Development over Time in Cognitive Function among European 55-69-Year-Olds from 2006 to 2015, and Differences of Region, Gender, and Education Development over Time in Cognitive Function among European 55-69-Year-Olds from 2006 to 2015, and Differences of Region, Gender, and Education Ying Zhou Abstract: With populations rapidly aging, the development over time in the cognitive function among the elderly approaching or reaching retirement is important for successful aging at work and planning pension policies. However, few studies in this fi eld focus on this age group. This study characterizes time trends in cognitive function among 55-69-year-old Europeans from 2006 to 2015, and compares these trends by region, gender, and education. This study analyzes 40,689 subjects in Waves 2, 4, 5 and 6 of the Survey of Health, Aging and Retirement in Europe (SHARE) covering ten countries. Cognitive function was measured by Recall and Verbal Fluency. Educational levels were classifi ed by quartiles. A Generalized Estimating Equation (GEE) model was used to explore the association between cognitive function and development over time after controlling for confounders. Further stratifi cation analysis using GEE models was conducted, stratifi ed by region, gender and education. Cognitive function improved signifi cantly in southern and central Europe over the observed timeframe, whereas it did not in northern Europe. Those with relative low levels of formal education displayed the most rapid increases in cognitive function in southern and central Europe. Among those with lower education in southern Europe, males’ cognitive function improved more quickly than females’. The improvement of cognitive function at ages 55-69 in southern and central Europe may contribute to continuing engagement with productive activities in old age. Educational interventions for people with lower levels of education may be most effective in achieving such engagement. This paper extends the literature on the development over time in the cognitive function among the elderly close to retirement age in Europe by analysing southern, central and northern Europe, as well as differences by region, gender and education. The results may provide evidence for planning pension policies and educational interventions. Keywords: Aging · Cognitive function · Education · Flynn effect Comparative Population Studies Vol. 47 (2022): 119-142 (Date of release: 29.03.2022) Federal Institute for Population Research 2022 URL: www.comparativepopulationstudies.de DOI: https://doi.org/10.12765/CPoS-2022-04 URN: urn:nbn:de:bib-cpos-2022-04en0 • Ying Zhou120 1 Introduction and review As populations rapidly age, public policies in many countries now encourage working into increasingly higher ages, so as to help manage the increased costs of government retirement pension and health care (Fisher et al. 2016, 2017). By legislation, about half of the OECD countries will increase their retirement age to approximately 66 years on average by 2060 (OECD 2017). Moreover, fi nancial incentives are now widely provided to delay retirement (Fisher et al. 2016). However, poor health – including cognitive health – is one of the leading causes of early and unplanned retirement (Fisher et al. 2016). Although age-related cognitive decline begins in early adulthood (Salthouse 2010b), a study of the United States found that between the ages of 55 and 69, approximately 10 percent of workers experienced steep cognitive decline (Belbase et al. 2015). These workers experiencing steep cognitive decline were more likely to have lower performance at work, to “downshift” to a less cognitively demanding job, or to retire earlier than workers not experiencing cognitive decline (Belbase et al. 2015; Fisher et al. 2017). The cognitive functioning among older adults at work, particularly during periods approaching or reaching retirement, is a critical factor to successfully aging at work, especially in cognitively-demanding jobs, such as professional, managerial, technical and other cognitively complex jobs (Fisher et al. 2017). Therefore, the development of cognitive function over time is a public health concern, particularly for the 55-69 age group. 1.1 Current research on trends in cognitive function Past research analysing the trends of cognitive function remains inconsistent regarding the theory (“success of success” hypothesis versus “failure of success” hypothesis, “Flynn-effect” versus “Anti-Flynn-effect”), as well as in study results and conclusions. On the one hand, the “success of success” hypothesis states that there exist factors which may improve the cognitive function of a population over time, postpone the onset of disability, and thereby result in more people living longer and healthier lives (Fries 1980). These positive factors for younger generations include higher education, better cardiovascular risk factor profi les (i.e. lower blood pressure and total cholesterol levels, less smoking, and more physical activity), better nutrition, and the improved treatment of cardiovascular diseases and diabetes (Pietschnig/Voracek 2015; Ahrenfeldt et al. 2018). Moreover, according to the “cognitive reserve” hypothesis, more recent generations have also more experience in jobs that require considerable cognitive investment, which leads to increases in neuronal development and consequently serves as a buffer against cognitive decline (Fisher et al. 2017). Recent studies have shown more recent birth years are associated with better cognitive function in Europe (Ahrenfeldt et al. 2018), and with slower rates of age-related cognitive decline and a lower risk of incident cognitive impairment in the United States (Downer et al. 2019; Vonk et al. 2019). The Berlin Aging Study reports that later-born cohort (in 2013-14, mean age of 75 years Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 121 old) showed higher levels of cognitive performance than their counterparts born 20 years earlier (Gerstorf et al. 2015). In addition, the tendency toward a cumulation of advantages or disadvantages, the so-called “Matthew effect”, is observed in human aging and life-course patterns within cohorts, as well as between cohorts (Dannefer 1987, 2003). If a larger proportion of younger generations is exposed to more factors benefi cial to cognitive function, taking the “Matthew effect” into consideration, their cognitive development over time at older ages may also develop positively. The increase of cognitive abilities from one generation to the next, the so called “Flynn effect”, has been reported in both developed countries (including the United States, 15 European nations, Japan, South Korea, Australia, Canada, and New Zealand) and developing countries (including China, India, Brazil, Dominica, Saudi Arabia, and Sudan) (Flynn 2012, 1987; Pietschnig/Voracek 2015). A meta-analysis has estimated a global increase of 3 IQ points per decade on average from 1909 to 2013 (Pietschnig/Voracek 2015). On the other hand, according to the “failure of success” hypothesis, reduced age-specifi c mortality and longer lifespans might increase the prevalence of chronic conditions including cognitive impairment, leading to worse overall cognitive health of the elderly population (Fries 1980). Studies examining trends in cognitive function showed an increase over time in prevalent cognitive impairment in the Netherlands (van den Kommer et al. 2018) and the United States (Downer et al. 2019), and in prevalence of dementia in Japan (Sekita et al. 2010) and Sweden (Mathillas et al. 2011). These studies investigating cognitive impairment and dementia usually focus on the population aged 65+, 75+, or 85+. However, little is known about the cognitive functioning and its development over time for the elderly approaching or reaching retirement (55-69 years old), despite research revealing that some workers experience steep cognitive decline over this period (Belbase et al. 2015). The plateauing or decline of cognitive abilities from one generation to the next, the so-called “negative Flynn effect” or “anti-Flynn effect”, has been reported in European countries including Norway (Sundet et al. 2004), Denmark (Teasdale/ Owen 2008), the Netherlands (Woodley/Meisenberg 2013), Finland (Dutton/Lynn 2013), France (Dutton/Lynn 2015), and Britain (Shayer/Ginsburg 2009). However, the majority of research on the negative Flynn effect used samples of under-20-year- olds (Dutton et al. 2016; Sundet et al. 2004; Teasdale/Owen 2008; Shayer/Ginsburg 2009; Woodley/Meisenberg 2013; Dutton/Lynn 2013, 2015). Taking dynamic time periods, geographic areas, and gender into consideration, the trends in the cognitive function studies are more complicated, since development over time in cognitive function may vary among different population subgroups of geographic regions, genders, and further potential factors (Prince et al. 2016; Dutton et al. 2016; Vonk et al. 2019). Studies of both Denmark and Norway found early gains of the Flynn effect, followed by stagnation through the 1990s in both countries, and a slight decline since the late 1990s in Denmark (Teasdale/Owen 2005; Sundet et al. 2004). A recent study indicated greater increases in cognitive performance for southern Europeans (compared to central and northern Europe) and for females (compared to males) (Weber et al. 2017). However, the authors noted that further research is required for the inconclusive and statistically insignifi cant • Ying Zhou122 fi ndings, namely that males had larger gains than females in southern Europe, whereas females had larger gains in northern and central Europe (Weber et al. 2017). Moreover, the study did not analyse nor adjust for educational levels explicitly. However, in addition to region and gender, education appears to be one of the key factors for time trends in cognitive ability in some countries, while not appearing infl uential in others (Neisser 1998; Teasdale/Owen 2005). More education in Europe may help cognitive levels, with a long-term effect through the pathways of occupation, social interaction, and lifestyle choices (Bennett et al. 2014; Banks/ Mazzonna 2012; Alley et al. 2007; Pavlova/Silbereisen 2012; Schneeweis et al. 2014). Well-educated people are more likely to achieve higher earnings and income, to engage in greater social and cultural participation, and to have more healthy habits and lifestyles, which may prevent the age-related cognitive decline (Banks/ Mazzonna 2012; Rowe/Kahn 1997). Moreover, people with higher education have more opportunities to work jobs demanding higher cognitive ability, and the “use it or lose it” hypothesis seems to be supported by results across studies (Fisher et al. 2017). Yet it remains unclear whether and how education moderates the trajectory of cognitive change (Lenehan et al. 2015; Alley et al. 2007). Some of the inconsistency in previous evidence may lie in the threshold of education for its effect (Lenehan et al. 2015). 1.2 Novel contributions of this study First, this study focuses on the age group of 55-69. As mentioned above, the cognitive ability and its trends for this age group has a considerable infl uence on sustained productive work in old age and desired retirement timing. Moreover, this age group is less affected by sample attrition related to death or other health conditions than older adults. However, most research examining time trends in cognitive limitations or dementia focuses on persons 70 years of age or older (Langa 2015; Choi et al. 2018); while the majority studies which found the anti-Flynn effect examined those younger than 20 years old (Larson et al. 2013; Alley et al. 2007; Dutton/Lynn 2013; Teasdale/Owen 2005). Second, this study investigates trends in cognitive function considering region, gender, and education by using a large representative European sample covering a total of ten countries in three European regions over nine years. The countries are Austria, Belgium, Czech Republic, Denmark, France, Germany, Italy, Spain, Sweden, and Switzerland. Third, this study applies a new approach to operationalising education, which elsewhere has been successfully applied to time trend investigation by examining the relative rank of educational attainment or socio-economic status in proportion to the overall distribution (Choi et al. 2018; Bound et al. 2015). In contrast, the conventional classifi cation of educational attainment by credentials or International Standard Classifi cation of Education (ISCED) may lead to less comparable subgroups in the study of time trends, because the high educational group (ISCED ≥ 5) has grown over time, and the low educational group (ISCED ≤ 2) has shrunk and is becoming an increasingly vulnerable segment of the population (Bound et al. 2015). Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 123 Insuffi cient numbers of participants with low or high education levels may limit statistical power and lead to bias (Lenehan et al. 2015). Furthermore, relying only on years of education may cloud international comparisons, as each country has its own system with varying programme durations and differing types of certifi cation, such as vocational and academic degrees. 1.3 Research question This study aims to investigate changes in cognitive function over time at ages 55-69 in Europe and the differences linked to region, gender, and education. This study seeks to provide evidence whether and how cognitive performance becomes better over time at the population level for the elderly close to retirement ages in Europe. These results may contribute to the maintenance of productive activities in old age. 2 Data and methods 2.1 Sample This study uses data of people between the ages of 55 and 69 in Waves 2, 4, 5, and 6 of the SHARE survey, a longitudinal, nationally representative, population- based, international, and harmonized survey (Börsch-Supan et al. 2013). Wave 1 was excluded because education was measured differently than in subsequent waves. Wave 3 was also excluded because it focused on respondents’ life histories. Wave 7 was excluded because it was designed to focus on both life history and panel information, and the panel information was collected by a shorter, condensed set of questions from the regular panel questionnaire. Thus, Verbal Fluency in Wave 7, one of the key variables in this study, has a great deal of missing values because it was only asked for respondents who participated in SHARELIFE in Wave 3. This study included only those countries which participated in all selected waves, except for the Netherlands due to differences in its mixed mode experiment in Wave 6 compared to the regular SHARE waves for other countries. Thus, ten countries were included in the analysis: Austria, Belgium, Czech Republic, Denmark, France, Germany, Italy, Spain, Sweden, and Switzerland. Although some waves covered two calendar years, the fi rst calendar year was used as the survey year, namely the years 2006, 2011, 2013, and 2015 for Waves 2, 4, 5, and 6, respectively. Italy and Spain were assigned to the category of southern Europe; Austria, Belgium, Czech Republic, France, Germany, and Switzerland were grouped together as central Europe; Denmark and Sweden were listed as northern Europe. 2.2 Cognitive function measures SHARE uses temporal orientation, numeracy, immediate and delayed recall testing, and verbal fl uency to measure cognitive function. This study focuses on recall tests and verbal fl uency for four reasons. First, episodic recall tests and semantic • Ying Zhou124 verbal fl uency are sensitive to cognitive aging (Souchay et al. 2000; Tomer/Levin 1993). Second, recall and verbal fl uency refl ect two key dimensions of cognitive performance separately, namely fl uid intelligence/working memory and crystallized intelligence/knowledge (Arpino/Bordone 2014; Salthouse 2006; Weber et al. 2017). Crystallized intelligence refers to an individual’s store of knowledge and learned operation; fl uid intelligence refers to the ability to think logically and solve novel problems independently of acquired knowledge, and is highly related to working memory (Nisbett et al. 2012; Arpino/Bordone 2014; Salthouse 2006; Horn/Cattell 1967; Jaeggi et al. 2008). Third, the temporal orientation in SHARE had a skewed distribution and thus was strongly affected by ceiling and fl oor effects. Fourth, the ordinal variable resulting from the numeracy test had limited variability, and the numeracy tests had a high number of missing values in Wave 6 (Adam et al. 2013; Mehrbrodt et al. 2019). The recall test was used to assess episodic memory performance. The respondent listened to a list of ten words, and immediately did a recall test (immediate recall test), followed by another after a delay (delayed recall test). The number of words the respondent recalled correctly was recorded, ranging from 0 to 10 for both tests. The Recall variable is the average score of the immediate and delayed recall tests. Verbal fl uency was used to assess executive function. Respondents were asked to list animals, and the number of animals the respondent said correctly in one minute was counted and stored as the raw variable of verbal fl uency. Outliers were defi ned as the values more than 2.5 standard deviations away from the mean per wave. These outliers were replaced by the wave-specifi c maximum values within 2.5 standard deviations. Thus, the verbal fl uency without outliers ranged from 0 to approximately 40. Additionally, the variable was multiplied by 0.25, so that it had a comparable scale of 0 to approximately 10, similar to the Recall variable. 2.3 Educational level For an international comparison of education, the ISCED 1997 was used with categories ranging from 0 to 6, with higher categories indicating higher levels of formal education. In order to improve comparability across groups over time, the ISCED 1997 grades were multiplied by years of education and further categorized into groups of “low”, “middle”, and “high” Adjusted Educational Levels for each survey wave. These three groups are made up of the bottom 25 percent (lowest to fi rst quartile Q1), middle 50 percent (fi rst quartile Q1 to third quartile Q3), and top 25 percent (third quartile Q3 to highest) for each survey wave. 2.4 Modelling strategies and statistical analysis Generalized Estimating Equation (GEE) models were estimated using a linear link function with normally distributed error terms and an unstructured covariance matrix to account for the dependency of observations between waves. The GEE model is also capable of handling missing values and unbalanced cases and is fairly robust with large sample sizes in particular. Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 125 Separate GEE models for Recall and Verbal Fluency were run to explore their association with the time trend measured by the numeric survey year, adjusting for gender, region, education, and the interactions of gender-by-survey year, education- by-survey year, and region-by-survey year. The adjustment was conducted further for age at interview and whether the subject had participated in SHARE for the fi rst time. In subsequent stratifi cation analyses the association between cognitive function and time was further investigated by GEE models stratifi ed by region, gender, and education. A sensitivity analysis was conducted for the other common correlation structures (autoregressive, exchangeable, and independent). The goodness of fi t statistic Quasilikelihood under the Independence Model Information Criterion (QIC) was used for the GEE model comparison (Hardin/Hilbe 2013). Moreover, a sensitivity analysis was conducted to include Wave 7. Finally, in order to investigate differences and similarities between the two operationalizations of education, the conventional classifi cation of education by ISCED (ISCED-97 = 0-2, 3-4, 5-6 as low, middle, and high levels of education) was additionally applied in a separate analysis. All analyses were performed in SAS 9.4 (SAS Institute, Cary NC). 3 Results The sample comprised 40,689 subjects with a mean age of approximately 62; roughly 54 percent of whom were female and 46 percent were male (see Table 1). From Wave 2 to Wave 6, the scores for Recall and Verbal Fluency increased on average. The proportions of low, middle, and high Adjusted Educational Level were approximately 25 percent, 50 percent, and 25 percent for each survey wave as defi ned, although these were not as precise as defi ned due to the tied values. The proportion of subjects who participated in the SHARE survey for the fi rst time decreased over time (see Table 1). Figure 1 illustrates the mean values and 95 percent confi dence intervals for Recall and Verbal Fluency, showing a consistent increase of Verbal Fluency over time for each European region, and of Recall over time in central Europe. A less consistent trend of increase of Recall can also be found in northern and southern Europe. Overall, central Europe shows higher cognitive function than southern Europe, while northern Europe has the highest values. However, the differences between regions are shrinking due to different speeds of increase over time. Table 2 shows the results of the two overall models for the association of cognitive function and time after controlling for age, gender, region, education, their interaction, and whether the subject had participated in SHARE for the fi rst time. Gender: Females had statistically signifi cantly better Recall than males (ß = 0.44, p < 0.0001). There was a general improvement in cognitive function over time (Recall: ß = 0.02, p < 0.0001; Verbal Fluency: ß = 0.04, p < 0.0001). However, there was no signifi cant overall difference in the trend between the two genders (Recall: ß = 0.00, p = 0.322; Verbal Fluency: ß = 0.01, p = 0.176). • Ying Zhou126 Region: There were statistically signifi cant regional differences, with northern Europe showing higher cognitive function than southern and central Europe (central vs. northern Europe: Recall: ß = -0.56, p < 0.0001; Verbal Fluency: ß = -0.57, p < 0.0001; southern vs. northern Europe: Recall: ß = -0.97, p < 0.0001; Verbal Fluency: ß = -1.73, p < 0.0001). However, southern Europe experienced faster improvements in Recall (ß = 0.03, p < 0.0001) and Verbal Fluency (ß = 0.02, p = 0.0009) than northern Europe; in central Europe Recall and Verbal Fluency also improved more quickly than in northern Europe (Recall: ß = 0.07, p < 0.0001; Verbal Fluency: ß = 0.02, p = 0.0006). Education: In comparison to those with relative low levels of education, people with middle and higher levels of education had higher cognitive function (Middle vs. Low educational levels: Recall: ß = 0.65, p < 0.0001; Verbal Fluency: ß = 0.61, p < 0.0001; High vs. Low educational level: Recall: ß = 1.18, p < 0.0001; Verbal Fluency: ß = 1.22, p < 0.0001). Nevertheless, those with lower levels of education Tab. 1: Demography, cognitive function, educational level, and other characteristics for the sample in SHARE Wave 2 Wave 4 Wave 5 Wave 6 n (%) n (%) n (%) n (%) N 12556 19338 24464 21151 Age 61.7 (4.27) 61.9 (4.22) 62.2 (4.21) 62.4 (4.25) Gender Female 6778 (54.0) 10562 (54.6) 13268 (54.2) 11666 (55.2) Male 5778 (46.0) 8776 (45.4) 11196 (45.8) 9485 (44.8) Country Austria 649 (5.2) 2549 (13.2) 2128 (8.7) 1578 (7.5) Belgium 1553 (12.4) 2597 (13.4) 2814 (11.5) 2883 (13.6) Czech Republic 1442 (11.5) 3067 (15.9) 3094 (12.6) 2515 (11.9) Denmark 1233 (9.8) 1122 (5.8) 2162 (8.8) 1932 (9.1) France 1361 (10.8) 2767 (14.3) 2322 (9.5) 1905 (9.0) Germany 1429 (11.4) 886 (4.6) 2731 (11.2) 2181 (10.3) Italy 1591 (12.7) 1789 (9.3) 2280 (9.3) 2489 (11.8) Spain 1048 (8.3) 1629 (8.4) 3018 (12.3) 2483 (11.7) Sweden 1536 (12.2) 1026 (5.3) 2308 (9.4) 1705 (8.1) Switzerland 714 (5.7) 1906 (9.9) 1607 (6.6) 1480 (7.0) Recall 4.6 (1.67) 5.0 (1.69) 5.1 (1.71) 5.2 (1.70) Verbal fl uency 5.1 (1.79) 5.3 (1.80) 5.5 (1.82) 5.6 (1.82) Adjusted Educational Level Low 3179 (26.0) 4001 (25.6) 5214 (26.9) 3889 (26.5) Middle 5895 (48.3) 7697 (49.2) 9267 (47.9) 7098 (48.3) High 3143 (25.7) 3946 (25.2) 4879 (25.2) 3707 (25.2) Participate in SHARE-survey First time 5786 (46.1) 13175 (68.1) 10664 (43.6) 3827 (18.1) Not the fi rst time 6770 (53.9) 6163 (31.9) 13800 (56.4) 17324 (81.9) Source: Own calculations, based on data from SHARE Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 127 Fig. 1: Bar chart for mean values of cognitive function with 95 percent confi dence interval by region and survey year 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 Northern Europe Central Europe Southern Europe Recall Recall by region and survey year 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 Northern Europe Central Europe Southern Europe 2006 2011 2013 2015 Verbal fluency Verbal fluency by region and survey year Survey year Source: Own calculations, based on data from SHARE • Ying Zhou128 improved their Recall faster than those with middle educational levels did (ß = -0.01, p = 0.005), and improved their Verbal Fluency more quickly than the highly-educated (ß = -0.02, p = 0.006). The models are stratifi ed by regions in Table 3 (Columns I and IV), permitting us to study the differences in regional time trends. The models reveal that cognitive function improved signifi cantly in southern and central Europe (southern Europe: Recall: ß = 0.08, p < 0.0001; Verbal fl uency: ß = 0.07, p < 0.0001; central Europe: Recall: ß = 0.09, p < 0.0001; Verbal fl uency: ß = 0.06, p < 0.0001), whereas in northern Europe Recall has stagnated (ß = -0.02, p = 0.145) and Verbal Fluency improved somewhat, but not statistically signifi cantly according to a signifi cance level of 0.05 (ß = 0.02, p = 0.066). Moreover, in southern Europe, the interaction of time-by-gender showed that males tended to benefi t more from improving cognitive function than females (Recall: ß = -0.02, p = 0.003; Verbal fl uency: ß = -0.01, p = 0.081). Additionally, in southern and central Europe, those with lower education levels tended to improve their cognitive function faster than the other educational groups, as most of the coeffi cients for the time-by-education interaction Tab. 2: Overall models: Association of the cognitive function and time after controlling for gender, region, education, their interactions, and potential confounders in SHARE by GEE models1 Recall Verbal Fluency Intercept 7.14*** 7.84*** Gender (Female)2 0.44*** 0.00 Survey year 0.02*** 0.04*** Survey year*Gender (Female)2 0.00 0.01 Region (Central Europe)3 -0.56*** -0.57*** Region (Southern Europe)3 -0.97*** -1.73*** Survey year * Region (Central Europe)3 0.07*** 0.02*** Survey year * Region (Southern Europe)3 0.03*** 0.02*** Adjusted educational level (Middle)4 0.65*** 0.61*** Adjusted educational level (High)4 1.18*** 1.22*** Survey year * Adjusted educational level (Middle)4 -0.01*** -0.01 Survey year * Adjusted educational level (High)4 0.00 -0.02*** Age -0.05*** -0.04*** Participate in SHARE-survey for the 1st time -0.18*** -0.13*** 1 *** p<=0.01; ** p<=0.05; *p<=0.1. 2 Male is the reference group for gender. 3 Regions: Southern Europe: Italy, Spain; Central Europe: Austria, Belgium, Czech Republic, France, Germany, Switzerland; Northern Europe: Denmark, Sweden (reference group). 4 Adjusted educational level: low: lowest-Q1 (reference group); Middle: Q1-Q3; High: Q3-highest. Source: Own calculations, based on data from SHARE Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 129 R e ca ll V e rb a l F lu e n cy O v e ra ll (I ) F e m a le ( II ) M a le ( II I) O v e ra ll (I V ) F e m a le ( V ) M a le ( V I) S o u th e rn E u ro p e 2 In te rc e p t 5 .6 7 ** * 6 .2 2 ** * 5 .3 2 ** * 5 .5 5 ** * 5 .6 6 ** * 5 .2 5 ** * M id d le E d u ca ti o n 3 1. 01 ** * 0 .9 5 ** * 1. 0 6 ** * 0 .5 8 ** * 0 .6 1 ** * 0 .5 5 ** * H ig h E d u ca ti o n 3 1. 6 9 ** * 1. 71 ** * 1. 6 9 ** * 1. 0 0 ** * 0 .9 4 ** * 1. 0 4 ** * S u rv e y y e a r 0 .0 8 ** * 0 .0 5 ** * 0 .0 9 ** * 0 .0 7 ** * 0 .0 5 ** * 0 .0 7 ** * S u rv e y y e a r * M id d le E d u ca ti o n 3 -0 .0 3 ** * -0 .0 2 * -0 .0 4 ** * -0 .0 1 0 .0 0 -0 .0 1 S u rv e y y e a r * H ig h E d u ca ti o n 3 -0 .0 5 ** * -0 .0 4 ** -0 .0 7 ** * 0 .0 0 0 .0 1 -0 .0 2 G e n d e r (F e m a le )4 0 .2 7 ** * -0 .1 3 ** S u rv e y y e a r * G e n d e r (F e m a le )4 -0 .0 2 ** * -0 .0 1 * C e n tr a l E u ro p e 2 In te rc e p t 6 .8 1 ** * 7. 31 ** * 6 .7 7 ** * 7. 3 0 ** * 7. 51 ** * 7. 10 ** * M id d le E d u ca ti o n 3 0 .5 4 ** * 0 .5 6 ** * 0 .5 0 ** * 0 .7 2 ** * 0 .7 1 ** * 0 .7 4 ** * H ig h E d u ca ti o n 3 1. 12 ** * 1. 12 ** * 1. 12 ** * 1. 41 ** * 1. 3 9 ** * 1. 4 3 ** * S u rv e y y e a r 0 .0 9 ** * 0 .0 8 ** * 0 .1 0 ** * 0 .0 6 ** * 0 .0 7 ** * 0 .0 7 ** * S u rv e y y e a r * M id d le E d u ca ti o n 3 -0 .0 1 ** -0 .0 1 -0 .0 2 ** -0 .0 1 ** -0 .0 1 -0 .0 2 ** S u rv e y y e a r * H ig h E d u ca ti o n 3 0 .0 0 0 .0 0 -0 .0 1 -0 .0 3 ** * -0 .0 2 * -0 .0 5 ** * G e n d e r (F e m a le )4 0 .4 6 ** * 0 .0 5 S u rv e y y e a r * G e n d e r (F e m a le )4 0 .0 0 0 .0 1 ** N o rt h e rn E u ro p e 2 In te rc e p t 7. 01 ** * 7. 8 0 ** * 6 .7 4 ** * 8 .4 2 ** * 8 .9 9 ** * 7. 7 6 ** * M id d le E d u ca ti o n 3 0 .4 3 ** * 0 .4 0 ** * 0 .4 6 ** * 0 .4 1 ** * 0 .4 6 ** * 0 .3 6 ** * H ig h E d u ca ti o n 3 0 .7 8 ** * 0 .8 6 ** * 0 .6 7 ** * 0 .9 1 ** * 0 .9 1 ** * 0 .9 1 ** * T a b . 3: S tr a ti fi ca ti o n a n a ly se s: A ss o ci a ti o n o f co g n it iv e f u n c ti o n a n d t im e in S H A R E b y G E E m o d e ls , st ra ti fi e d b y re g io n , an d f u rt h e r st ra ti fi e d b y g e n d e r1 • Ying Zhou130 R e ca ll V e rb a l F lu e n cy O v e ra ll (I ) F e m a le ( II ) M a le ( II I) O v e ra ll (I V ) F e m a le ( V ) M a le ( V I) S u rv e y y e a r -0 .0 2 0 .0 0 -0 .0 3 ** 0 .0 2 * 0 .0 3 * 0 .0 2 S u rv e y y e a r * M id d le E d u ca ti o n 3 0 .0 2 * 0 .0 1 0 .0 3 * 0 .0 1 0 .0 0 0 .0 1 S u rv e y y e a r * H ig h E d u ca ti o n 3 0 .0 4 ** * 0 .0 1 0 .0 6 ** * 0 .0 2 0 .0 2 0 .0 1 G e n d e r (F e m a le )4 0 .5 9 ** * 0 .0 1 S u rv e y y e a r * G e n d e r (F e m a le )4 0 .0 1 0 .0 1 T a b . 3: C o n ti n u a ti o n 1 * ** p < = 0 .0 1 ; ** p < = 0 .0 5 ; *p < = 0 .1 . M o d e ls a re f u rt h e r ad ju st e d b y a g e a n d w h e th e r th e s u b je ct p ar ti ci p at e d in t h e S H A R E s u rv e y fo r th e fi r st t im e . 2 R e g io n s: S o u th e rn E u ro p e : It al y, S p ai n ; C e n tr al E u ro p e : A u st ri a, B e lg iu m , C ze ch R e p u b lic , Fr an ce , G e rm an y, S w it ze rl an d ; N o rt h e rn E u ro p e : D e n m ar k, S w e d e n . 3 A d ju st e d e d u ca ti o n al le v e l: lo w : lo w e st -Q 1 ( re fe re n ce g ro u p ); M id d le : Q 1 -Q 3 ; H ig h : Q 3 -h ig h e st . 4 M al e is t h e r e fe re n ce g ro u p f o r g e n d e r. S o u rc e : O w n c al cu la ti o n s, b as e d o n d at a fr o m S H A R E Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 131 T a b . 4: S tr a ti fi ca ti o n a n a ly si s: A ss o ci a ti o n o f th e c o g n it iv e f u n c ti o n a n d t im e in S H A R E b y G E E m o d e ls , st ra ti fi e d b y re g io n a n d e d u ca ti o n 1 R e ca ll V e rb a l F lu e n cy O v e ra ll2 L o w M id d le H ig h O v e ra ll2 L o w M id d le H ig h E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 (I ) (I I) (I II ) (I V ) (V ) (V I) (V II ) (V II I) S o u th e rn E u ro p e 4 In te rc e p t 5 .6 7 ** * 6 .4 2 ** * 5 .8 7 ** * 6 .5 2 ** * 5 .5 5 ** * 5 .5 9 ** * 5 .6 0 ** * 8 .1 5 ** * G e n d e r (F e m a le )5 0 .2 7 ** * 0 .2 7 ** * 0 .2 3 ** 0 .3 5 * -0 .1 3 ** -0 .1 6 ** -0 .0 4 -0 .2 4 S u rv e y y e a r 0 .0 8 ** * 0 .0 8 ** * 0 .0 4 ** * 0 .0 2 0 .0 7 ** * 0 .0 7 ** * 0 .0 5 ** * 0 .0 5 ** * S u rv e y y e a r * G e n d e r5 -0 .0 2 ** * -0 .0 3 ** * -0 .0 1 -0 .0 1 -0 .0 1 * -0 .0 2 ** -0 .0 1 0 .0 0 M id d le E d u ca ti o n 3 1. 01 ** * 0 .5 8 ** * H ig h E d u ca ti o n 3 1. 6 9 ** * 1. 0 0 ** * S u rv e y y e a r * M id d le E d u ca ti o n 3 -0 .0 3 ** * -0 .0 1 S u rv e y y e a r * H ig h E d u ca ti o n 3 -0 .0 5 ** * 0 .0 0 C e n tr a l E u ro p e 4 In te rc e p t 6 .8 1 ** * 7. 5 7 ** * 7. 2 8 ** * 7. 7 5 ** * 7. 3 0 ** * 7. 4 4 ** * 8 .0 2 ** * 8 .7 1 ** * G e n d e r (F e m a le )5 0 .4 6 ** * 0 .4 0 ** * 0 .4 9 ** * 0 .4 2 ** * 0 .0 5 0 .0 6 0 .0 5 0 .0 5 S u rv e y y e a r 0 .0 9 ** * 0 .1 0 ** * 0 .0 7 ** * 0 .0 8 ** * 0 .0 6 ** * 0 .0 7 ** * 0 .0 5 ** * 0 .0 3 ** * S u rv e y y e a r * G e n d e r5 0 .0 0 -0 .0 1 0 .0 0 0 .0 1 0 .0 1 ** 0 .0 0 0 .0 1 * 0 .0 2 ** M id d le E d u ca ti o n 3 0 .5 4 ** * 0 .7 2 ** * H ig h E d u ca ti o n 3 1. 12 ** * 1. 41 ** * S u rv e y y e a r * M id d le E d u ca ti o n 3 -0 .0 1 ** -0 .0 1 ** S u rv e y y e a r * H ig h E d u ca ti o n 3 0 .0 0 -0 .0 3 ** * N o rt h e rn E u ro p e 4 In te rc e p t 7. 01 ** * 7. 16 ** * 7. 4 9 ** * 7. 61 ** * 8 .4 2 ** * 9 .6 0 ** * 8 .6 5 ** * 9 .0 7 ** * G e n d e r (F e m a le )5 0 .5 9 ** * 0 .5 4 ** * 0 .4 7 ** * 0 .7 3 ** * 0 .0 1 -0 .0 6 0 .0 7 0 .0 0 S u rv e y y e ar -0 .0 2 -0 .0 3 0 .0 0 0 .0 3 ** * 0 .0 2 * 0 .0 3 0 .0 4 ** * 0 .0 3 ** * S u rv e y y e a r * G e n d e r5 0 .0 1 0 .0 3 0 .0 2 -0 .0 1 0 .0 1 0 .0 1 -0 .0 1 0 .0 2 • Ying Zhou132 T a b . 4: C o n ti n u a ti o n R e ca ll V e rb a l F lu e n cy O v e ra ll2 L o w M id d le H ig h O v e ra ll2 L o w M id d le H ig h E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 E d u ca ti o n 3 (I ) (I I) (I II ) (I V ) (V ) (V I) (V II ) (V II I) M id d le E d u ca ti o n 3 0 .4 3 ** * 0 .4 1 ** * H ig h E d u ca ti o n 3 0 .7 8 ** * 0 .9 1 ** * S u rv e y y e a r * M id d le E d u ca ti o n 3 0 .0 2 * 0 .0 1 S u rv e y y e a r * H ig h E d u ca ti o n 3 0 .0 4 ** * 0 .0 2 1 * ** p < = 0 .0 1 ; ** p < = 0 .0 5 ; *p < = 0 .1 . M o d e ls a re f u rt h e r ad ju st e d b y a g e a n d w h e th e r th e s u b je ct p ar ti ci p at e d in t h e S H A R E s u rv e y fo r th e fi r st t im e . 2 T h e o v e ra ll m o d e l i s th e s am e a s th at in t h e T ab le 3 . It is li st e d h e re a g ai n f o r e as ie r co m p ar is o n w it h o th e r m o d e ls . 3 A d ju st e d e d u ca ti o n al le v e l: lo w : lo w e st -Q 1 ( re fe re n ce g ro u p ); M id d le : Q 1 -Q 3 ; H ig h : Q 3 -h ig h e st . 4 R e g io n : S o u th e rn E u ro p e : It al y, S p ai n ; C e n tr al E u ro p e : A u st ri a, B e lg iu m , C ze ch R e p u b lic , Fr an ce , G e rm an y, S w it ze rl an d ; N o rt h e rn E u ro p e : D e n m ar k, S w e d e n . 5 M al e is t h e r e fe re n ce g ro u p f o r g e n d e r. S o u rc e : O w n c al cu la ti o n s, b as e d o n d at a fr o m S H A R E Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 133 were statistically signifi cantly negative values, and none of these coeffi cients were statistically signifi cantly positive values. Further stratifi cation analysis by region and education, shown in Table 4, confi rmed a statistically signifi cant interaction of time and gender favoring males with lower education levels in southern Europe (Table 4). (Column II, interaction time-by-gender for Recall: ß = -0.03, p = 0.002; Column VI, for Verbal fl uency: ß = -0.02, p = 0.039). The sensitivity analysis regarding the correlation structures (autoregressive, exchangeable, and independent correlation structures) resulted in consistent parameter estimations and inference p-values, as well as similar goodness of fi t statistics (QICs). Results remained stable when the data of Wave 7 was included, despite a large number of missing values for Verbal Fluency. The sensitivity analysis using the conventional classifi cation of education by ISCED produced similar results, but was not able to show the interaction of survey year and educational level consistently (data not shown). 4 Discussion This study has three important results. First, cognitive function among 55-69-year- olds improved signifi cantly from 2006 to 2015 in southern and central Europe, whereas it did not in northern Europe, although there was already a generally higher level in northern Europe to begin with. Thus, the gap between European regions is getting smaller. Second, in southern and central Europe, the group with relative low education, starting from the lowest level of cognitive function, tended to improve its cognitive functions more quickly than the middle and higher educational groups did. Thus, the gap between the relative low educational group and other educational groups is growing smaller. Third, among those with lower education in southern Europe, males improved their cognitive functions more quickly than females. More generally, in southern Europe, males seemed to benefi t more than females in terms of improvement of cognitive function after controlling for education. The disparity of regions, education, and genders over time may be explained by the threshold theory, as Lyketsos et al. tried to interpret the benefi t of education for preventing cognitive decline in adulthood (i.e. beyond a certain “dose” of education, such as 8 years of schooling, additional years of education are no longer associated with less cognitive decline) (Lenehan et al. 2015; Lyketsos et al. 1999). Once the biological limits or environmental threshold on the improvement of cognitive function are reached or approaching, gains of cognitive ability might arrive at or get close to saturation and the improvements may yield diminishing returns, cease, or even reverse (Weede 2006; Rindermann 2008; Pietschnig/Voracek 2015; Dutton et al. 2016). In other words, the “success of success” or “failure of success” hypotheses alone may not explain the complete picture of time trends in cognitive function. • Ying Zhou134 4.1 Regional disparities The fi ndings of this study suggest that 55-69-year-old adults in southern and central Europe stay at the Flynn effect phase, whereas the same age group in northern Europe stays at the anti-Flynn effect phase. Flynn et al. have summarized that some Scandinavian nations had robust intelligence gains in the twentieth century but peaked around 1990 and may have gone into mild decline since then (Flynn 2012). If we look at the trends in cognitive function dynamically, the Flynn effect in southern and central Europe indicates that these countries may have the potential for further improvement on cognitive function from one generation to the next, whereas the anti-effect in northern Europe suggests that these countries may have arrived at or be approaching the maximum plateau. A recent study showed improvements in cognitive function for those aged 50+ in all European regions (Ahrenfeldt et al. 2018). However, more closely examining those aged 55-69 in that study, southern and central Europeans had considerable improvements of cognitive function, whereas northern Europeans didn’t (Ahrenfeldt et al. 2018). The mortality selection and different survival patterns for the oldest among the old (Hülür et al. 2013; Gerstorf et al. 2015) may explain the inconsistency between the present study and theirs. Furthermore, the fi nding of the present study is consistent with some previous studies on trends in cognitive function (Pietschnig/Voracek 2015; Sundet et al. 2004), but inconsistent with others (Dutton/Lynn 2015; Weber et al. 2017). The inconsistency may result from the age groups of the samples, the measures and analysis of cognitive function, and other potential infl uencing factors. Nonetheless, a major strength of the present study is that it provides evidence from three European regions with internationally comparable data and approaches. Cognitive health is a key prerequisite for deciding the age at which people retire, although the decision of when to retire is complex and often based on many factors including macroeconomic, family, work-related, and individual factors (Fisher et al. 2016, 2017). This study indicates that increases of the retirement age may be reasonable in southern and central European countries from the perspective of cognitive health. 4.2 Educational disparities A threshold for the cognitive health benefi ts of education may explain the association between education and the rates of cognitive change. According to a review paper published in 2000, the majority of the studies between 1985 and 1999 reported protective effects of education on the rate of cognitive decline (Anstey/Christensen 2000). In contrast, another review paper published in 2015 indicated that the majority of the studies from 2000 to 2015 reported no such protective effects (Lenehan et al. 2015). The benefi cial impact of education on cognitive decline may no longer be observable as education levels extend beyond a certain threshold, such as an eight-year threshold (Lyketsos et al. 1999; Lenehan et al. 2015). Having eight years or less of formal education was reported to be associated with greater cognitive Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 135 decline. However, having nine years of education or more was unrelated to any further reduction of cognitive decline (Lyketsos et al. 1999). In line with this threshold explanation, this study indicates that most benefi ts of the Flynn effect in those with the least education are probably due to the fact that those people with the lowest education levels had the lowest baseline values for cognitive function, and these remained lower than the threshold throughout the analysis period in southern and central Europe. Consequently, those with low education in these regions have disproportionally improved their cognitive function over time. On the other hand, this was not observed in northern Europe. A possible explanation is that the majority of the northern Europeans attained the maximum cognitive reserve capacity with increasing levels of education as the norm (Lenehan et al. 2015). Interestingly, previous studies have revealed that school reform in England led to improvements in old-age memory of people with less education (Banks/ Mazzonna 2012). Other studies also found that the early gains of the Flynn effect in Denmark and Norway were primarily at the low end of the cognitive test score distributions, which were attributed to the developments of educational system (e.g. improved pre-school care, starting school earlier and leaving later, and improved special education for physically or mentally disabled children) (Teasdale/Owen 2005; Sundet et al. 2004). Support for active aging was recommended to target less-educated and other disadvantaged groups in Germany (Pavlova/Silbereisen 2012). In a review paper, Lenehan et al. suggested that less-educated groups were an under-researched subgroup for the effect of education on the rate of cognitive function, and they assumed that this subgroup could benefi t most from education- based intervention approaches (Lenehan et al. 2015). Consistently, the fi ndings of this study provide further evidence that in the context of limited resources, it may be worth prioritizing those with relative low educational levels when it comes to education-based intervention approaches in southern and central Europe. This would also hold true for developing countries, since a continuing Flynn effect has been observed or has begun to be observed in the developing world (Flynn 2012; Nisbett et al. 2012). 4.3 Gender disparities An explanation for the greater improvement of cognitive function in males in southern Europe may be the larger biological vulnerability, variability, and fl exibility of cognitive function for males (Halpern 2012). Some types of mental retardation are X-linked and are therefore more prevalent in males (Halpern 2012). Moreover, previous studies have found that males show greater age-related decline than females, independent of whether males or females outperform for the parameter of cognitive function (Maylor et al. 2007). In other words, 55-69-year-old males may have lower baselines, and a larger potential to improve, particularly in those groups in southern Europe with the lowest educational levels. Banks et al. found that school reform in England improved executive function at old age only for males with less education (Banks/Mazzonna 2012). Schneeweis et • Ying Zhou136 al. similarly observed that the educational reforms in Europe had stronger protective effects on cognitive function and decline for males (Schneeweis et al. 2014). Both of these studies explained this as the positive fi nancial returns and prolonged labor force participation only for males from the educational reforms (Banks/Mazzonna 2012; Schneeweis et al. 2014). This may also explain the present study’s fi nding of greater increases in cognitive function favoring males in the group with lower education in southern Europe. Additionally , this study’s analysis of education gives further insights into the regional gender differences based on Weber’s papers (Weber et al. 2017). The larger gain for males in southern Europe which was observed by Weber et al. (Weber et al. 2017) may be attributed to the improvement in the lower educational group. 4.4 A more complete picture of disparity by region, gender, and education On the whole, threshold theory may be able to explain the various rates of cognitive improvement throughout all Europe by region, education, and gender. Southern and central Europeans, people with lower education, and males may have larger potentials to improve their cognitive function than the corresponding comparison groups, thus also showing quicker improvements in the data. Conversely, the cognitive function of the corresponding comparison groups may be closer to biological limits (saturation), hence improvements happen more slowly, if at all. In addition, threshold theory may also cover other factors, such as living conditions and heathy behaviors, which may infl uence cognitive improvements over time as well. Further research is needed to investigate and further support this explanation. 4.5 Strengths and limitations One of the major strengths of this study is that it investigated 55-69-year-olds, whose cognitive development over time is important for sustained productive work in old age and pension policies, but who have relatively rarely been investigated to date. Second, this study used SHARE datasets with large sample sizes covering numerous European countries with harmonized design and data (Börsch-Supan et al. 2013). Third, this study used a defi nition of education by quantiles, which decreased selection bias and increased comparability. Fourth, this study used the data of Waves 2, 4, 5 and 6, and considered the correlation of the same subject in its modelling strategy. This longitudinal approach can show a more precise picture than cross-sectional studies or studies with only two time points (Lenehan et al. 2015). Fifth, recall and verbal fl uency are performance-based measures, thus avoiding the biases which may occur in self-reported measures. This study also has limitations. First, the retest effect may infl uence the cognitive parameters (Salthouse 2010a). In order to minimize it, an adjustment for fi rst-time participation was included in the statistical models. Second, the two parameters only partially cover the overall concept of cognitive function. Further studies are needed to investigate other parameters of cognitive function. However, our Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 137 assessments of recall and verbal fl uency can be regarded as two key dimensions of intelligence, fl uid intelligence/working memory, and crystallized intelligence/ knowledge (Arpino/Bordone 2014; Salthouse 2006; Weber et al. 2017). Third, as an observational study without a causal identifi cation strategy, no causal relationship can be provided. Moreover, it is diffi cult to clearly distinguish the aging and cohort effects. Consideration must be taken when interpreting the results. Fourth, the overall effect sizes seem to be small. However, due to the division of values in our parameters for a comparable range, the true effect size might be larger than it seems to be. Moreover, the diffi culty of setting criteria of clinical meaningfulness for the effect size is a general limitation in research using cognitive tests (Halpern 2012). Fifth, the relative rank of educational attainment is used to minimize the selection bias of educational groups over time. However, it is diffi cult to use another golden standard measurement to validate this approach. A sensitivity analysis using conventional classifi cation of educational attainment was conducted to investigate the differences and similarities between the two operationalizations of education. 4.6 Conclusions Older adults close to retirement age (55-69-year-olds) in southern and central Europe improved their cognitive function signifi cantly from 2006 to 2015, whereas adults of the same age in northern Europe did not. Those with relative low levels of education in southern and central Europe improved their cognitive function most rapidly. Among those with lower education in southern Europe, males improved in cognitive function more quickly than females did. These results can be interpreted in line with threshold theory. Despite the complexity of the decisions about the retirement age, future increases of the retirement age may be reasonable in southern and central European countries from the perspective of cognitive function. It also may be worth prioritizing people with lower levels of formal education when it comes to education-based intervention approaches. Acknowledgements I would like to express my sincere thanks to Gabriele Doblhammer for her very valuable comments on this manuscript, and Renee Lueskow for her English editing service. This paper uses data from SHARE Waves 2, 4, 5, and 6, see Börsch-Supan et al. (2013) for methodological details. The SHARE data collection has been funded by the European Commission, DG RTD through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812), FP7 (SHARE-PREP: GA N°211909, SHARE-LEAP: GA N°227822, SHARE M4: GA N°261982, DASISH: GA N°283646) and Horizon 2020 (SHARE-DEV3: GA N°676536, SHARE-COHESION: GA N°870628, SERISS: GA N°654221, SSHOC: GA N°823782, SHARE-COVID19: GA N°101015924) and by DG Employment, Social Affairs & Inclusion through VS 2015/0195, VS 2016/0135, VS 2018/0285, VS 2019/0332, and VS 2020/0313. Additional funding from the German • Ying Zhou138 Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the U.S. National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C, RAG052527A) and from various national funding sources is gratefully acknowledged (see www.share-project.org). References Adam, Stéphane et al. 2013: Occupational activity and cognitive reserve. Implications in terms of prevention of cognitive aging and Alzheimer’s disease. In: Clinical Interventions in Aging 8: 377-390. https://doi.org/10.2147/CIA.S39921 Ahrenfeldt, Linda J. et al. 2018: Comparison of cognitive and physical functioning of Europeans in 2004-05 and 2013. In: International Journal of Epidemiology 47,5: 1518- 1528. https://doi.org/10.1093/ije/dyy094 Alley, Dawn; Suthers, Kristen; Crimmins, Eileen 2007: Education and cognitive decline in older Americans: Results from the AHEAD sample. In: Research on Aging 29,1: 73- 94. https://doi.org/10.1177/0164027506294245 Anstey, Kaarin; Christensen, Helen 2000: Education, activity, health, blood pressure and apolipoprotein E as predictors of cognitive change in old age: a review. In: Gerontology 46,3: 163-177. https://doi.org/10.1159/000022153 Arpino, Bruno; Bordone, Valeria 2014: Does grandparenting pay off? The effect of child care on grandparents' cognitive functioning. In: Journal of Marriage and Family 76,2: 337-351. https://doi.org/10.1111/jomf.12096 Banks, James; Mazzonna, Fabrizio 2012: The effect of education on old age cognitive abilities: evidence from a regression discontinuity design. In: The Economic Journal 122,560: 418-448. https://doi.org/10.1111/j.1468-0297.2012.02499.x Belbase, Anek et al. 2015: Slowed or Sidelined? The Effect of 'Normal' Cognitive Decline on Job Performance Among the Elderly. In: Center for Retirement Research at Boston College Working Paper 2015-12. Bennett, David A. et al. 2014: Cognitive and social lifestyle: links with neuropathology and cognition in late life. In: Acta Neuropathologica 127,1: 137-150. https://doi.org/10.1007/s00401-013-1226-2 Börsch-Supan, Axel et al. 2013: Data resource profi le: the Survey of Health, Ageing and Retirement in Europe (SHARE). In: International Journal of Epidemiology 42,4: 992- 1001. https://doi.org/10.1093/ije/dyt088 Bound, John et al. 2015: Measuring recent apparent declines in longevity: The role of increasing educational attainment. In: Health Affairs (Project Hope) 34,12: 2167-2173. https://doi.org/10.1377/hlthaff.2015.0481 Choi, HwaJung et al. 2018: Trends in the Prevalence and Disparity in Cognitive Limitations of Americans 55-69 Years Old. In: The Journals of Gerontology. Series B, Psychological sciences and social sciences 73,S1: S29-S37. https://doi.org/10.1093/geronb/gbx155 Dannefer, Dale 1987: Aging as intracohort differentiation: Accentuation, the Matthew effect, and the life course. In: Sociological Forum 2,2: 211-236. https://doi.org/10.1007/BF01124164 Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 139 Dannefer, Dale 2003: Cumulative advantage/disadvantage and the life course: Cross- fertilizing age and social science theory. In: The journals of Gerontology. Series B, Psychological sciences and social sciences 58,6: S327-S337. https://doi.org/10.1093/geronb/58.6.S327 Downer, Brian et al. 2019: Cohort Differences in Cognitive Impairment and Cognitive Decline Among Mexican-Americans Aged 75 Years or Older. In: American Journal of Epidemiology 188,1: 119-129. https://doi.org/10.1093/aje/kwy196 Dutton, Edward; Lynn, Richard 2013: A negative Flynn effect in Finland, 1997-2009. In: Intelligence 41,6: 817-820. https://doi.org/10.1016/j.intell.2013.05.008 Dutton, Edward; Lynn, Richard 2015: A negative Flynn Effect in France, 1999 to 2008-9. In: Intelligence 51: 67-70. https://doi.org/10.1016/j.intell.2015.05.005 Dutton, Edward; van der Linden, Dimitri; Lynn, Richard 2016: The negative Flynn effect: A systematic literature review. In: Intelligence 59: 163-169. https://doi.org/10.1016/j.intell.2016.10.002 Fisher, Gwenith G. et al. 2017: Cognitive functioning, aging, and work: A review and recommendations for research and practice. In: Journal of Occupational Health Psychology 22,3: 314-336. https://doi.org/10.1037/ocp0000086 Fisher, Gwenith G.; Chaffee, Dorey S.; Sonnega, Amanda 2016: Retirement timing: A review and recommendations for future research. In: Work, Aging and Retirement 2,2: 230-261. https://doi.org/10.1093/workar/waw001 Flynn, James R. 1987: Massive IQ gains in 14 nations: What IQ tests really measure. In: Psychological Bulletin 101,2: 171-191. https://doi.org/10.1037//0033-2909.101.2.171 Flynn, James R. 2012: Are we getting smarter? Rising IQ in the twenty-fi rst century. Cambridge/New York: Cambridge University Press. https://doi.org/10.1017/CBO9781139235679 Fries, James F. 1980: Aging, natural death, and the compression of morbidity. In: Bulletin of the World Health Organization 1980 80,3: 245-250. Gerstorf, Denis et al. 2015: Secular changes in late-life cognition and well-being: Towards a long bright future with a short brisk ending? In: Psychology and Aging 30,2: 301-310. https://doi.org/10.1037/pag0000016 Halpern, Diane F. 2012: Sex differences in cognitive abilities. 4th ed. New York: Psychology press. Hardin, James W.; Hilbe, Joseph M. 2013: Generalized estimating equations. 2nd ed. Boca Raton: CRC Press. Horn, John L.; Cattell, Raymond B. 1967: Age differences in fl uid and crystallized intelligence. In: Acta Psychologica 26: 107-129. https://doi.org/10.1016/0001-6918(67)90011-X Hülür, Gizem et al. 2013: Cohorts based on decade of death: no evidence for secular trends favoring later cohorts in cognitive aging and terminal decline in the AHEAD study. In: Psychology and Aging 28,1: 115-127. https://doi.org/10.1037/a0029965 Jaeggi, Susanne M. et al. 2008: Improving fl uid intelligence with training on working memory. In: Proceedings of the National Academy of Sciences of the United States of America 105,19: 6829-6833. https://doi.org/10.1073/pnas.0801268105 Langa, Kenneth M. 2015: Is the risk of Alzheimer's disease and dementia declining? In: Alzheimer's Research & Therapy 7,1: 34. https://doi.org/10.1186/s13195-015-0118-1 • Ying Zhou140 Larson, Eric B.; Yaffe, Kristine; Langa, Kenneth M. 2013: New insights into the dementia epidemic. In: New England Journal of Medicine 369,24: 2275-2277. https://doi.org/10.1056/NEJMp1311405 Lenehan, Megan Elizabeth et al. 2015: Relationship between education and age-related cognitive decline: a review of recent research. In: Psychogeriatrics 15,2: 154-162. https://doi.org/10.1111/psyg.12083 Lyketsos, Constantine G.; Chen, Li-Shiun; Anthony, James C. 1999: Cognitive decline in adulthood: an 11.5-year follow-up of the Baltimore Epidemiologic Catchment Area study. In: American Journal of Psychiatry 156,1: 58-65. https://doi.org/10.1176/ajp.156.1.58 Mathillas, Johan; Lövheim, Hugo; Gustafson, Yngve 2011: Increasing prevalence of dementia among very old people. In: Age and Ageing 40,2: 243-249. https://doi.org/10.1093/ageing/afq173 Maylor, Elizabeth A. et al. 2007: Gender and sexual orientation differences in cognition across adulthood: age is kinder to women than to men regardless of sexual orientation. In: Archives of Sexual Behavior 36,2: 235-249. https://doi.org/10.1007/s10508-006-9155-y Mehrbrodt, Tabea; Gruber, Stefan; Wagner, Melanie 2019: Scales and Multi-Item Indicators in the Survey of Health, Ageing and Retirement in Europe. SHARE Working Paper Series – Technical Reports. Nr. 45-2019 Neisser, Ulric 1998: The rising curve: Long-term gains in IQ and related measures. Washington, DC: American Psychological Association. https://doi.org/10.1037/10270-000 Nisbett, Richard E. et al. 2012: Intelligence: new fi ndings and theoretical developments. In: American Psychologist 67,2: 130-159. https://doi.org/10.1037/a0026699 OECD 2017: Pensions at a Glance 2017: OECD and G20 Indicators. Paris: OECD Publishing. https://doi.org/10.1787/pension_glance-2017-en Pavlova, Maria K.; Silbereisen, Rainer K. 2012: Perceived level and appraisal of the growing expectations for active ageing among the young-old in Germany. In: Research on Aging 34,1: 80-99. https://doi.org/10.1177/0164027511416371 Pietschnig, Jakob; Voracek, Martin 2015: One Century of Global IQ Gains: A Formal Meta-Analysis of the Flynn Effect (1909-2013). In: Perspectives on Psychological Science 10,3: 282-306. https://doi.org/10.1177/1745691615577701 Prince, Martin et al. 2016: Recent global trends in the prevalence and incidence of dementia, and survival with dementia. In: Alzheimer's Research & Therapy 8,23. https://doi.org/10.1186/s13195-016-0188-8 Rindermann, Heiner 2008: Relevance of education and intelligence for the political development of nations: Democracy, rule of law and political liberty. In: Intelligence 36,4: 306-322. https://doi.org/10.1016/j.intell.2007.09.003 Rowe, John W.; Kahn, Robert L. 1997: Successful aging. In: The Gerontologist 37,4: 433- 440. https://doi.org/10.1093/geront/37.4.433 Salthouse, Timothy A. 2006: Mental exercise and mental aging: Evaluating the validity of the “use it or lose it” hypothesis. In: Perspectives on Psychological Science 1,1: 68-87. https://doi.org/10.1111/j.1745-6916.2006.00005.x Salthouse, Timothy A. 2010a: Major issues in cognitive aging. Oxford/New York: Oxford University Press. Development over Time in Cognitive Function among European 55-69-Year-Olds ... • 141 Salthouse, Timothy A. 2010b: Selective review of cognitive aging. In: Journal of the International Neuropsychological Society 16,5: 754-760. https://doi.org/10.1017/S1355617710000706 Schneeweis, Nicole; Skirbekk, Vegard; Winter-Ebmer, Rudolf 2014: Does education improve cognitive performance four decades after school completion? In: Demography 51,2: 619-643. https://doi.org/10.1007/s13524-014-0281-1 Sekita, Atsuko et al. 2010: Trends in prevalence of Alzheimer's disease and vascular dementia in a Japanese community: the Hisayama Study. In: Acta Psychiatrica Scandinavica 122,4: 319-325. https://doi.org/10.1111/j.1600-0447.2010.01587.x Shayer, Michael; Ginsburg, Denise 2009: Thirty years on – a large anti-Flynn effect/ (II): 13- and 14-year-olds. Piagetian tests of formal operations norms 1976-2006/7. In: British Journal of Educational Psychology 79,3: 409-418. https://doi.org/10.1348/978185408X383123 Souchay, Céline; Isingrini, Michel; Espagnet, Laurence 2000: Aging, episodic memory feeling-of-knowing, and frontal functioning. In: Neuropsychology 14,2: 299-309. https://doi.org/10.1037/0894-4105.14.2.299 Sundet, Jon Martin; Barlaug, Dag G.; Torjussen, Tore M. 2004: The end of the Flynn effect? A study of secular trends in mean intelligence test scores of Norwegian conscripts during half a century. In: Intelligence 32,4: 349-362. https://doi.org/10.1016/j.intell.2004.06.004 Teasdale, Thomas W.; Owen, David R. 2005: A long-term rise and recent decline in intelligence test performance: The Flynn Effect in reverse. In: Personality and Individual Differences 39,4: 837-843. https://doi.org/10.1016/j.paid.2005.01.029 Teasdale, Thomas W.; Owen, David R. 2008: Secular declines in cognitive test scores: A reversal of the Flynn Effect. In: Intelligence 36,2: 121-126. https://doi.org/10.1016/j.intell.2007.01.007 Tomer, Rachel; Levin, Bonnie E. 1993: Differential effects of aging on two verbal fl uency tasks. In: Perceptual and Motor Skills 76,2: 465-466. https://doi.org/10.2466/pms.1993.76.2.465 van den Kommer, Tessa N. et al. 2018: Time Trend in Persistent Cognitive Decline. Results From the Longitudinal Aging Study Amsterdam. In: The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences 73,suppl_1: S57-S64. https://doi.org/10.1093/geronb/gbx151 Vonk, Jet M. J. et al. 2019: Secular trends in cognitive trajectories of diverse older adults. In: Alzheimer's & dementia: the journal of the Alzheimer's Association 15,12: 1576- 1587. https://doi.org/10.1016/j.jalz.2019.06.4944 Weber, Daniela et al. 2014: The changing face of cognitive gender differences in Europe. In: Proceedings of the National Academy of Sciences 111,32: 11673-11678. https://doi.org/10.1073/pnas.1319538111 Weber, Daniela; Dekhtyar, Serhiy; Herlitz, Agneta 2017: The Flynn effect in Europe – Effects of sex and region. In: Intelligence 60: 39-45. https://doi.org/10.1016/j.intell.2016.11.003 Weede, Erich 2006: Economic freedom and development: New calculations and interpretations. In: Cato Journal 26,3: 511-524. Woodley, Michael A.; Meisenberg, Gerhard 2013: In the Netherlands the anti-Flynn effect is a Jensen effect. In: Personality and Individual Differences 54,8: 871-876. https://doi.org/10.1016/j.paid.2012.12.022 • Ying Zhou142 Date of submission: 05.02.2020 Date of acceptance: 05.11.2021 Ying Zhou (). Institute for Sociology and Demography, University of Rostock. Rostock, Germany. E-mail: smile.zhou@hotmail.com Published by Federal Institute for Population Research (BiB) D-65180 Wiesbaden / Germany Managing Publisher Dr. Nikola Sander 2022 Editor Prof. Frans Willekens Managing Editor Dr. Ralina Panova Dr. Katrin Schiefer Editorial Assistant Beatriz Feiler-Fuchs Wiebke Hamann Layout Beatriz Feiler-Fuchs E-mail: cpos@bib.bund.de Scientifi c Advisory Board Karsten Hank (Cologne) Ridhi Kashyap (Oxford) Michaela Kreyenfeld (Berlin) Natalie Nitsche (Rostock) Zsolt Spéder (Budapest) Alyson van Raalte (Rostock) Rainer Wehrhahn (Kiel) Comparative Population Studies www.comparativepopulationstudies.de ISSN: 1869-8980 (Print) – 1869-8999 (Internet) Board of Reviewers Bruno Arpino (Barcelona) Kieron Barclay (Rostock) Laura Bernardi (Lausanne) Gabriele Doblhammer (Rostock) Anette Eva Fasang (Berlin) Michael Feldhaus (Oldenburg) Tomas Frejka (Sanibel) Alexia Fürnkranz-Prskawetz (Vienna) Birgit Glorius (Chemnitz) Fanny Janssen (Groningen) Frank Kalter (Mannheim) Stefanie Kley (Hamburg) Bernhard Köppen (Koblenz) Anne-Kristin Kuhnt (Duisburg) Hill Kulu (St Andrews) Nadja Milewski (Wiesbaden) Roland Rau (Rostock) Thorsten Schneider (Leipzig) Tomas Sobotka (Vienna) Jeroen J. A. Spijker (Barcelona) Heike Trappe (Rostock) Helga de Valk (The Hague) Sergi Vidal (Barcelona) Michael Wagner (Cologne)