Dementia affects an estimated 6 million Americans, and is the leading cause of neurological disability and the third leading cause of overall disability among older adults (over 65 years). Employment near retirement age appears to reduce the risk of cognitive decline, which is a key precursor to dementia (Coe et al. 2012, Bonsang and Perelman 2012, Celidoni et al. 2017, Mazzonna and Peracchi 2017). The potential link between work and forestalling cognitive decline and dementia suggests there may be large social benefits from higher employment at older ages. 

However, we know less about whether employment at mid-life, when cognitive decline already starts, delays cognitive decline. Whereas severe cognitive impairment (and dementia onset) typically occurs after age 65, pathologically meaningful cognitive decline is detectable as early as age 40 (Salthouse 2009, Singh-Manoux et al. 2012, Yang et al. 2024), and over 32% of adult men aged 51-64 years in the US are not working. The goal of our new research (Kouchekinia et al. 2026) is to assess the causal impact of employment on cognitive decline to employment before the retirement window.

In the face of population ageing, policymakers are already focused on promoting employment beyond normal retirement ages, motivated by many factors including – in the US – the solvency of Social Security. But given the high social costs of dementia and cognitive decline, evidence of an employment–cognition link at pre-retirement ages could alter the benefit-cost calculation regarding policies intended to increase employment at these younger ages – such as reforms to Social Security Disability Insurance (Maestas 2019) or broader hiring subsidies – although ideally we would like more evidence of cognitive benefits of employment for the specific target populations. 

The potential benefits of higher employment at pre-retirement ages could extend beyond improved cognitive functioning, and also include prolonged work, less reliance on disability insurance, and enhanced retirement security via later claiming of Social Security benefits and greater accumulation of savings from working longer. In addition, of course, the evidence that exogenous increases in employment around retirement ages helps forestall cognitive decline points to an additional benefit from policies that push back retirement (which are generally being adopted for different reasons – to improve the finances of public pension systems). Importantly, however, the ability of policy to boost employment at younger, pre-retirement ages, compared to boosting employment at ages very close to retirement, is likely much stronger because employment rates are much higher and public pension payments are not yet available to support non-employment.

Prior work

Descriptive work finds that working longer is associated with reduced cognitive decline (e.g. Dufouil et al. 2014, Leist et al. 2013, Lupton et al. 2009, Roberts et al. 2011). But this work does not establish a causal impact of employment in reducing cognitive decline, because unmeasured factors related to mental health – such as adverse mental health shocks – may drive decisions about the timing of leaving paid work, or lead to sub-par work performance that can precede leaving a job. Hence, researchers have used quasi-random variation in employment to isolate the causal effect of employment. 

Most of these empirical studies with arguably more rigorous causal study designs focus on policy changes to ages at which retirement benefits or Social Security are received, or other exogenous changes affecting employment around the retirement age – including discontinuities created by retirement and pension policies (e.g. Atalay et al. 2019, Bonsang et al. 2012 Celidoni et al. 2017, Coe et al. 2012). This work generally shows that extending employment near retirement ages delays the onset of cognitive decline, consistent with the hypothesis that stimulating cognitive and/or social activities in the workplace may delay cognitive decline (Salthouse 1991). 

Because the studies cited above rely on plausibly exogenous retirement policy changes, they focus on a narrow age range (the mid- to late-60s). Thus, it remains an open question whether employment at earlier pre-retirement ages delays cognitive decline. 

Our approach

Our research exploits a different source of exogenous variation in employment than does research using retirement reforms or sharp policy changes at particular ages.  We use 12 waves of the Health and Retirement Study (HRS) dataset spanning 1996-2018. The HRS has a large number of older adults, rich information on employment history and trajectories of cognitive decline, and broad geographic coverage.

Our primary research strategy uses County Business Patterns data to construct a predictor of employment changes at the local labour market level (commuting zones, or CZs).  In particular, we predict changes in local labour demand based on baseline variation in the composition of industries across CZs, which we use to weight national industry-specific growth rates to predict changes in local labour demand. Thus, for example, if a CZ has a large share of employment in the auto industry, and auto industry employment declines nationally, then that CZ would have a predicted negative shock to labour demand. We then use predicted changes in local labour demand to generate variation in employment of older workers that is not related to other changes that might drive both their employment and their cognition. This procedure relies on ‘exogenous’ changes in employment, rather than, say, a decline in cognitive capacity that leads them to leave the workforce – exactly the kind of variation that would provide misleading evidence of a positive causal effect of employment on cognition.

There is substantial geographic variation in CZ level employment that supports the utility of local area shocks as an instrument. Figure 1, Panel A, shows this for men ages 51-64. We focus on this group because the predicted local labour demand shocks work best for them for predicting actual employment changes. The different shading illustrates the geographic variation in employment change, as well as large reductions in employment for 51-64 year-old men in many areas of the southeast. Panel B displays the predicted change in the employment-to-population ratio by CZ. Note there is a correlation with the geography of realised changes in employment, which is the basis for our empirical strategy. 

Figure 1 Change in (employment/population) by commuting zone

A) Men ages 51-64

B) Shift-share instrument prediction

The evidence

We find that employment declines driven by local labour demand shocks are associated with cognitive decline of men ages 51-64. There is a straightforward way to display this visually. Among the HRS participants in our sample, cognitive performance gradually decreases with age. In Figure 2, we display the trajectory of the global (Langa-Weir) cognitive score available in the HRS, broken down by CZs in the first to fourth quartiles of employment growth (from lowest to highest). Respondents in the CZs in the lowest quartile of employment growth start their early 50s with scores at the same level as the other three quartiles. However, a gap emerges over time, particularly as respondents enter potential retirement years. Relative to the top quartile, there is a full one-point gap for participants in their 70s between the bottom and top quartiles of CZ employment growth. This suggests a role for work – related to local variation in employment – in slowing cognitive decline.

Figure 2 Cognitive trajectories (global cognitive score) based on quartiles of CZ employment growth

Our estimate that identifies the causal effect of employment conveys the same message: higher employment in men’s pre-retirement years helps forestall cognitive decline. Moreover, in our view the magnitude is sizable. To put the estimate in perspective, in our sample period, the employment rate decline of men from age 51 to 61 (before eligibility for Social Security at age 62) is 18 percentage points. Suppose this employment decline was halved, so it fell by only 9 percentage points. Based on our estimate, this change would boost the average cognitive score by one-tenth of a standard deviation. This turns out to be approximately the same as the average cognitive decline that men experience over this age range, implying that cutting the pre-retirement employment rate decline in half would forestall the average cognitive decline that occurs over this age range.  

Conclusion

Our work has some limitations, including the inability to determine which aspects of work, or of employment loss, causally affect cognition. We suspect that the relation between work and cognition at pre-retirement ages is heterogeneous in ways that depend on both the nature of work and the characteristics of the people studied. More research would be useful in understanding other ways that both work, and non-work activities, might help forestall cognitive decline. Finally, it is always useful to be cautious about drawing strong policy conclusions from limited amounts of evidence, and as we have explained, there is little extant work on the causal impact of work on cognition at pre-retirement ages. Nonetheless, our findings – especially if replicated by others – hold clear policy implications. Federal efforts to promote work at pre-retirement ages would not only reduce reliance on disability insurance and enhance retirement security, but would also promote healthy aging through delaying cognitive decline. 

References

Atalay, K, G F Barrett and A Staneva (2019), “The effect of retirement on elderly cognitive functioning”, Journal of Health Economics 66: 37-53.

Bonsang, E, S Adam and S Perelman (2012), “Does retirement affect cognitive functioning?”, Journal of Health Economics 31: 490-501.

Celidoni, M, C Dal Bianco and G Weber (2017), “Retirement and cognitive decline. A longitudinal analysis using SHARE data”, Journal of Health Economics 56: 113-125.

Coe, N B, H-M von Gaudecker, M Lindeboom and J Maurer (2012), “The effect of retirement on cognitive functioning”, Health Economics 21: 913–927.

Dufouil, C (2014), “Older age at retirement is associated with decreased risk of dementia”, European Journal of Epidemiology 29: 353–361.

Kouchekinia, N, D Neumark and T Bruckner (2026), “Does employment slow cognitive decline? Evidence from labor market shocks”, NBER Working Paper No. 35117.

Leist, A K, M M Glymour, J P Mackenbach, F J van Lenthe and M Avendano (2013), “Time away from work predicts later cognitive function: Differences by activity during leave”, Annals of Epidemiology 23: 455–462.

Lupton, et al. (2009), “Education, occupation and retirement age effects on the age of onset of Alzheimer’s disease”, International Journal of Geriatric Psychiatry 25: 30–36.

Maestas, N (2019), “Identifying work capacity and promoting work: A strategy for modernizing the SSDI program”, Annals of the American Academy of Political and Social Science 686: 93–120.

Mazzonna, F and F Peracchi (2017), “Unhealthy retirement?”, Journal of Human Resources 52: 128–151.

Roberts, B A, R Fuhrer, M Marmot and M Richards (2011), “Does retirement influence cognitive performance? The Whitehall II study”, Journal of Epidemiology and Community Health 65: 958–963.

Salthouse, T A (2009), “When does age-related cognitive decline begin?”, Neurobiology of Aging 30: 507–514.

Singh-Manoux, A, et al. (2012), “Timing of onset of cognitive decline: Results from Whitehall II prospective cohort study”, BMJ 344.

Yang, Y C, et al. (2024), “An early and unequal decline: Life course trajectories of cognitive aging in the United States”, Journal of Aging and Health 36(3–4): 230–245.