effects at all levels of cognitive performance. In this study, we employ an analytical approach that enables us to go beyond averages to examine the association between education and five measures of global and domain-specific cognitive performance across the outcome distributions.

Methods

This cross-sectional study included 1,780 older adults aged 58–68 years from the Longitudinal Aging Study Amsterdam. Conditional quantile regression was used to examine variation across the outcome distribution. Cognitive outcomes included Mini-Mental State Examination (MMSE) score, crystallized intelligence, information processing speed, episodic memory, and a composite score of global cognitive performance.

Results

The results showed that the associations between education and different cognitive measures varied across the outcome distributions. Specifically, we found that education had a stronger association with crystallized intelligence, MMSE, and a composite cognitive performance measure in the lower tail of performance distributions. The associations between education and information processing speed and episodic memory were uniform across the outcome distributions.

Conclusion

Larger associations between education and some domains of cognitive performance in the lower tail of the performance distributions imply that inequalities are primarily generated among individuals with lower performance rather than among average and high performers. Additionally, the varying associations across some of the outcome distributions indicate that estimating a single average effect through standard regression methods may overlook variations in cognitive performance between educational groups. Future studies should consider heterogeneity across the outcome distribution.

Keywords: Conditional quantile regression, Crystallized intelligence, Episodic memory, Mini-Mental State Examination, Educational inequalities

Introduction

More schooling is associated with higher cognitive performance [1–4]. Evidence from quasi-experimental studies suggests that this relationship may partly be caused by educational attainment increasing cognitive ability, and not only as a consequence of selection into higher education based on prior cognitive ability [5–8]. This relationship is maintained in old age, and education is often cited as an early-life protective factor against cognitive impairment and dementia [9]. Studies that explore the relationship between education and cognitive performance typically estimate one average effect of education on the cognitive outcome of interest. While the average effect of an exposure on an outcome can be a useful metric, it provides limited information of how the effect of an exposure might differ across various levels of the outcome. Recent studies have indicated that educational differences in cognitive performance among younger people [10] and adults [11] may vary at different levels of performance distribution, with larger differences found at low levels of performance. If these findings translate to older adults, the seemingly protective effect of education on cognitive aging could be weak or even nonexistent for persons with higher cognitive performance. In this study, we employ an analytical approach that enables us to go beyond averages to examine educational differences in five measures of global and domain-specific cognitive performance across the cognitive performance distribution in a sample of older adults in the Netherlands.

One of the most extensively documented processes that contribute to the association between education and late-life cognition is the relationship between higher cognitive ability in early life and a greater likelihood of attaining higher education and socioeconomic success [12–14]. However, while cognitive ability is a substantial factor, educational choices are also influenced by other factors [15], and not all individuals with high cognitive ability pursue higher education. Consequently, there is likely to be considerable heterogeneity in cognitive ability among lower-educated individuals. Conversely, among higher educated individuals, selection into education based on prior cognitive ability is likely to result in a more homogenous group in terms of cognitive ability compared to lower-educated groups. Selection processes that occur in early life could, therefore, generate distributions of cognitive performance that vary across educational groups in adulthood and old age.

Besides prior selection on cognitive ability, higher educational attainment may increase the level of cognitive performance by systematically and frequently offering cognitively stimulating tasks, challenges, and social interactions. It has been proposed that a high level of cognition achieved in younger and middle ages may delay the time it takes for cognitive decline to reach the threshold when it starts affecting daily functioning in older ages [16]. Alternatively, education has been proposed to increase resilience against age-related neuropathological changes [17]. The cognitive reserve hypothesis suggests that longer education can enhance more adaptable functional brain processes, enabling higher-educated persons to compensate for the negative effects of age-related neuropathological changes, thus postponing the clinical manifestation of cognitive impairment [17, 18]. Education could also protect cognition by enabling the absorption of age-related neuropathological changes, resulting in no or minimal brain alterations and preserved cognition (brain maintenance hypothesis) [19, 20]. Regardless of the specific resilience mechanisms involved, if the effects of brain aging on cognitive performance are delayed in higher educated groups, educational differences among older persons may be larger at low levels of performance.

As described above, there are several reasons to expect differences in the distribution of cognitive performance between educational groups. This variation cannot be detected by only describing average differences in cognitive performance between different educational groups as is typically done in studies using traditional regression methods. The limitations of the mean effects have previously been highlighted in prior studies from various research fields, such as the association between education and BMI [21, 22] and the motherhood wage-penalty [23]. Recently, Ford and Leist [11] showed that income and educational differences in the Mini-Mental State Examination (MMSE) were larger at the bottom of MMSE distribution. Considering heterogeneity in cognitive performance across different educational groups is essential for several reasons. First, from a methodological perspective, relying on average scores per educational group may not accurately represent educational inequalities in cognitive performance for everyone in the data. Second, from a conceptual perspective, educational differences in cognitive performance are likely generated by different processes at different levels of performance. For example, education may have a larger impact on persons with lower initial levels of cognitive performance than those with higher levels. Third, from a policy perspective, to understand and ultimately reduce educational inequalities in cognitive performance among older adults, it is crucial to identify groups where the greatest potential for improvement exists. Therefore, analyzing variation beyond average differences will provide new insights into understanding and addressing educational differences in cognitive performance among older adults.

In this paper, we address the possibility that the association between education and cognitive performance varies across cognitive performance distribution among older adults. We aim to answer the question: Does the association between educational and cognitive performance vary at different levels of cognitive performance among older adults? To do so, we apply quantile regression to estimate the association between education and five measures of global and domain-specific cognitive performance at each 5th percentile of the outcome distribution using data from the Longitudinal Aging Study Amsterdam (LASA).

Materials and Methods

This study utilizes data on older adults aged between 58 and 68 from the LASA that were collected by face-to-face and medical interviews. LASA is based on representative samples of three cohorts of older adults in the Netherlands that were initially included in 1992–1993, 2002–2003, and 2012–2013, respectively. Response rates in LASA are around 60% for the initial samples with higher rates (around 80 percent) for the subsequent follow-up waves [24].

For this study, we selected the second wave that each of the three cohorts participated in, as this was the wave in which all measures of cognitive performance were included. The first cohort, born between 1926 and 1930, was measured in 1995–1996 (n = 303). The second cohort, born between 1938 and 1947, was measured in 2005–2006 (n = 798), and the third cohort, born between 1948 and 1957, was measured in 2015–2016 (n = 679). These three cohorts were pooled to generate one sample of 1,780 persons aged between 58 and 68. Persons who had full information on all variables were included in the analytical sample. No exclusion criteria were applied based on cognitive impairment; however, due to the relatively young age of the respondents, the sample included few individuals with severe cognitive impairment (see Fig. 1 for the distribution of MMSE scores among respondents).

Fig. 1.

Fig. 1.

Open in a new tab

Density distributions of five cognitive performance measures by low and intermediate-high education (n = 1,780).

Dependent Variables

Crystallized intelligence was assessed with a vocabulary test that was based on a subtest of the Groninger Intelligence Test [25]. The respondents were presented with 20 words of increasing difficulty that they then had to pair with a correct synonym out of five alternatives. The total score ranged between 0 and 20, with higher scores indicating a higher level of intelligence.

Global cognitive functioning was assessed with the MMSE and was measured with a 30-point screening questionnaire that covered seven cognitive domains [26]. The total score ranged between 0 and 30, with higher scores indicating better cognitive functioning.

Information processing speed was assessed by a letter substitution task [27], adapted from the Alphabet Coding Task-15 [28]. In this task, the respondent is first presented with two rows of characters; each character in the upper row is matched with a character in the bottom row. The respondent is then presented with another two rows, where only the upper row is filled with characters, and the respondent must fill in the matching character in the bottom row using the same pattern that was presented in the first two rows. The respondent had 1 min to complete as many combinations as possible, and this procedure was performed three times. The mean value of the three trials was used in the analysis.

Episodic memory was assessed with a Dutch version of the auditory verbal learning test. The respondents were presented with 15 words that they were then asked to recall immediately after, and this was done three times using the same list of words. The sum score of these three trials was used in the analysis. The total score ranged between 0 and 45.

These four cognitive performance variables were Z-standardized by taking each respondent’s score minus the mean of the sample divided by the standard deviation of the sample, resulting in measures where the mean is zero and the standard deviation equals one. More detailed information about the cognitive tests is available at the LASA website (lasa-vu.nl).

To capture a more complete measure of cognitive performance that resembles a person’s general intelligence (g-score), a composite score was created based on the four different indicators, namely, crystallized intelligence, MMSE, information processing speed, and episodic memory. The composite score was calculated from the four separate cognitive performance indicators using principal component analysis. The principal component analysis indicated that the four measures all contributed significantly and substantially to the composite score (all loadings were above 0.5; online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000535717). Like the four separate cognitive performance measures, the composite score had a standardized mean of 0 and a standard deviation of 1.

Independent Variable

Education was assessed using information on the respondents’ educational achievements, which were initially recorded as categorical levels in accordance with the Dutch education system. In order to facilitate a more intuitive interpretation of this variable, these educational categories were converted into a scale denoting years of education ranging from 5 to 18 or more years of formal education.

Covariates

Three variables were adjusted for in the regressions: age measured in years, a binary variable indicating sex, and a three-category variable indicating the birth cohort.

Analytical Strategy

To show the distribution of each cognitive performance measure across educational groups, we dichotomized the educational variable into low education (10 years or less) and intermediate-to-high education (11 years or more). These two groups comprised around half the sample each (Table 1).

Table 1.

Sample descriptive statistics and nonstandardized cognitive performance scores by low and intermediate-high education (n = 1,780)

Minimum Median Maximum
Age, years 57.7 64.22 68.9
Education, years 5 10 18
Number of respondents %
Education
 Low (<11 years) 921 52
 High (>10 years) 859 48
Sex
 Men 862 48
 Women 918 52
Cohort
 1926–1930 303 17
 1938–1947 798 45
 1948–1957 679 38
Total sample Low education High education
Crystallized intelligence
 Mean 13.4 12.1 14.7
 SD 3.5 3.5 3.0
 Min-max 0–20.0 1.0–20.0 0–20.0
Information processing speed
 Mean 29.5 28.1 31.1
 SD 6.4 6.7 5.8
 Min-max 5.0–51.0 5.0–46.7 11.3–51.0
MMSE
 Mean 28.2 27.8 28.5
 SD 1.8 2.1 1.4
 Min-max 13.0–30.0 13.0–30.0 21.0–30.0
Episodic memory
 Mean 23.5 22.2 24.9
 SD 6.2 6.0 6.2
 Min-max 0–42.0 0–42.0 0–41.0
Composite score
 Mean 0.0 −0.3 0.4
 SD 1.0 1.0 0.8
 Min-max −5.5–2.3 −5.5–2.2 −2.8–2.3
Open in a new tab

We used conditional quantile regression (CQR) to estimate the association between education and cognitive performance at different percentiles of the cognitive performance distribution. This method is appropriate for exploring whether the effect of interest is uniform or varies across the conditional outcome distribution [29, 30]. The statistical models were estimated using the quantreg package version 5.95 for R. In all regression models, the continuous education variable described above was used. The quantile regression estimates were compared against the average educational effect obtained from a linear regression model with the same model specification.

Analyses were performed for each cognitive performance indicator separately. To remove the influences of the basic demographic compositional differences across the educational groups, we adjust the models for age, sex, and birth cohort. No additional variables were adjusted for since the goal was to describe educational differences in cognitive performance rather than to model causal effect estimates. To test whether educational differences were present at all levels of education, a sensitivity analysis using an educational variable with four categories was performed. The results were robust and showed that associations similar to the main results were present at all educational levels (online suppl. Fig. 1; online suppl. Table 3).

Results

Table 1 presents sample characteristics and the distribution of five cognitive performance scores per educational group. The analytical sample consisted of 1,780 participants, with ages ranging from 57.7 to 68.9 years and a median age of 64.2. Educational attainment ranged from 5 to 18 years, with a median of 10 years. The sample was composed of 17% of individuals from the first cohort born between 1926 and 1930 and surveyed in 1995–1996, 45% from the second cohort born between 1938 and 1947 and surveyed in 2005–2006, and 38% from the third cohort born between 1948 and 1957 and surveyed in 2015–2016.

The four cognitive performance scores (nonstandardized in Table 1) and the standardized composite score are presented for the entire sample and stratified by the educational variable dichotomized in two categories: low and intermediate-high education. Cognitive scores were consistently lower in the low-education group compared to the intermediate-high-education group. The standard deviation was substantially larger in the low-education group for all cognitive measures, except for episodic memory, which showed similar variability across both education groups.

In Figure 1, the density distributions for the standardized cognitive performance scores are presented for individuals with low and intermediate-high education. Individuals with lower education exhibit lower cognitive performance scores across all points of distributions. Furthermore, in crystallized intelligence, information processing speed, and the composite measure, the range of the distribution was wider among the low education group compared to the narrower distribution seen in the intermediate-high education group.

Figure 2 shows estimates from CQR and linear regression (LR) of the continuous education variable on four cognitive performance measures and the composite score, adjusted for age, sex, and birth cohort (see online suppl. Table 2 for full regression estimates). The CQR estimate indicates the change in cognitive performance for each additional year of education at a specified quantile, while the LR estimate represents the average change in cognitive performance for each additional year of education. The LR estimates showed that respondents with higher education had on average higher cognitive performance in all measures. Specifically, each additional year of education was associated with an average increase in standard deviation of 0.137 in the composite score, 0.146 in crystallized intelligence, 0.082 in episodic memory, 0.091 in information processing speed, and 0.084 in MMSE.

Fig. 2.

Fig. 2.

Open in a new tab

Conditional quantile regression (CQR) and linear regression (LR) estimates of education on five cognitive performance measures estimated at each 5th quantile of the outcome distribution (n = 1,780).

The CQR showed varying effect sizes at different levels of the outcome. Specifically, for the composite score, the CQR estimates showed larger associations between education and the composite score at the lower tail of the distribution, at approximately 0.2 standard deviations, compared to smaller associations at the higher tail, at around 0.08–0.09 standard deviations. Across the entire composite score distribution, the association between education and the outcome became progressively weaker at higher quantiles. Similarly, in crystallized intelligence, the CQR estimate for education was 0.234 standard deviations at the lowest quantile (5th) and 0.094 at the highest percentile (95th), indicating that educational differences in crystallized intelligence were more pronounced at lower values of crystallized intelligence and smaller at higher values. For MMSE, the largest educational differences were observed at the lower quantiles, while no differences were observed at the highest three quantiles.

The association between education and information processing speed showed little variation across the outcome distribution, with the CQR estimates fluctuating around the effect size for education obtained from the LR estimates and overlapping confidence intervals. Likewise, the association between education and episodic memory was consistent across the entire outcome distribution, with the CQR estimates closely approximating the effect size from the LR at all quantiles.

Discussion

The findings from this descriptive cross-sectional study showed that the association between education and cognitive performance varied across the outcome distribution in three out of five measures. The association was larger in the lower tail of the outcome distribution in crystallized intelligence, MMSE, and in a composite score. In contrast, the association between education and information processing speed and episodic memory was relatively uniform across the outcome distribution. Variation across the outcome scale and variation between measures of cognitive performance suggests a complex relationship between education and different cognitive domains.

The complex relationship between education and cognitive performance has been explored in previous studies. For example, Ritchie et al. [3, 31] showed that the general relationship between higher education and higher g-factor scores was driven by varying domain-specific cognitive abilities. Our study supports this finding by demonstrating differences in the size of the association across five measures of global and domain-specific cognitive performance and education.

The variation in the associations between education and cognitive performance across the outcome scale also suggests that the mechanisms underlying educational differences in cognitive performance may differ at different levels of performance. This observation is consistent with both the causal and selection-based accounts of the relationship between education and cognitive performance. Selection into education based on cognitive ability may be especially strong among individuals with lower performance, which will result in larger educational differences at low levels of performance. Additionally, the hypothesis that additional education enhances brain resilience and is protective against age-related brain changes and pathology [17] may be especially effective among individuals with lower initial levels of cognitive performance, yielding diminishing protective returns at higher levels of cognitive performance. Future studies exploring these processes should consider the possibility of heterogeneous effects of education on cognitive aging.

The association between education and crystallized intelligence was found to be the largest and most variable across the outcome scale among the separate cognitive dimensions. This result aligns with the notion that crystallized abilities are strongly connected to learning, formal education, and acculturation [32]. The weakened relationship between education and crystallized intelligence as it reached the upper end of intelligence distribution hints at a possible saturation point where the influence of education levels off. However, it is important to note that the association still persisted at higher levels of education, albeit with a reduced strength. Conversely, information processing speed is a fluid ability that is less related to experience or acquired knowledge and more related to inherent abilities and, among older adults, reflective of neurobiological decline [3, 33]. Consistent with this, we found a weaker and uniform association between education and information processing speed across the outcome distribution. Like information processing speed, episodic memory is age-sensitive and declines from mid-life and onward [34, 35]. We observed a weaker association between episodic memory and education, with no variation in educational differences across levels of episodic memory. However, it should be noted that the relatively young sample (age 58–68) may have experienced only mild declines in information processing speed and episodic memory. Nyberg et al. [34] noted that there was substantial interindividual variation in memory performance, and some individuals may show preserved memory functioning in their 70s. Therefore, it is possible that with advancing age, the uniform educational differences we observed for fluid abilities may change.

MMSE is a commonly used clinical screening tool for dementia, but it may lack sensitivity to milder cognitive deficits due to ceiling effects, particularly in individuals with higher education levels [36]. Therefore, the lack of association between education and MMSE at higher levels of MMSE may be a result of such ceiling effects, which were also noted by Ford and Leist [11]. Additionally, as discussed above, the sample in our study was relatively young (aged 58–68), and in these ages, few persons have experienced cognitive decline to the point of scoring in the lower half of the MMSE scale. Nonetheless, the larger association between education and MMSE observed at the lower end of the MMSE scale may suggest an earlier onset of severe cognitive impairment among individuals with lower levels of education, as has been previously shown [37].

Throughout the 20th century, later-born cohorts have shown improvements in standardized intelligence test scores compared to earlier-born cohorts [38]. Some of these gains can be attributed to the expansion and improvement of education [39]. Additionally, Hessel et al. [40] demonstrated that improvements in word recall among older individuals were starting to plateau in European countries with longer histories of higher socioeconomic development compared to less developed countries. Our findings of weakened associations between education and some measures of cognitive performance at higher performance levels raise the question whether cohort gains in cognitive performance are of equal size across cognitive distribution. Future research should investigate whether improvements in cognitive performance occur across the entire spectrum of cognitive performance or if they are confined to specific segments of distribution.

The present study has several limitations. One limitation is that we included only one age group and applied a cross-sectional design. The variations in the relationship between education and cognitive performance that we observed may be different, either weaker or exacerbated, in ages where more severe cognitive decline has occurred. Future studies should examine whether there are similar variations across the cognitive performance scale in older ages and whether the rate of decline varies at different performance levels.

Another limitation concerns the measures used for assessing cognitive performance and educational attainment. In this study, each dimension of cognitive performance is assessed using only one measure, which may not comprehensively reflect a person’s cognitive abilities within that specific domain. For example, crystallized intelligence encompasses the influences of learning, experience, and acculturation, and while it is a prevalent practice to employ verbal tests to operationalize crystallized intelligence, such assessments may not encompass the entire concept [32].

Similarly, education is operationalized as years of formal education that respondents completed several decades ago. This approach does not consider the ongoing process of learning and informal education that individuals may have pursued throughout their life course. Despite these inherent measurement limitations, our findings can offer valuable insights into the association between education and cognitive outcomes both for researchers that employ similar measurements as well as studies that seek to apply more complex measurement models.

Conclusion

Few, if any, studies have examined whether the association between education and cognitive performance varies across performance distribution among older adults. Our study adds to this literature by demonstrating that for three of the five cognitive measures assessed, educational differences were greater at the lower tail of performance distribution. As a result, relying solely on an average association between education and cognitive outcomes will obscure important variations in this association that exist across different levels of performance.

Furthermore, the observation of larger associations between education and cognitive performance in the lower tail of performance distributions implies that inequalities are primarily generated among individuals with lower performance rather than among average and high performers. This finding has potential policy implications as it suggests that the greatest potential for reducing educational inequalities in cognitive performance may be among individuals with lower performance levels. Future studies investigating the relationship between education and cognition should consider heterogeneity across the outcome scale and explore the mechanisms that underlie this heterogeneity.

Acknowledgments

The Longitudinal Aging Study Amsterdam is supported by a grant from the Netherlands Ministry of Health, Welfare, and Sport, Directorate of Long-Term Care. The baseline data collection for the LASA third cohort was financially supported by the Netherlands Organization for Scientific Research (NWO) in the framework of the project “New cohorts of young-old in the 21st century” (file number 480-10-014).

Statement of Ethics

The use of data for the purposes of this study was granted by the Swedish Central Ethical Review Board (Dnr 2021-03895). Written informed consent was obtained for all participants in the Longitudinal Aging Study Amsterdam at the time of the interviews.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This work was supported by the Swedish Research Council for Health, Working Life and Welfare, grant number 2020-00071.

Author Contributions

J.R. and S.F. contributed to the conception of the study design. J.R. drafted the first manuscript of the study and performed the statistical analyses. J.R., M.H., S.F., A.M., and A.K. contributed to development of the final analyses, results, and theoretical design and contributed to the revision of the manuscript.

Funding Statement

This work was supported by the Swedish Research Council for Health, Working Life and Welfare, grant number 2020-00071.

Data Availability Statement

The data used in this study contains sensitive personal information that cannot be made publicly available. The data from the Longitudinal Aging Study Amsterdam (LASA) database are available for use for specific research that are evaluated and approved by the LASA steering group. Documentation of data handling and statistical procedures that were used in this study can be obtained from the corresponding author. The current study was not preregistered.

Supplementary Material

References

  • 1. Caamaño-Isorna F, Corral M, Montes-Martínez A, Takkouche B. Education and dementia: a meta-analytic study. Neuroepidemiology. 2006;26(4):226–32. [DOI] [PubMed] [Google Scholar]
  • 2. Karp A, Kåreholt I, Qiu C, Bellander T, Winblad B, Fratiglioni L. Relation of education and occupation-based socioeconomic status to incident Alzheimer’s disease. Am J Epidemiol. 2004;159(2):175–83. [DOI] [PubMed] [Google Scholar]
  • 3. Ritchie SJ, Bates TC, Der G, Starr JM, Deary IJ. Education is associated with higher later life IQ scores, but not with faster cognitive processing speed. Psychol Aging. 2013;28(2):515–21. [DOI] [PubMed] [Google Scholar]
  • 4. Wilson RS, Hebert LE, Scherr PA, Barnes LL, Mendes de Leon CF, Evans DA. Educational attainment and cognitive decline in old age. Neurology. 2009;72(5):460–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Brinch CN, Galloway TA. Schooling in adolescence raises IQ scores. Proc Natl Acad Sci U S A. 2012;109(2):425–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Hamad R, Elser H, Tran DC, Rehkopf DH, Goodman SN. How and why studies disagree about the effects of education on health: a systematic review and meta-analysis of studies of compulsory schooling laws. Soc Sci Med. 2018;212:168–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Lager A, Seblova D, Falkstedt D, Lövdén M. Cognitive and emotional outcomes after prolonged education: a quasi-experiment on 320,182 Swedish boys. Int J Epidemiol. 2017;46(1):303–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Meghir C, Palme M, Simeonova E. Education, cognition and Health: evidence from a social experiment. Cambridge, MA: National Bureau of Economic Research; 2013. [Google Scholar]
  • 9. Livingston G, Sommerlad A, Orgeta V, Costafreda SG, Huntley J, Ames D, et al. Dementia prevention, intervention, and care. Lancet. 2017;390(10113):2673–734. [DOI] [PubMed] [Google Scholar]
  • 10. Hajovsky DB, Villeneuve EF, Schneider WJ, Caemmerer JM. An alternative approach to cognitive and achievement relations research: an introduction to quantile regression. J Pediatr Neuropsychol. 2020;6(2):83–95. [Google Scholar]
  • 11. Ford KJ, Leist AK. Returns to educational and occupational attainment in cognitive performance for middle-aged South Korean men and women. Gerontol Geriatr Med. 2021;7:23337214211004366. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Deary IJ, Strand S, Smith P, Fernandes C. Intelligence and educational achievement. Intelligence. 2007;35(1):13–21. [Google Scholar]
  • 13. Fors S, Torssander J, Almquist YB. Is childhood intelligence associated with coexisting disadvantages in adulthood? Evidence from a Swedish cohort study. Adv Life Course Res. 2018;38:12–21. [Google Scholar]
  • 14. Strenze T. Intelligence and socioeconomic success: a meta-analytic review of longitudinal research. Intelligence. 2007;35(5):401–26. [Google Scholar]
  • 15. Ganzach Y. Parents’ education, cognitive ability, educational expectations and educational attainment: interactive effects. Br J Educ Psychol. 2000;70 (Pt 3)(3):419–41. [DOI] [PubMed] [Google Scholar]
  • 16. Lövdén M, Fratiglioni L, Glymour MM, Lindenberger U, Tucker-Drob EM. Education and cognitive functioning across the life span. Psychol Sci Public Interest. 2020;21(1):6–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Stern Y, Arenaza‐Urquijo EM, Bartrés-Faz D, Belleville S, Cantilon M, Chetelat G, et al. Whitepaper: defining and investigating cognitive reserve, brain reserve, and brain maintenance. Alzheimers Dement. 2020;16(9):1305–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Barulli D, Stern Y. Efficiency, capacity, compensation, maintenance, plasticity: emerging concepts in cognitive reserve. Trends Cogn Sci. 2013;17(10):502–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Cabeza R, Albert M, Belleville S, Craik FI, Duarte A, Grady CL, et al. Maintenance, reserve and compensation: the cognitive neuroscience of healthy ageing. Nat Rev Neurosci. 2018;19(11):701–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Stern Y, Barnes CA, Grady C, Jones RN, Raz N. Brain reserve, cognitive reserve, compensation, and maintenance: operationalization, validity, and mechanisms of cognitive resilience. Neurobiol Aging. 2019;83:124–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Bann D, Fitzsimons E, Johnson W. Determinants of the population health distribution: an illustration examining body mass index. Int J Epidemiol. 2020;49(3):731–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Liu SY, Kawachi I, Glymour MM. Education and inequalities in risk scores for coronary heart disease and body mass index: evidence for a population strategy. Epidemiology. 2012;23(5):657–64. [DOI] [PubMed] [Google Scholar]
  • 23. Budig MJ, Hodges MJ. Differences in disadvantage: variation in the motherhood penalty across white women’s earnings distribution. Am Sociol Rev. 2010;75(5):705–28. [Google Scholar]
  • 24. Hoogendijk EO, Deeg DJH, de Breij S, Klokgieters SS, Kok AAL, Stringa N, et al. The Longitudinal Aging Study Amsterdam: cohort update 2019 and additional data collections. Eur J Epidemiol. 2020;35(1):61–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Luteijn F, Ploeg FAE. GIT: groninger intelligentie test:[handleiding]. Swets & Zeitlinger; 1983. [Google Scholar]
  • 26. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12(3):189–98. [DOI] [PubMed] [Google Scholar]
  • 27. Piccinin AM, Rabbitt P. Contribution of cognitive abilities to performance and improvement on a substitution coding task. Psychol Aging. 1999;14(4):539–51. [DOI] [PubMed] [Google Scholar]
  • 28. Savage R. Alphabet coding task 15. Perth, Western Australia: Murdoch University; 1984. Unpublished manuscript. [Google Scholar]
  • 29. Borgen NT, Haupt A, Wiborg ØN. Quantile regression estimands and models: revisiting the motherhood wage penalty debate. Eur Socio Rev. 2022;39(2):317–31 jcac052. [Google Scholar]
  • 30. Koenker R, Hallock KF. Quantile regression. J Econ Perspect. 2001;15(4):143–56. [Google Scholar]
  • 31. Ritchie SJ, Bates TC, Deary IJ. Is education associated with improvements in general cognitive ability, or in specific skills? Dev Psychol. 2015;51(5):573–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Schipolowski S, Wilhelm O, Schroeders U. On the nature of crystallized intelligence: the relationship between verbal ability and factual knowledge. Intelligence. 2014;46:156–68. [Google Scholar]
  • 33. Tucker-Drob EM, Johnson KE, Jones RN. The cognitive reserve hypothesis: a longitudinal examination of age-associated declines in reasoning and processing speed. Dev Psychol. 2009;45(2):431–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Nyberg L, Lövdén M, Riklund K, Lindenberger U, Bäckman L. Memory aging and brain maintenance. Trends Cogn Sci. 2012;16(5):292–305. [DOI] [PubMed] [Google Scholar]
  • 35. Old SR, Naveh-Benjamin M. Differential effects of age on item and associative measures of memory: a meta-analysis. Psychol Aging. 2008;23(1):104–18. [DOI] [PubMed] [Google Scholar]
  • 36. Hoops S, Nazem S, Siderowf A, Duda J, Xie S, Stern M, et al. Validity of the MoCA and MMSE in the detection of MCI and dementia in Parkinson disease. Neurology. 2009;73(21):1738–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Terrera GM, Minett T, Brayne C, Matthews FE. Education associated with a delayed onset of terminal decline. Age Ageing. 2014;43(1):26–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Trahan LH, Stuebing KK, Fletcher JM, Hiscock M. The Flynn effect: a meta-analysis. Psychol Bull. 2014;140(5):1332–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Beller J, Kuhlmann BG, Sperlich S, Geyer S. Secular improvements in cognitive aging: contribution of education, health, and routine activities. J Aging Health. 2022;34(6–8):807–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Hessel P, Kinge JM, Skirbekk V, Staudinger UM. Trends and determinants of the Flynn effect in cognitive functioning among older individuals in 10 European countries. J Epidemiol Community Health. 2018;72(5):383–9. [DOI] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

Supplementary_material-Suppl.1-s1.pdf (253.7KB, pdf)

Data Availability Statement

The data used in this study contains sensitive personal information that cannot be made publicly available. The data from the Longitudinal Aging Study Amsterdam (LASA) database are available for use for specific research that are evaluated and approved by the LASA steering group. Documentation of data handling and statistical procedures that were used in this study can be obtained from the corresponding author. The current study was not preregistered.

Articles from Gerontology are provided here courtesy of Karger Publishers

RESOURCES