Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, ECONOMICS AND FINANCE (economics.oxfordre.com). (c) Oxford University Press USA, 2016. All Rights Reserved. Personal use only; commercial use is strictly prohibited (for details see Privacy Policy and Legal Notice).

date: 17 August 2018

The Economics of Cognitive Aging

Summary and Keywords

Population aging, the combined effect of declining fertility and rising life expectancy, is one of the fundamental trends observed in developed counties and, increasingly, in developing countries as well. A key aspect of the aging process is the decline of cognitive ability. Cognitive aging is an important and complex phenomenon, and its risk factors and economic consequences are still not well understood. For instance, the relationship between cognitive aging and productivity matters for long-term economic growth. Cognitive functioning is also crucial for decision-making because it influences individuals’ ability to process information and to make the right choices, and older individuals are increasingly required to make complex financial, health, and long-term-care decisions that might affect their health, resources, and welfare. This article presents evidence from economics and other fields that have investigated this phenomenon from different perspectives.

A common empirical finding is the hump-shaped profile of cognitive performance over the life cycle. Another is the large variability of observed age profiles, not only at the individual level but also across sociodemographic groups and countries. The age profiles of cognitive performance also vary depending on the cognitive task considered, reflecting the different combinations of cognitive skills that they require. The literature usually distinguishes between two main types of cognitive skills: fluid intelligence and crystallized intelligence. The first consists of the basic mechanisms of processing new information, while the second reflects acquired knowledge. Unlike fluid intelligence, which declines rapidly as people get older, crystallized intelligence tends to be maintained at older ages. Differences in the age profiles of cognitive performance across tasks partly reflect differences in the importance of these two types of intelligence. For instance, tasks where learning, problem-solving, and processing speed are essential tend to be associated with a faster decline, while tasks where experience matters more tend to be associated with a slower decline. Various life events and behaviors over the life cycle also contribute to the large heterogeneity in the observed age profiles of cognitive performance. This source of variation includes not only early-life events and investments (e.g., formal education), but also midlife and later-life events (e.g., health shocks) and individual choices (e.g., health behaviors or retirement).

From an economic viewpoint, cognitive abilities may be regarded as one dimension of human capital, along with education, health, and noncognitive abilities. Economists have mainly focused their attention on human capital accumulation, and much less so on human capital deterioration. One explanation is that early-life investments appears to be more profitable than investments later in life. However, recent evidence from neuropsychology suggests that the human brain is malleable and open to enhancement even later in adulthood. Therefore, more economic research is needed to study how human capital depreciates over the life cycle and whether cognitive decline can be controlled.

Keywords: cognitive aging, decision-making, productivity, retirement

Basic Concepts

This section briefly defines the basic concepts of aging and cognitive ability. After discussing the multidimensional nature of cognitive ability, it describes the average profiles by age of the two main dimensions of cognitive capital—fluid and crystallized intelligence—and the large heterogeneity observed in individual age profiles.

Aging, Health, and Cognitive Ability

Aging may be defined as “the progressive loss of function with advancing age,” and “increasing rates of health problems including mortality are one of its main manifestations” (Kirkwood & Austad, 2000). The causes of aging are not clear. One view, known as the damage theory, regards aging as the result of an accumulation of damage, especially to deoxyribonucleic acid (DNA), which eventually causes death. Another view, the programmed aging theory, regards aging as the result of largely predetermined internal processes that are supposed to confer evolutionary benefits.

The aging process, only imperfectly expressed by chronological age, involves both physical and cognitive functions, with a strong interplay between physical health and cognitive changes (Alwin & Hofer, 2011; Bishop, Ehhum-Wilkens, Haas, & Kronenfeld, 2016). For example, a severe health shock (e.g., a stroke or the onset of important limitations on physical function) may lead to cognitive impairment, and severe cognitive impairment in turn might affect an individual’s health. Despite their strong interrelation, the age-related processes of health and cognitive decline do not coincide and might be affected by different biological and behavioral mechanisms. Indeed, it is not rare to observe both people in poor health, but with intact cognitive functioning, and people in relatively good health, but with poor cognitive functioning.

From an economic viewpoint, cognitive ability may be considered a type of capital that generates a flow of services in the form of our ability to perform a variety of tasks—from simple to very complex—through the process of mental and intellectual functioning. Consequently, in the remainder of this discussion, the terms cognitive ability and cognitive capital will be used interchangeably.

The age-related decline of cognitive capital is similar to that observed for other forms of human capital (e.g., muscular strength or physical health), but a unique feature of cognitive capital is that it does not appear to deteriorate due to wear. For example, health tends to deteriorate faster in manual jobs than in other jobs because of the amount of exertion required (Case & Deaton, 2005). On the contrary, the evidence from neuropsychology strongly suggests that our cognitive capital depreciates faster if it is not sufficiently stimulated (the “use it or lose it hypothesis,” as described by Hultsch, Hertzog, Small, & Dixon, 1999).

The literature usually distinguishes between the smooth process of normal age-related cognitive decline—i.e., the deterioration in cognitive performance that is part of the normal aging process—and neurological pathologies, such as Alzheimer’s disease and other forms of dementia, that lead to severe cognitive impairment (Leshner et al., 2017) and dramatically affect activities of daily living or social functioning. In between the expected cognitive decline of normal aging and the pathological decline of dementia is mild cognitive impairment. This intermediate stage, whose boundaries are not well defined, has received considerable attention in the medical literature because it is considered a risk state for dementia (Petersen, 2011).

Although it is not simple to draw clear distinctions among the three different stages, the focus of this article will be on the determinants and consequences of normal cognitive aging, especially from the economic viewpoint.

Multidimensionality of Cognitive Ability

Cognitive ability is better regarded as a multidimensional concept that comprises the wide range of skills required to perform the various mental activities most closely associated with learning and problem-solving.

The psychological literature usually distinguishes between two main types of skills (Horn & Cattell, 1967). The first, fluid intelligence, consists of the basic mechanisms of information processing, which are closely related to biological and physical factors and encompasses fundamental skills such as memory, abstract reasoning, executive function, and processing speed. Fluid intelligence is subject to a clear decline as people get older, and this decline may already start in the early 20s. Salthouse (1996) suggests that decline in processing speed is the most relevant of these issues.

The age-related decline in fluid intelligence tends to be partially offset by the different behavior of so-called crystallized intelligence, which consists of the knowledge and experience acquired during life through formal education, work history, and other life events. Crystallized intelligence tends to accumulate with age, but at a diminishing rate (Salthouse, 2012). The result is a hump-shaped age profile of the average level of cognitive functioning.

The importance of fluid and crystallized intelligence depends on the particular task considered, as different tasks are generally based on different combinations of the two types of intelligence. This implies that although individual cognitive functioning on average follows a hump-shaped pattern (see Figure 1), humans’ ability to perform a specific task may either decline or improve over time, depending on the task considered.

The Economics of Cognitive AgingClick to view larger

Figure 1. This graph shows the theoretical age profile of fluid and crystallized intelligence and the resulting average cognitive performance.

Heterogeneity of Cognitive Profiles

The aging process is clearly recognizable in the observed average age profiles of cognitive functioning, which are typically downward sloping after about age 50, regardless of gender, education, birth cohort, or country. Equally important, however, is the substantial heterogeneity of individual age profiles, even across individuals who are very similar in terms of observed characteristics (see, e.g., Yaffe et al., 2009). There is also evidence of large cross-country heterogeneity in average age profiles (see, e.g., Skirbekk, Loichinger, & Weber, 2012).

Both genetic differences (nature) and life-course events (nurture) contribute to this heterogeneity. Cunha and Heckman (2007) provide a useful framework that overcomes the traditional distinction between nature and nurture. They argue that cognitive and noncognitive abilities are created, not solely inherited, through the interaction between genes and environment. Particularly important seems to be the role of formal education (see, e.g., Banks & Mazzonna, 2012; Cook & Fletcher, 2015) and early childhood circumstances (see, e.g., Heckman, 2007; Case & Paxson, 2008; Currie, 2009), including exposure to war-related hardship (Havari & Peracchi, 2017). With regard to education, Cook and Fletcher (2015) suggest that it affects cognitive decline through changing brain processes (i.e., how we think) and building cognitive reserves, with more of an effect than socioeconomic status and resources. As for early childhood circumstances, their importance is not surprising because the first years of life is a period of intense physical, cognitive, and emotional development.

The importance of childhood circumstances does not mean that later-life events and behavior have no effect on cognitive trajectories. Health or psychological shocks (e.g., a car accident, a stroke, or the loss of a family member) may strongly affect cognitive trajectories at any age (e.g., van den Berg, Deeg, Lindeboom, & Portrait, 2010). Moreover, the levels of cognitive functioning are malleable and open to enhancement later in adulthood (Hertzog, Kramer, Wilson, & Lindenberger, 2008; Leshner et al., 2017). This explains why an association has been found between better cognitive trajectories and a variety of risk factors, such as occupational history (Mazzonna & Peracchi, 2012; Antonova, Bucher-Koenen, & Mazzonna, 2017), social relationships (Börsch-Supan & Schuth, 2013), and physical activity (Leshner et al., 2017).

Even in the case of dementia, where the genetic component might play a very important role, many behaviors during the lifetime, such as poor diet, smoking, physical inactivity, and social isolation, are now considered modifiable risk factors for which late intervention is possible (Livingston et al., 2017). Thus, cognitive aging is no longer regarded as a completely exogenous phenomenon. Instead, it is now recognized that events and behavior later in life may strongly affect individuals’ trajectories and partially explain the large heterogeneity observed in cognitive profiles.

Measurement and Data

To assess individuals’ cognitive ability and its different dimensions one needs reliable and objective tests. We begin this section with a brief overview of standard cognitive tests, with a focus on those used to measure the age-related process of cognitive decline. Then we discuss the available types of data, distinguishing between experimental and observational data. We end this section by describing some of the methodological issues that arise when using panel data to measure changes in cognitive performance over time.

Cognitive Tests

Cognitive testing has a long tradition in psychometrics, going back to the work of Francis Galton in the late 1800s. Many commonly used tests derive from standardized intelligence tests initially developed for particular needs, like the Binet-Simon scale developed in 1905 to identify primary-grade children with cognitive problems. Examples include the Armed Force Qualification Test (AFQT), which is used to determine basic qualifications for enlistment in the U.S. armed forces, and the Scholastic Aptitude Test (SAT), which is meant to assess student readiness for college in the United States. The AFQT is often employed as a proxy for unobserved ability when estimating wage equations (e.g., Griliches & Mason, 1972), while the SAT is often used in the economics of education to evaluate academic performance. These tests consist of multiple-choice questions that measure a combination of fluid and crystallized abilities, and the overall score is often regarded as a general measure of intelligence. Because these tests are typically taken at a young age, and only a few times, they are not adequate for studying age-related decline in cognitive abilities.

The recent progress in the economics of cognitive aging has been made possible by the availability of easy-to-use cognitive tests and by their adoption in experimental settings and sample surveys. Many cognitive tests used in old-age surveys are derived from the Mini-Mental State Exam, or MMSE (Folstein, Folstein, & McHugh, 1975), a multidimensional measure of cognitive functioning widely used as a screening instrument for cognitive impairment and dementia. The main advantages of the MMSE include its short administration period and ease of use, the fact that it requires no specialized equipment or training for administration, and its validity and reliability for the diagnosis and longitudinal assessment of Alzheimer’s disease.

Since the MMSE was originally proposed as a test for severe cognitive impairment, modifications have been introduced to test normal cognitive aging. For example, the Telephone Interview for Cognitive Status (TICS) has been used in many old-age surveys (as discussed later in this article) to assess cognitive function in healthy elderly people (de Jager, Budge, & Clarke, 2003). The TICS includes simple questions and problems in a number of areas, such as orientation in time or space, memory, verbal fluency, and numeracy. The test of orientation in time typically consists of questions about the date and day of the week. In general, this test shows very little variability across respondents. The test of memory typically consists of verbal registration and recall of a list of 10 words. The respondent hears the complete list only once, and the test is carried out two times, immediately after the encoding phase (immediate recall) and after a few minutes (delayed recall). The test of verbal fluency typically consists of counting how many distinct elements from a particular category the respondent can name in a specific time interval. The test of numeracy typically consists of a few questions involving simple arithmetical calculations, some of which may be based on real-life situations. Given the strong correlation between numeracy skills and financial literacy, additional numeracy tests are also included in many old-age surveys (e.g., basic questions about compound interest).

Clearly, the cognitive tests mentioned so far do not represent an exhaustive list of all cognitive tests used to measure the age-related process of cognitive decline. In particular, there are other types of tests, such as the Hodkinson abbreviated mental test score or the General Practitioner Assessment of Cognition, which are used as screening tools by practitioners for cognitive impairment and dementia.

Experimental Data

Experimental data usually come from randomized controlled trials (RCTs), in which a researcher randomly assigns sample individuals to either the treatment or the control group. RCTs are sometimes considered the gold standard for research, especially in the medical literature, although the credibility of their results can be undermined by excessive heterogeneity, especially when the distribution of the effects is asymmetric (Deaton & Cartwright, 2016).

Most available experimental evidence in the area of cognitive decline compares the effects of drugs aimed at slowing down the process by affecting the cardiovascular pathologies that lead to cognitive impairments and dementia, such as hypertension or diabetes. For instance, Livingston et al. (2017) report experimental evidence that antihypertensive drugs may reduce cognitive decline in people older than 60 years, while statins do not.

There is also substantial experimental evidence about the effect of cognitive-stimulating activities, such as training programs that target specific cognitive domains. The ACTIVE study (Rebok et al., 2014) is perhaps the largest and most convincing RCT implemented so far in this area, with a sample of 2,802 participants followed for 10 years. In this study, the treated group shows significant short- and long-term improvements in cognitive performance in several domains (e.g., memory, reasoning, and processing speed) compared to the control group. Based on these results, Livingston et al. (2017) conclude that “cognitive reserve is not a static property, but might be amenable to manipulation by cognitive interventions in later life.” On the other hand, the effects on other cognitive domains not targeted by the training are often small (Leshner et al., 2017). Moreover, attrition rates are particularly high for the 5- and 10-year follow-up periods—a finding that raises concerns about the internal validity of the results (Leshner et al., 2017).

Recent studies have used multimodal intervention, in which a combination of interventions—targeting physical activity, diet, social engagement, and cognitive training—are carried out simultaneously instead of one at a time. For example, the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) provided intensive multimodal intervention (300 hours) over a 2-year period to over 600 people who were older than 60 years and at high risk of dementia (Ngandu et al., 2015). The positive and significant effect on the treated group suggests a potential for lifestyle modifications to improve cognitive function, but the effects are rather small considering the intensity of the interventions (Livingston et al., 2017). Even less encouraging are the results from the so-called PreDiva study (van Charante et al., 2016), a 6-year multidomain intervention on over 3,500 participants that failed to show evidence of any effect on dementia incidence.

Unfortunately, the set of issues that one can study using RCTs is rather limited. In particular, it is very difficult to design and implement RCTs aimed at studying the effects on cognitive decline of risk factors that reflect behavioral choices. Although diet, physical inactivity, and social isolation have been identified in many longitudinal cohort studies as preventable risk factors for cognitive impairment and dementia, designing and implementing RCTs to investigate their effect (and the impact of other life events, such as unemployment or retirement) are extremely challenging and pose a variety of issues, including delicate ethical ones. Because these behavioral choices have long-run consequences, studies aimed at understanding their effects should start in midlife and would require repeated exposure and very long follow-up periods. Further, all one can hope to uncover are “intention to treat effects,” as researchers cannot force individuals to comply with a treatment. All these issues explain why RCTs on many of these behavioral risk factors have not been as successful as expected (Livingston et al., 2017).

This largely explains why most of the economic research on cognitive aging is based on observational data, especially from nationally representative surveys of the elderly population.

Observational Data

Until the early 1990s, empirical research on cognitive aging was mainly in neuropsychology and related fields and was mostly based on experimental evidence or small-scale surveys. The latter were not nationally representative but sometimes adopted very innovative longitudinal designs. Perhaps the most important among them is the Seattle Longitudinal Study, which started in 1956 and is considered one of the most extensive psychological studies of how cognitive skills develop and change through adulthood (Schaie, 1996).

A major impulse to the economics of cognitive aging came from the availability, beginning in the mid-1990s, of data collected through large-scale, nationally representative panel surveys of the aging population. In addition to rich longitudinal information about demographic, health, and socioeconomic aspects, these surveys included a battery of cognitive tests.

The first such data set was the U.S. Health and Retirement Study (HRS), a longitudinal project financed by the U.S. National Institute of Aging (NIA) and the U.S. Social Security Administration, which has followed a core cohort of adult Americans born between 1931 and 1941 and their spouses regardless of age, with biennial surveys since 1992. In addition, a group of elderly adults born before 1924 (the AHEAD cohort) has been followed since 1993, and additional cohorts have been added in 1998, 2004, and 2010 for respondents who were then 51–56 years old. The survey is still ongoing and currently covers about 20,000 Americans aged 50+.

Using a modified version of the TICS test, the HRS collects several cognitive measures selected to satisfy the following considerations:1 (a) to provide descriptive information about a comprehensive range of cognitive functions; (b) to span all difficulty levels, from competent cognitive functioning to cognitive impairment; (c) to be sensitive to change over time; (d) to be administrable in a survey environment with lay interviewers, over the telephone, in a short time; and (e) to be valid and reliable. The available cognitive measures include memory (immediate and delayed word recall), mental status (backward counting, date naming, object naming, and U.S. president/vice president naming), abstract reasoning, fluid reasoning, vocabulary, dementia, and numeracy.

The HRS has been the model for a number of surveys in many other countries. These include the English Longitudinal Survey of Ageing (ELSA), launched in 2002; the Survey of Health, Ageing, and Retirement in Europe (SHARE), launched in 2004; the Korean Longitudinal Study of Ageing (KLoSA), launched in 2006; the Japanese Study on Aging and Retirement (JSTAR), launched in 2007; the Chinese Health and Retirement Longitudinal Study (CHARLS), launched in 2008; the Irish Longitudinal Study on Ageing (TILDA), launched in 2009; and the Longitudinal Aging Study in India (LASI), launched in 2010.

The HRS and its sister surveys are part of a concerted, worldwide effort aimed at improving our knowledge of how different institutions, cultures, and policies can understand and prepare for population aging, and the similarity of the survey design contributes to scientific insights and policy development. These surveys also have been instrumental in experimenting new testing methodologies, including the collection of biomarkers, which can be analyzed to provide researchers with objective health information.

To illustrate the kind of empirical findings that these data provide, Figure 2 uses SHARE data to trace the average age profiles of test scores for three cognitive domains. We average by gender and over countries, ignoring existing differences across sociodemographic groups. Two findings are clearly revealed. First, average test scores remain quite stable until 60 or 65 years of age, but then decrease rapidly at older ages. Second, the age-related cognitive decline varies for the different cognitive domains. The recall test, which is meant to evaluate the respondents’ episodic memory, displays the largest degree of age-related decline. Consistently with the theoretical distinction between fluid and crystallized intelligence, episodic memory is considered a component of fluid intelligence. On the other hand, the verbal fluency and numeracy tests display a less steep age-related decline because they also rely on the respondent’s accumulated knowledge (e.g., vocabulary). These results are not specific to SHARE and are very similar to those obtained from the other surveys.

The Economics of Cognitive AgingClick to view larger

Figure 2. This graph shows the cross-sectional average age profile of three different cognitive test scores in SHARE (namely, Recall, Fluency, and Numeracy). The age profiles are constructed by pooling all observations from waves 4 and 5 of SHARE (2010–2013) and then collapsing the standardized test scores at their mean value by age.

Measuring Individual Changes Over Time

Panel data represents a powerful design for empirical analysis, for it allows a researcher to describe how an outcome of interest changes over time at the individual level, while controlling for the role of time-invariant, unobserved individual characteristics. In particular, the administration of cognitive tests in longitudinal surveys offers the possibility of analyzing the progression of cognitive impairment by following the same individuals as their cognitive functioning changes over time. Longitudinal surveys also offer the possibility of distinguishing time effects from age or cohort effects, which would be impossible with cross-sectional data. However, several issues complicate the use of the longitudinal information collected in panel surveys.

One issue is how to separate age effects from cohort effects (Deaton & Paxson, 1994). Another is how to interpret differences in mean cognitive trajectories by education, birth cohort, etc., in the presence of differential mortality (for two opposing views, see Agarwal, Driscoll, Gabaix, & Laibson, 2009; Rohwer, 2016). A third issue is attrition across waves of a survey. If attrition is correlated with cognitive functioning, the possibility of sample selection bias arises. A fourth issue, specific to cognitive functioning, is retesting effects. To avoid confounding measurement changes with substantive changes, ideally one would like to use identical tests, or at least the same test format, in each wave of a longitudinal survey. Unfortunately, repeated exposure to the same tests or the same test format may induce learning effects that cause spurious improvement in the cognitive scores of the respondents (see, e.g., Rabbitt, Diggle, Holland, & McInnes, 2004; Salthouse, Schroeder, & Ferrer, 2004; Mazzonna & Peracchi, 2012). Such learning effects also might be heterogeneous across cognitive domains and sociodemographic characteristics (e.g., age and education). This might misleadingly suggest that performance declines more rapidly on some task or for a particular sociodemographic group.

A proposed solution to account for retesting effects is to break the perfect correlation between age and testing occasions. This might be achieved using a refreshment sample to compare the performances of individuals tested for the second time with people of the same age interviewed for the first time. Although straightforward, this method is also not perfect if the two samples are not equivalent. This happens especially when the attrition rate across waves is high and particularly selective. Salthouse (2010) proposes an alternative solution that can take into account both the retesting and the selective attrition problems. The main idea is to administer the cognitive test of interest only to a randomly selected half of the participants on the first occasion, and to all participants on the second occasion. All in all, after adjusting for retesting effects and attrition, all the studies show that most of the longitudinal gains were reduced, with little discrepancy between the cross-sectional and the longitudinal age profiles.

Modeling

This section reviews the main models that have driven the research on cognitive aging. After briefly discussing several popular models from neuroscience, a version of the model proposed by Grossman (1972) is presented, which provides a useful way of formalizing the link between the natural aging process and cognitive-stimulating activities. This section concludes by discussing the main statistical models employed in the literature.

Models From Neuroscience

The neuropsychological literature offers several theories and hypotheses that try to explain the heterogeneity in the process of age-related cognitive decline described in the previous sections. For instance, the cognitive reserve hypothesis (Stern, 2002) postulates that individuals with similar observed characteristics are characterized by different levels of cognitive reserve (i.e., general knowledge and cognitive ability). This reserve allows them to limit the consequences of the process of neurodegeneration associated with aging and to cope with brain damage after a shock.2 In particular, once the brain reserve falls below a certain critical threshold, specific cognitive impairments emerge. In this model, the individual heterogeneity in cognitive reserve may be due to both genetic differences and personal choices such as education, occupation, or engaging in cognitive-stimulating activities.

Other models focus more on individual heterogeneity in the process of cognitive aging in order to understand to what extent one can enhance cognition in old age and slow cognitive decline. One popular theory is the use it or lose it hypothesis mentioned previously (Hultsch et al., 1999). A drawback of this theory, however, is that it does not help explain individual differences in the time and effort spent on these activities (Stine-Morrow, 2007). The cognitive enrichment hypothesis encompasses both theories and explicitly allows for individual differences in behavior. This theory states that individuals’ behavior has a meaningful impact on the level of their cognitive ability in old age (Hertzog et al., 2008). Based on the evidence of brain plasticity in old age (Leshner et al., 2017; Greenwood, 2007), individuals at any age can partially improve their cognitive ability via their behavior.

Economic Models

A useful way of formalizing the link between the natural aging process, cognitive ability or cognitive capital, and cognitive-stimulating activities or cognitive investment is the following version of the model originally proposed by Grossman (1972) for the evolution of physical health.3 This model offers two key insights. First, the observed age-related decline in cognitive ability need not be the same as biological deterioration because people may respond to aging by investing in cognitive-promoting behaviors (e.g., reading, social interactions, and cultural and other intellectually stimulating activities). Second, the amount of cognitive investment depends on market and nonmarket incentives, relative prices, discount rates, and other factors.

At the beginning of the relevant planning period (t=0), an individual chooses a sequence {(ct,it)}t=0T of consumption and cognitive investment to maximize her lifetime utility:

U=t=0Tut(ct,it)(1+ρt)t,

where T is life length and ρt is the rate of time preference. The period utility function may depend on t and is assumed to be strictly increasing in both arguments, with decreasing marginal utilities uct and uit. Although preferences also may depend on the level Kt of cognitive capital, this simple specification is enough for our purposes, as the main results do not change when cognitive capital gets included in the utility function.

The individual, who takes her initial levels K0 and A0 of cognitive capital and other assets as given, faces three constraints. The first is the law of motion of the level of cognitive capital:

Kt+1Kt=γtitδtKt,t=0,...,T1,
(1)

where γt is the efficiency of cognitive investment, which may depend on education, medical knowledge, and other factors; and δt is the natural deterioration rate of cognitive capital, subject to the terminal condition KT+1=0. Notice that δt is not constant over the life cycle. This is important because if δt were constant, one would have the counterintuitive result that the natural deterioration is greater when Kt is larger, which tends to occur when an individual is relatively young. The second is a nonnegative constraint on cognitive investment, which implies that

Kt+1Kt(1δt)Kt,t=0,...,T1.
(2)

It follows from Equation (1) that cognitive capital declines whenever the effective investment rate γtit/Kt falls below δt, but because of the constraint expressed in Equation (2), its rate of decline cannot exceed δt. The third is the lifetime budget constraint:

t=0Tct(1+r)t+t=0Tptit(1+r)t=A0+t=0Tyt(1+r)t,

where ρt is the price of cognitive investment, r is the real interest rate, and yt is income, which is allowed to be a nondecreasing function yt=Ft(Kt) of the level of cognitive capital.

The first-order conditions for an interior solution require the marginal rate of substitution (MRS) between cognitive investment and consumption to equal the effective price of cognitive investment in terms of foregone consumption:

uituct=MRSt=ptYtit,
(3)

where Yt is the discounted value at time t of all future incomes. For a fixed pt, since MRSt plausibly decreases with it, changes in cognitive investment and changes in Yt/it are positively correlated. As Yt declines with t, and since it is equal to zero at the end of life, cognitive investment is increasingly determined by the pure comparison of MRSt with pt. Thus, toward the end of life, cognitive investment should be small, especially if pt is high.

Notice that cognitive-damaging behaviors enter this model not through it but by increasing the natural deterioration rate δt. Further notice that differences in the utility function, such as a preference for more cognitive-stimulating activities by more educated individuals, may slow the rate of decline of the cognitive stock. Since education also may increase the efficiency of investment γt and lower the deterioration rate δt, more educated people may end up with a higher stock of cognitive capital throughout their lives.

To gain insight into the optimal solution to this model, consider the extreme case when cognitive investment does not enter the utility function (the so-called pure investment model), which implies that the condition in Equation (3) is satisfied by setting pt=Yt/it. If the ratio pt/γt takes the constant value p*, and if Ft has a strictly decreasing derivative ft, this is equivalent to the following condition:

πt=ft(Kt),
(4)

where πt=p*(r+δt) may be interpreted as the user cost of cognitive capital. Thus, when the nonnegativity constraint in Equation (2) is not binding, an increase in πt (due to an increase in p*, δt, or r) causes Kt to fall.

The Grossman model may be criticized for its exclusive attention to adult cognitive investment decisions, thus ignoring schooling decisions or the role of early childhood environment, as well as for its absence of dynamic complementarities in cognitive investment. Introducing some of these aspects is relatively straightforward, however.

An interesting alternative to the Grossman model is the deficit accumulation model recently proposed by Dalgaard and Strulik (2014). As in the Grossman model, individuals consume and make investments to slow the aging process. The key difference is how the aging process is understood. Drawing on reliability theory from engineering, Dalgaard and Strulik (2014) view the human organism as consisting of a given number of individual parts (called blocks) that are connected in parallel. Blocks do not age, but they have a fixed probability of failure. The system as a whole is assumed to survive so long as there is one functioning block remaining. This set of assumptions implies the following law of motion for the stock of health or cognitive capital:

Kt+1Kt=γtitμ(K¯Kt),t=0,...,T1,
(5)

where μ is an age-invariant physiological parameter driving the aging process and K¯ is the upper limit to an individual’s cognitive ability. Notice that Equation (5) implies that natural deterioration increases with K¯Kt and therefore is higher at later ages. This in turn means that attempts to contrast natural deterioration cost much more, in terms of foregone consumption, at later ages than at younger ages. If the efficiency of cognitive investment is lower at later ages, they also may be much less effective. Under the pure investment model and the additional assumption that pt/γt takes the constant value p*, again Equation (4) is obtained, except that now πt=(rμ)p*, which requires r>μ to make economic sense.

Statistical Models

The economic literature on cognitive aging tends to use a different range of statistical models than other disciplines (e.g., epidemiology, sociology, and psychology) also interested in the subject. This variation reflects differences in the focus of the analysis, the kind of data that are typically used, and the weight of the psychometric tradition.

At the cost of some simplification, the economic literature tends to focus on the role played by income, educational attainment, and personal decisions such as retirement and public policies, as well as on the effect that cognitive aging may have on other outcomes of interest, such as saving and portfolio choices. Epidemiology and sociology tend instead to concentrate their attention on the relative importance of genetic and sociodemographic factors of cognitive aging, while psychology tends to concentrate on the effect of training and the prediction of dementia.

Although all disciplines recognize the importance of answering questions about cause and effect, the economic literature appears to be especially concerned with the question of what can be learned from observational (i.e., nonexperimental) data, perhaps because of its extensive use of survey data and its increasing use of administrative data.

The economic literature also tends to focus more on the direct modeling of a small number of selected cognitive measures, often treated as continuous measurements, while the other disciplines tend to follow the psychometric tradition more by modeling cognitive abilities as an unobservable or latent variable (more rarely, as a set of latent variables) of which the various cognitive tests are only imperfect measures (see, e.g., Salthouse, 2001).

As a result, the economic literature tends to rely much more than other disciplines on linear models, which are estimated using a variety of methods—particularly ordinary least squares, fixed effects, difference-in-differences, regression discontinuity, or instrumental variables—that depend on the type of data (e.g., cross-sectional or panel) and the additional information at hand. Interestingly, despite the availability of panel data and the fact that cognitive aging is a dynamic process, the use of dynamic models is not very common in the empirical literature, perhaps because of their technical difficulty and the need to address delicate selectivity issues explicitly.

Cognitive Aging and Decision-Making

The trend observed in many developed economies toward the declining importance of publicly provided safety nets, especially social security and healthcare, places an increasing emphasis on individual decision-making skills, as older adults are asked more and more to make complex decisions about medical treatments, retirement plans, and insurance coverage, which will crucially affect their lifetime resources and health.

A long-standing body of literature, mainly in psychology, studies the extent to which aging and cognitive decline affect the decision-making process (for a review, see Carpenter & Yoon, 2011). This research shows that older adults are more likely to use heuristic or biased strategies in their decision-making, which has been linked to cognitive decline. Given their cognitive constraint, older adults in fact may choose to limit both the quantity and the complexity of the information that they use. Hess (2014) argues that the age-related decline in processing speed leads to an increasing cost of engagement in effortful cognitive activities. In particular, because of the aging process, more effort would be needed to achieve a particular level of performance.

This may explain why older adults prefer a simplified choice process, with limited information search and clear answers. In certain situations, such as the case of frequent consumer choices in which experience plays a greater role, the strategy of reducing the number of options and focusing on the essentials often leads older adults to better decisions than younger adults (Carpenter & Yoon, 2011). On the other hand, when the choice process requires considering several relevant options and searching for new information, cognitive decline may lead older adults to suboptimal decisions, as in the case of complex financial, medical, and health insurance decisions.

For example, Fang, Keane, and Silverman (2008) show that cognitive ability emerges as one of the strongest predictors of purchase of Medicare supplemental insurance (Medigap). Their results provide support for the idea that many people in the United States face difficulties in understanding Medicare and Medigap rules. Using a unique data set of prescription drug claims, Abaluck and Gruber (2011) evaluate elderly choices under Medicare Part D, a program created to subsidize the cost of prescription drugs. Consistent with the insights from the psychological literature, they find that the elderly tend to focus on a narrow range of dimension-making choices, which is inconsistent with optimization using full information. In particular, the elderly tend to place much more weight on plan premiums and financial characteristics and less weight on expected out-of-pocket costs and financial risk.

Cognitive Aging and Risk Aversion

Because of the fundamental role that preferences play in economic models, economists have long been interested in the relationship between individual preferences and cognitive abilities. Preferences toward risk and impatience have been found to be strongly correlated with cognitive abilities (e.g., Dohmen, Falk, Huffman, & Sunde, 2010; Benjamin, Brown, & Shapiro, 2013).4 Not surprisingly, Bonsang and Dohmen (2015) find that the association between aging and risk aversion is mediated by numerical ability. The increasing cost of cognitive engagement may explain the positive relationship between aging and risk aversion. As pointed out by Hess (2014), if risks are perceived as high but cognitive resources are shrinking, older adults may be less motivated to take risks.

A few studies investigate the relationship between risk aversion, aging, and cognitive skills using experimental methodologies to elicit and assess individual risk attitudes. Such methodologies include various lottery experiments (Dohmen et al., 2010; Andersson, Holm, Tyran, & Wengström, 2016) and methods conducted in a lab environment. For example, in the Balloon Analogue Risk Task (BART), participants are asked to pump a series of virtual balloons and can decide when to stop pumping. An individual earns points with each pump but risks losing the earned points if the balloon explodes. Using this experiment, Koscielniak, Rydzewska, and Sedek (2016) find that aging is positively associated with risk aversion and worsens performances (i.e., the points gained). They also show that this age effect is mediated by processing speed and the motivational mechanisms mentioned previously.

Savings, Wealth, and Portfolio Choices

In the last decade, partly motivated by the financial crisis of 2007–2008, a rapidly growing body of research in economics has investigated the causes and consequences of financial literacy. Financial ability is becoming more and more important because households have access to increasingly complex financial products and must make important financial decisions about their wealth. This is especially true for the older populations that, in most developed countries, hold a substantial share of a nation’s total financial wealth.5 There is strong evidence that a large fraction of households are unfamiliar with the most basic economic and financial concepts, like the compounding of interest rates (for a review of the literature, see Lusardi, Mitchell, & Curto, 2014). In addition, Lusardi and Mitchell (2007) and Behrman, Mitchell, Soo, and Brava (2012) show that individuals with greater financial literacy tend to accumulate larger retirement wealth, while Banks, O’Dea, and Oldfield (2010) report evidence of the different age-wealth profiles of more and less financially educated people.

Although often used interchangeably, financial literacy and cognitive ability are distinct concepts. Financial literacy may be considered the output of a human capital production function that takes as input various types of cognitive ability, especially numeracy (Banks, 2010). Christelis, Jappelli, and Padula (2010) find a strong positive relationship between cognitive abilities, especially numeracy, and the propensity to invest in risky assets (stock, mutual funds, etc.). More recently, Agarwal and Mazumder (2013) show that performance on cognitive tests strongly predicts the quality of financial decisions related to the use of credit cards, while Brown, Kapteyn, Luttmer, and Mitchell (2017) show that cognitive ability strongly predicts an individual’s ability to value annuities. To disentangle the effect of cognitive skills from that of other unobserved factors (e.g., preferences, constraints, information, or beliefs), Choi, Kariv, Müller, and Silverman (2014) conduct a large-scale experiment and find that people with better decision-making ability tend to accumulate more wealth.

Another related strand of the literature has investigated the age profile of financial literacy or financial performance to figure out whether the age-related process of cognitive decline affects individuals’ ability to make good financial decisions. Agarwal et al. (2009) show that the quality of credit decisions (specifically, financial mistakes in loans and mortgages) declines after peaking between 50 and 55 years of age. Korniotis and Kumar (2011) try to disentangle the effects of age and experience. They show that older and more experienced investors exhibit greater investment knowledge, but aging negatively affects their decision skills, and also that investment performance peaks around 70 years of age and declines sharply afterward. More recently, Finke, Howe, and Huston (2016) directly estimate the relationship between age and financial literacy, reporting evidence of a linear decline in financial literacy after the age of 60.

Overall, this research finds clear evidence of a hump-shaped profile of financial literacy, which is consistent with the age profile of cognitive skills discussed in this article. On the other hand, these studies come to different conclusions regarding the estimated peak age and the speed of the subsequent decline. This might be because they consider different measures of financial literacy or performance, each of which corresponds to different combinations of fluid and crystallized intelligence.

From a policy perspective, it is particularly important to understand whether people recognize their own cognitive decline and how they react to it. If people perceive or predict their own cognitive decline, they can delegate financial decisions to someone they trust—another family member or a financial advisor—without incurring any financial loss. In this case, policy interventions might not be needed (Agarwal et al., 2009). The case is different, though, if people do not realize that their cognitive performance are declining. In this case, older individuals not only are more likely to incur financial losses (because of bad investment decisions), but they also might be vulnerable to financial frauds and swindles. The empirical evidence is still scarce, but the few papers investigating the issue provide some support for this second hypothesis. In particular, Finke et al. (2016) show that older individuals seem to be unaware of the age decline in their cognitive and decision-making abilities, and this seems to contribute to the observed decline in financial literacy and performance. Moreover, Lusardi et al. (2014) argue that financial swindles are more frequent among the elderly because the mismatch between actual and perceived levels of financial literacy increases with age.

Cognitive Aging, Productivity, and Retirement

The relationship between cognitive aging, productivity, and retirement is particularly important due to both the demographic transition and the rapid pace of technological change. Population aging implies an aging workforce, and the relationship between cognitive aging and productivity determines how demographic transitions affect economic growth. If individual productivity strictly follows the hump-shaped profile of cognitive ability, then an aging workforce may reduce economic growth and decrease fiscal sustainability. At the same time, cognitive skills are becoming increasingly relevant to labor market performance because of technological change (Spitz-Oener, 2006). This leads to the important distinction between biological skill obsolescence—the result of the age-related decline in cognitive ability—and economic skill obsolescence—the result of technological innovation or changes in the pattern of trade. Even if workers are unaffected by cognitive aging, the second type of skill obsolescence may affect the value of their human capital.

Cognitive Aging and Productivity

From the viewpoint of a company, older workers may represent a loss if their wages exceed their productivity levels (Lazear & Moore, 1984). This argument often has been used as a motivation or justification for early retirement policies. However, given the different age profiles of fluid and crystallized intelligence, the effect of aging on individual productivity, either theoretically or empirically, is less clear. While younger workers may acquire new skills faster, older workers may buffer their decline in fluid intelligence by relying on their crystallized intelligence (i.e., the knowledge and experience accumulated in the workplace). This implies that the effect of cognitive aging may be heterogeneous across jobs, depending on the amount of fluid intelligence required. For instance, cognitive aging may not severely affect the productivity of workers who perform routine tasks (e.g., those in clerical jobs), for which there is no need to acquire new skills. Workers’ productivity also may be affected by other individual characteristics, such as noncognitive skills (e.g., the ability to work in teams and to deal with human nature), which may follow an age profile that differs from that observed for cognitive ability.

Which workers are more vulnerable to cognitive aging partly depends on the type of work that they do. For example, Belbase and Sanzenbacher (2016) argue that there are two types of workers who may struggle to maintain their productivity. The first are those in high-skilled occupations with intense demand for fluid intelligence (e.g., air traffic controllers). The second are those who experience unusually severe cognitive decline. In particular, a sharp decline in fluid intelligence, perhaps following a health shock, severely limits the ability of a worker to acquire the new skills required for a new job.

Following the insight of the human capital model, the trade-off between the benefits and costs of cognitive investment at older ages should be considered as well. The model presented here predicts that the incentives to cognitive investment crucially depend on the working time horizon, which implies that there may be little incentive for workers close to retirement to acquire new skills (Bartel & Sicherman, 1993; Montizaan, Cörvers, & De Grip, 2010). The limited working horizon also affects a company’s willingness to invest in the human capital of older workers because of the short time window to recoup the investment that would have to be made. Similar considerations apply to older workers who end up switching firms or occupations.

Assessing empirically the extent that cognitive aging affects workers’ productivity is not simple. First, measuring productivity directly is complicated. Theoretically, in a competitive labor market, the wage rate should be equal to the value of a worker’s marginal productivity (VMP). Such a relationship may hold on average, but the age-wage profile of an individual worker can negatively affect this, as wages often increase with seniority at a job but rarely decrease at older ages, even if an older worker is less productive. One reason is that, to solve an agency problem, workers are often paid less than their VMP when young, and more when old (Lazear & Moore, 1984). Moreover, other institutional features, such as minimum wages and collective agreements, may prevent the decline of wages at older ages.

For this reason, several studies use matched employer-employee microdata and focus on plant- or company-level outcomes, such as value added or sales per worker (see, e.g., Haltiwanger, Lane, & Spletzer, 1999; Hellerstein, Neumark, & Troske, 1999). Most of these studies confirm the presence of a hump-shaped age profile of workers’ performance, but the age where performance is estimate to peak varies substantially across studies, ranging between 35 and 55 years of age. Unfortunately, measures of workers’ productivity at the plant- or company-level may be severely biased. Indeed, the observed age composition of the workforce may be endogenous (Börsch-Supan & Weiss, 2016). One reason is that firms are likely to retain workers who are more productive and fire those who are less so. A second reason is that plant closures and early retirement lead to additional selection. Finally, better firms tend to expand their workforce by new hires, which typically results in a younger workforce.

To overcome some of these challenges, Börsch-Supan and Weiss (2016) measure productivity at the work-team level by using longitudinal data from a German truck assembly plant.6 One of their interesting findings is that the number of errors that the workers make increases slightly with age, but the severity of the errors steadily declines. This suggests that worker experience is an important determinant of productivity, and a declining age profile of cognitive ability in a worker does not always imply a declining age profile of the person’s performance. Again, such evidence confirms that cognitive aging may have heterogeneous effects across jobs, depending on the amount of fluid intelligence required by each job. Indeed, when the focus is on creative jobs or top performance at the individual level, such as that of painters (Galenson & Weinberg, 2000) or Nobel prizewinners (Jones, 2010), the evidence points toward an earlier decline in productivity.

Cognitive Aging and Retirement

During the 1980s and 1990s, many developed economies experienced a rapid decline in labor force participation rates among older adults, especially males. To a large extent, this phenomenon was due to generous early retirement policies, often justified by the argument of declining productivity at older ages (see, e.g., the country-specific data in Gruber & Wise, 1999, 2004). However, declining fertility and increasing life expectancy have led most developed economies to rethink their retirement policies by removing early retirement incentives and raising the average retirement age, in order to guarantee the financial stability of social security programs.

The changes in retirement policies and the lengthening of the retirement period due to the rapid increase in life expectancy have led many scholars to analyze the potential spillover effects of retirement on individuals’ health and cognitive abilities. There are dissenting opinions about these effects because the empirical evidence is mixed, especially in the case of health (see, e.g., Mazzonna & Peracchi, 2017). From the economic literature, three main mechanisms can be identified through which retirement may positively affect health: the relief from work-related stress and strain (see, e.g., Case & Deaton, 2005; Mazzonna & Peracchi, 2017), the increase in sleep duration (Eibich, 2015), and the increase in physical activity (Eibich, 2015; Kämpfen & Maurer, 2016). Some of these mechanisms also may be relevant for cognition, but other aspects, such as the lack of work-related incentives to preserve cognitive capital after retirement and the reduction in cognitive stimulation (use it or lose it), are likely to play an important role and lead to negative effects of retirement on cognitive ability. A negative effect on health and cognitive abilities also may be expected if retirement reduces social interactions (Börsch-Supan & Schuth, 2013).

Theoretically, the effect of retirement on cognitive aging can be modeled by making two changes to the model, as presented in this article. The first is to consider a worker who knows for sure that she will retire at age R. The second is to assume that her income before retirement is entirely from labor earnings, which in turn are an increasing function yt=Ft(Kt) of the current level of cognitive capital Kt, while her income after retirement is entirely from a pension and no longer depends on Kt.

To understand the age profile of cognitive capital in this case, consider the extreme case of the pure investment model. If income depends on the current level of cognitive capital only up to retirement, then in a model without the nonnegativity constraint given in Equation (2), the optimal level of cognitive capital should drop to zero immediately after retirement. Since Equation (2) does not allow this, cognitive capital can decline only at its maximal rate δt. Thus, in the pure investment model with strictly decreasing marginal productivity of cognitive capital, ft=yt/Kt, we find

Kt={ft1(πt),tR,(1δt1)Kt1,t>R.
(6)

Another implication of Equation (6) is that the cognitive gap between initially identical individuals who differ only in their retirement pattern widens rapidly with age. Thus, simply introducing a retirement dummy, as often done in the literature (see, e.g., Rohwedder & Willis, 2010), is an inadequate way of modeling the effect of retirement on cognitive ability.

The simplicity of Equation (6) is lost with a model that makes the more plausible assumption that cognitive investment (or cognitive stock) enters the utility function, but the result remains that the rate of decline of the cognitive stock will depend on the efficiency of cognitive investment through the user cost of cognitive capital, πt, and it will increase after retirement. However, this rate of decline will be less than the pure deterioration rate δt if there are nonmarket incentives to cognitive investment (i.e., cognition is also good for other aspects of life besides earnings).

Empirically, providing convincing evidence on the causal effect of retirement is not simple. Individuals may retire because of bad health, declining cognitive ability (reverse causality), or other unobserved factors. Therefore, even the comparison of the same individual over time (i.e., the fixed-effect strategy) may lead to wrong conclusions. For this reason, the large body of epidemiological literature that has investigated this relationship using longitudinal studies (for a recent review, see Meng, Nexø, & Borg, 2017) cannot provide convincing evidence about the causal effect of retirement on cognitive ability. To address the endogeneity of retirement, most recent studies in economics exploit the large early-retirement incentives built into the eligibility rules for public old-age pensions (Gruber & Wise, 1999, 2004) as sources of exogenous variation. Indeed, the presence of large spikes in the retirement hazard at the eligibility ages for early or normal retirement benefits is well documented empirically (see, e.g., Peracchi & Welch, 1994).

Several studies find evidence of negative effects of retirement on cognitive functioning (see, among others, Rohwedder & Willis, 2010; Bonsang, Adam, & Perelman, 2012; Mazzonna & Peracchi, 2012; Clouston & Denier, 2017; Mazzonna & Peracchi, 2017). The main exceptions are Coe and Zamarro (2011) and Coe, von Gaudecker, Lindeboom, and Maurer (2012), who find no evidence of negative effects.

The varying results in the literature mostly reflect differences in model specifications and identification strategies (Nishimura, Oikawa, & Motegi, 2017). For differences in model specifications, it matters whether retirement is treated only as a parallel shift or as a change in the slope of the age profile of cognition. In cross-country studies, it also matters whether country dummies are included in the specification. Further, some of these differences depend on whether the effect of retirement is allowed to be heterogeneous across groups (e.g., education or occupation). Mazzonna and Peracchi (2017) show evidence of large heterogeneity in the effect of retirement across jobs. In particular, for people working in physically demanding jobs, retirement has an immediate beneficial effect on health and cognitive abilities, while it has negative long-run effects on the age profile of health and cognitive abilities for the rest of the workforce.

Regarding differences in identification strategies, the only notable exception to the prevalent use of legislated changes in retirement ages is Coe et al. (2012), who use offers of early retirement windows by the employer instead of age of eligibility for early or normal retirement benefits. Still, differences in the results across studies may arise from the fact that some use only between-country variations (e.g., Coe & Zamarro, 2011), while others use both between- and within-country variations (e.g., Mazzonna & Peracchi, 2012). Finally, although often based on panel surveys such as HRS or SHARE, most studies rely only on cross-sectional differences, given the relevant attrition rate that often affects old-age surveys. Only Bonsang et al. (2012) and Mazzonna and Peracchi (2017) fully exploit the longitudinal dimension of the data.

Overall, there is robust evidence of a negative effect of retirement on cognitive ability, but many open issues remain. First, the local nature of the effect identified by instrumental variable (IV) methods should be recognized (Angrist & Pischke, 2009), as we can identify the average treatment effect only on the subpopulation that reacts to the financial incentive provided by social security regulation. More generally, the instruments currently employed in the literature only exploit exogenous variations in the labor supply of older workers. Given the local nature of IV methods, it would be interesting to find instruments that are able to exploit demand-side variation, such as unanticipated company or plant closures.

Second, different from a past in which most people used to retire directly from full-time work, forms of partial retirement are becoming widespread. Knowing the effect of this kind of transition also might be relevant. Unfortunately, the usual instruments cannot help much to identify the effect of partial retirement separate from that of full retirement.

Finally, the effect of retirement on cognitive ability might be nonlinear. For instance, most literature relies on early retirement incentives, but the effect of raising the retirement age from 59 to 60 may not be the same as raising it from 65 to 66. It is clear that this is an empirical issue that may have important policy implications because many countries are gradually raising the eligibility age for public pension benefits to 66 years or beyond (e.g., Germany, Italy, and the Netherlands).

Conclusions

This account of the economics of cognitive aging has been somewhat selective and has ignored many important issues. Some of them are mentioned here as a reminder for future research.

One of these points is how to mitigate the impact of cognitive aging. The costs associated with cognitive decline include direct monetary costs and indirect costs in terms of extra time and effort devoted to care, changes in behavior, and stress (or even depression) among other family members. These costs may be quite large, and their structure may differ across countries based on culture, history, and other factors. There is also evidence that certain forms of intervention may be effective, but the precise cost-benefit calculations are not easy to perform. One of the reasons for this is that cognitive aging has an impact not only on the individuals affected, but also on their families and the public healthcare system.

Another issue is what determines the passage from normal aging to mild cognitive impairment (MCI), the intermediate stage before the pathological decline of dementia. Understanding the functional and economic implications of MCI is important, given the refinement of the diagnostic criteria for cognitive impairment and earlier-stage Alzheimer’s disease, as well as the emergence of new technologies that may facilitate earlier diagnosis and treatment.

A related issue is insurance. MCI and its treatment carry costs, but too few people buy long-term care (LTC) insurance. For example, Ko (2016) reports that only 13% of the elderly in the United States own LTC insurance. In many other countries, the share of the elderly population covered by LTC insurance is even smaller. Why do so few people buy LTC insurance? And if people do not buy insurance, how are the costs of LTC split between their families and the public healthcare system? The huge variations observed across countries suggest that an analysis of the role of demand-side factors (e.g., cultural or historical) and supply-side factors (e.g., regulations and market structure) may be important.

Finally, the policy implications of cognitive aging are relatively clear in general terms. For example, educational policies, retraining, and regulations are important. But the design of incentives to retraining and the details of regulations are also crucial. As often happens, the devil is in the details—and knowledge of the details is clearly what is missing.

References

Abaluck, J., & Gruber, J. (2011). Choice inconsistencies among the elderly: Evidence from plan choice in the Medicare Part D program. American Economic Review, 101(4), 1180–1210.Find this resource:

Agarwal, S., Driscoll, J. C., Gabaix, X., & Laibson, D. (2009). The age of reason: Financial decisions over the life cycle and implications for regulation. Brookings Papers on Economic Activity, 2, 51–117.Find this resource:

Agarwal, S., & Mazumder, B. (2013). Cognitive abilities and household financial decision-making. American Economic Journal: Applied Economics, 5(1), 193–207.Find this resource:

Alwin, D. F., & Hofer, S. M. (2011). Health and cognition in aging research. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 66(Suppl. 1), i9–i16.Find this resource:

Andersson, O., Holm, H. J., Tyran, J.-R., & Wengström, E. (2016). Risk aversion relates to cognitive ability: Preferences or noise? Journal of the European Economic Association, 14(5), 1129–1154.Find this resource:

Angrist, J. D., & Pischke, J.-S. (2009). Mostly harmless econometrics: An empiricist’s companion. Princeton, NJ: Princeton University Press.Find this resource:

Antonova, L., Bucher-Koenen, T., & Mazzonna, F. (2017). Long-term health consequences of recessions during working years. Social Science & Medicine, 187, 134–143.Find this resource:

Banks, J. (2010). Cognitive function, financial literacy and financial outcomes at older ages: Introduction. Economic Journal, 120(548), F357–F362.Find this resource:

Banks, J., & Mazzonna, F. (2012). The effect of education on old age cognitive abilities: evidence from a regression discontinuity design. Economic Journal, 122(560), 418–448.Find this resource:

Banks, J., O’Dea, C., & Oldfield, Z. (2010). Cognitive function, numeracy and retirement saving trajectories. Economic Journal, 120(548), F381–F410.Find this resource:

Bartel, A. P., & Sicherman, N. (1993). Technological change and retirement decisions of older workers. Journal of Labor Economics, 11(1, Part 1), 162–183.Find this resource:

Behrman, J. R., Mitchell, O. S., Soo, C. K., & Brava, D. (2012). How financial literacy affects household wealth accumulation. American Economic Review, 102(3), 300–304.Find this resource:

Belbase, A., & Sanzenbacher, G. T. (2016). Cognitive aging and ability to work. Technical Report 16–18, Center for Retirement Research at Boston College.Find this resource:

Benjamin, D. J., Brown, S. A., & Shapiro, J. M. (2013). Who is “behavioral”? Cognitive ability and anomalous preferences. Journal of the European Economic Association, 11(6), 1231–1255.Find this resource:

Bishop, N. J., Ehhum-Wilkens, N. D., Haas, S. A., & Kronenfeld, J. J. (2016). Estimating the co-development of cognitive decline and physical mobility limitations in older U.S. adults. Demography, 53(2), 337–364.Find this resource:

Bonsang, E., Adam, S., & Perelman, S. (2012). Does retirement affect cognitive functioning? Journal of Health Economics, 31(3), 490–501.Find this resource:

Bonsang, E., & Dohmen, T. (2015). Risk attitude and cognitive aging. Journal of Economic Behavior & Organization, 112, 112–126.Find this resource:

Börsch-Supan, A., & Schuth, M. (2013). Early retirement, mental health, and social networks. Active Ageing and Solidarity Between Generations in Europe: First Results From SHARE After the Economic Crisis, p. 337.Find this resource:

Börsch-Supan, A., & Weiss, M. (2016). Productivity and age: Evidence from work teams at the assembly line. Journal of the Economics of Ageing, 7, 30–42.Find this resource:

Brown, J. R., Kapteyn, A., Luttmer, E. F., & Mitchell, O. S. (2017). Cognitive constraints on valuing annuities. Journal of the European Economic Association, 15(2), 429–462.Find this resource:

Carpenter, S. M., & Yoon, C. (2011). Aging and consumer decision making. Annals of the New York Academy of Sciences, 1235(1), E1–E12.Find this resource:

Case, A., & Deaton, A. S. (2005). Broken down by work and sex: How our health declines. In Analyses in the economics of aging (pp. 185–212). Chicago: University of Chicago Press.Find this resource:

Case, A., & Paxson, C. (2008). Height, health, and cognitive function at older ages. American Economic Review, 98(2), 463–467.Find this resource:

Choi, S., Kariv, S., Müller, W., & Silverman, D. (2014). Who is (more) rational? American Economic Review, 104(6), 1518–1550.Find this resource:

Christelis, D., Jappelli, T., & Padula, M. (2010). Cognitive abilities and portfolio choice. European Economic Review, 54(1), 18–38.Find this resource:

Clouston, S. A., & Denier, N. (2017). Mental retirement and health selection: Analyses from the U.S. Health and Retirement Study. Social Science & Medicine, 178, 78–86.Find this resource:

Coe, N. B., von Gaudecker, H.-M., Lindeboom, M., & Maurer, J. (2012). The effect of retirement on cognitive functioning. Health Economics, 21(8), 913–927.Find this resource:

Coe, N. B., & Zamarro, G. (2011). Retirement effects on health in Europe. Journal of Health Economics, 30(1), 77–86.Find this resource:

Cook, C. J., & Fletcher, J. M. (2015). Can education rescue genetic liability for cognitive decline? Social Science & Medicine, 127, 159–170.Find this resource:

Cunha, F., & Heckman, J. (2007). The technology of skill formation. American Economic Review, 97(2), 31–47.Find this resource:

Currie, J. (2009). Healthy, wealthy, and wise? Socioeconomic status, poor health in childhood, and human capital development. Journal of Economic Literature, 47(1), 87–122.Find this resource:

Dalgaard, C.-J., & Strulik, H. (2014). Optimal aging and death: Understanding the Preston curve. Journal of the European Economic Association, 12(3), 672–701.Find this resource:

Deaton, A., & Cartwright, N. (2016). Understanding and misunderstanding randomized controlled trials. Technical Report 22595, National Bureau of Economic Research (NBER).Find this resource:

Deaton, A., & Paxson, C. (1994). Intertemporal choice and inequality. Journal of Political Economy, 102(3), 437–467.Find this resource:

de Jager, C. A., Budge, M. M., & Clarke, M. (2003). Utility of TICS-M for the assessment of cognitive function in older adults. International Journal of Geriatric Psychiatry, 18(4), 318–324.Find this resource:

Dohmen, T., Falk, A., Huffman, D., & Sunde, U. (2010). Are risk aversion and impatience related to cognitive ability? American Economic Review, 100(3), 1238–1260.Find this resource:

Eibich, P. (2015). Understanding the effect of retirement on health: Mechanisms and heterogeneity. Journal of Health Economics, 43, 1–12.Find this resource:

Fang, H., Keane, M. P., & Silverman, D. (2008). Sources of advantageous selection: Evidence from the Medigap insurance market. Journal of Political Economy, 116(2), 303–350.Find this resource:

Finke, M. S., Howe, J. S., & Huston, S. J. (2016). Old age and the decline in financial literacy. Management Science, 63(1), 213–230.Find this resource:

Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-mental state: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12(3), 189–198.Find this resource:

Galenson, D. W., & Weinberg, B. A. (2000). Age and the quality of work: The case of modern American painters. Journal of Political Economy, 108(4), 761–777.Find this resource:

Greenwood, P. (2007). Functional plasticity in cognitive aging: Review and hypothesis. Neuropsychology, 21(6), 657.Find this resource:

Griliches, Z., & Mason, W. M. (1972). Education, income, and ability. Journal of Political Economy, 80(3, Part 2), S74–S103.Find this resource:

Grossman, M. (1972). On the concept of health capital and the demand for health. Journal of Political Economy, 80(2), 223–255.Find this resource:

Gruber, J., & Wise, D. (1999). Social security programs and retirement around the world. Chicago: University of Chicago Press.Find this resource:

Gruber, J., & Wise, D. (2004). Social security programs and retirement around the world: Micro estimation. Chicago: University of Chicago Press.Find this resource:

Haltiwanger, J. C., Lane, J. I., & Spletzer, J. R. (1999). Productivity differences across employers: The roles of employer size, age, and human capital. American Economic Review, 89(2), 94–98.Find this resource:

Havari, E., & Peracchi, F. (2017). Growing up in wartime: Evidence from the era of the two world wars. Economics & Human Biology, 25(2), 9–32.Find this resource:

Heckman, J. J. (2007). The economics, technology, and neuroscience of human capability formation. Proceedings of the National Academy of Sciences, 104(33), 13250–13255.Find this resource:

Hellerstein, J. K., Neumark, D., & Troske, K. R. (1999). Wages, productivity, and worker characteristics: Evidence from plant-level production functions and wage equations. Journal of Labor Economics, 17(3), 409–446.Find this resource:

Hertzog, C., Kramer, A. F., Wilson, R. S., & Lindenberger, U. (2008). Enrichment effects on adult cognitive development: Can the functional capacity of older adults be preserved and enhanced? Psychological Science in the Public Interest, 9(1), 1–65.Find this resource:

Hess, T. M. (2014). Selective engagement of cognitive resources: Motivational influences on older adults’ cognitive functioning. Perspectives on Psychological Science, 9(4), 388–407.Find this resource:

Horn, J. L., & Cattell, R. B. (1967). Age differences in fluid and crystallized intelligence. Acta Psychologica, 26(2), 107–129.Find this resource:

Hultsch, D. F., Hertzog, C., Small, B. J., & Dixon, R. A. (1999). Use it or lose it: Engaged lifestyle as a buffer of cognitive decline in aging? Psychology and Aging, 14(2), 245.Find this resource:

Jones, B. F. (2010). Age and great invention. Review of Economics and Statistics, 92(1), 1–14.Find this resource:

Kämpfen, F., & Maurer, J. (2016). Time to burn (calories)? The impact of retirement on physical activity among mature Americans. Journal of Health Economics, 45, 91–102.Find this resource:

Kirkwood, T. B., & Austad, S. N. (2000). Why do we age? Nature, 408(6809), 233–238.Find this resource:

Ko, A. (2016). An equilibrium analysis of the long-term care insurance market. Technical report, Department of Economics, University of Pennsylvania.Find this resource:

Korniotis, G. M., & Kumar, A. (2011). Do older investors make better investment decisions? Review of Economics and Statistics, 93(1), 244–265.Find this resource:

Koscielniak, M., Rydzewska, K., & Sedek, G. (2016). Effects of age and initial risk perception on balloon analog risk task: The mediating role of processing speed and need for cognitive closure. Frontiers in Psychology, 7, 659.Find this resource:

Lazear, E. P., & Moore, R. L. (1984). Incentives, productivity, and labor contracts. Quarterly Journal of Economics, 99(2), 275–296.Find this resource:

Leshner, A., Landis, S., Stroud, C., & Downey, A. (2017). Preventing cognitive decline and dementia: A way forward. Washington, DC: National Academies Press.Find this resource:

Livingston, G., Sommerlad, A., Orgeta, V., Costafreda, S. G., Huntley, J., Ames, D., . . . Mukadam, N. (2017). Dementia prevention, intervention, and care. Lancet, 390(10113), 2673–2734.Find this resource:

Lusardi, A., & Mitchell, O. S. (2007). Baby boomer retirement security: The roles of planning, financial literacy, and housing wealth. Journal of Monetary Economics, 54(1), 205–224.Find this resource:

Lusardi, A., Mitchell, O. S., & Curto, V. (2014). Financial literacy and financial sophistication in the older population. Journal of Pension Economics & Finance, 13(4), 347–366.Find this resource:

Mazzonna, F., & Peracchi, F. (2012). Ageing, cognitive abilities, and retirement. European Economic Review, 56(4), 691–710.Find this resource:

Mazzonna, F., & Peracchi, F. (2017). Unhealthy retirement? Journal of Human Resources, 52(1), 128–151.Find this resource:

McFadden, D. (2008). Human capital accumulation and depreciation. Review of Agricultural Economics, 30(3), 379–385.Find this resource:

Meng, A., Nexø, M. A., & Borg, V. (2017). The impact of retirement on age related cognitive decline–a systematic review. BMC Geriatrics, 17(1), 160.Find this resource:

Montizaan, R., Cörvers, F., & De Grip, A. (2010). The effects of pension rights and retirement age on training participation: Evidence from a natural experiment. Labour Economics, 17(1), 240–247.Find this resource:

Ngandu, T., Lehtisalo, J., Solomon, A., Levälahti, E., Ahtiluoto, S., Antikainen, R., . . . Kivipelto, M. (2015). A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): A randomised controlled trial. Lancet, 385(9984), 2255–2263.Find this resource:

Nishimura, Y., Oikawa, M., & Motegi, H. (2017). What explains the differences in the effects of retirement on health? Evidence from global aging data. Journal of Economic Surveys, 31.Find this resource:

Peracchi, F., & Welch, F. (1994). Trends in labor force transitions of older men and women. Journal of Labor Economics, 12(2), 210–242.Find this resource:

Petersen, R. C. (2011). Mild cognitive impairment. New England Journal of Medicine, 364(23), 2227–2234.Find this resource:

Rabbitt, P., Diggle, P., Holland, F., & McInnes, L. (2004). Practice and drop-out effects during a 17-year longitudinal study of cognitive aging. Journal of Gerontology: Psychological Sciences, 59B(2), 84–97.Find this resource:

Rebok, G. W., Ball, K., Guey, L. T., Jones, R. N., Kim, H.-Y., King, J. W., . . . Willis, S. L. (2014). Ten-year effects of the ACTIVE cognitive training trial on cognition and everyday functioning in older adults. Journal of the American Geriatrics Society, 62(1), 16–24.Find this resource:

Rohwedder, S., & Willis, R. J. (2010). Mental retirement. Journal of Economic Perspectives, 24(1), 119–138.Find this resource:

Rohwer, G. (2016). A note on the dependence of health on age and education. Demography, 53(2), 325–335.Find this resource:

Salthouse, T. A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103(3), 403.Find this resource:

Salthouse, T. A. (2001). Structural models of the relations between age and measures of cognitive functioning. Intelligence, 29(2), 93–115.Find this resource:

Salthouse, T. A. (2010). Influence of age on practice effects in longitudinal neurocognitive change. Neuropsychology, 24(5), 563–572.Find this resource:

Salthouse, T. A. (2012). Consequences of age-related cognitive declines. Annual Review of Psychology, 63, 201–226.Find this resource:

Salthouse, T. A., Schroeder, D. H., & Ferrer, E. (2004). Estimating retest effects in longitudinal assessments of cognitive functioning in adults between 18 and 60 years of age. Developmental Psychology, 40(5), 813.Find this resource:

Schaie, K. W. (1996). Intellectual development in adulthood: The Seattle Longitudinal Study. Cambridge, UK: Cambridge University Press.Find this resource:

Skirbekk, V., Loichinger, E., & Weber, D. (2012). Variation in cognitive functioning as a refined approach to comparing aging across countries. Proceedings of the National Academy of Sciences, 109(3), 770–774.Find this resource:

Spitz-Oener, A. (2006). Technical change, job tasks, and rising educational demands: Looking outside the wage structure. Journal of Labor Economics, 24(2), 235–270.Find this resource:

Stern, Y. (2002). What is cognitive reserve? Theory and research application of the reserve concept. Journal of the International Neuropsychological Society, 8(3), 448–460.Find this resource:

Stine-Morrow, E. A. (2007). The Dumbledore hypothesis of cognitive aging. Current Directions in Psychological Science, 16(6), 295–299.Find this resource:

van Charante, E. P. M., Richard, E., Eurelings, L. S., van Dalen, J.-W., Ligthart, S. A., Van Bussel, E. F., . . . van Gool, W. A. (2016). Effectiveness of a 6-year multidomain vascular care intervention to prevent dementia (prediva), a cluster-randomised controlled trial. Lancet, 388(10046), 797–805.Find this resource:

van den Berg, G. J., Deeg, D. J., Lindeboom, M., & Portrait, F. (2010). The role of early-life conditions in the cognitive decline due to adverse events later in life. Economic Journal, 120(548), F411–F428.Find this resource:

Yaffe, K., Fiocco, A. J., Lindquist, K., Vittinghoff, E., Simonsick, E. M., Newman, A. B., . . . Health ABC Study. (2009). Predictors of maintaining cognitive function in older adults: The Health ABC Study. Neurology, 72(23), 2029–2035.Find this resource:

Notes:

(1.) A predecessor of the HRS is the US Retirement History Longitudinal Survey, which collected six waves of data from 1969 to 1979 but did not contain measures of cognitive ability.

(2.) The idea is that a stroke of a given magnitude may produce impairment in one patient, but minimal effects on another patient with a higher cognitive reserve.

(3.) For the discussion that follows, we largely draw upon our previous work (i.e., Mazzonna & Peracchi, 2012), which interested readers may refer to for more details.

(4.) Andersson et al. (2016) argue that this relation may be spurious because of nonrandom errors in choice behavior under risk.

(5.) Finke et al. (2016) estimate that households where the head is aged 60+ hold 51% of all financial wealth in the United States.

(6.) Using data at the work-team level allows one to take into account the fact that workers often work in a team, thus affecting one another’s productivity.