Changes in the health and longevity of the elderly in the US

Spending on public pensions through Social Security and on elderly healthcare through Medicare accounted for over one-third of federal expenditures in the US by 2024. This column explores how improvements in elderly health and life expectancy affected the relative finances of Social Security and Medicare. The elderly are not only living longer but also are in better health, so expected lifetime Medicare spending per elderly individual has grown much slower than expected lifetime Social Security spending. The authors analyse the forces that affect the optimal allocation of a given public budget across these two social insurance programmes.

The last several decades have seen considerable improvements in elderly health and life expectancy in most high-income countries. This is great for elderly wellbeing but challenging for government budgets. In the US, by 2024, spending on public pensions through Social Security and on elderly healthcare through Medicare accounted for over one-third of federal expenditures (Congressional Budget Office 2025), generating considerable discussion of possible entitlement reforms, such as raising the age of benefit eligibility, cutting benefit levels, and raising payroll taxes on the working population (e.g. National Research Council 2012, Poterba 2014, Sheiner 2021). 

But how have demographic changes affected the relative finances of Social Security and Medicare? We explored this question in a recent paper (Einav and Finkelstein 2026a). We found that the elderly are not only living longer but also are in better health, so expected lifetime Medicare spending by the elderly has grown much slower than expected lifetime Social Security payments.

To calculate this, we used data from the Medicare Current Beneficiary Surveys from 1992 through 2019. This rotating panel data contain detailed information on health conditions and functional limitations, healthcare spending, and mortality. We used these data to consider, for different years, the life of a hypothetical 66-year-old who faces the age-specific rates of morbidity and mortality experienced by the cross-section of individuals at different ages.

Between 1993 and 2017, remaining life expectancy for the average 66-year-old increased by 2.4 years, or about 14%, from 17.4 to almost 20 years (Figure 1). Heuristically, this increase in life expectancy can be attributed to some combination of ‘delayed’ ageing (individuals remaining healthier for more years) and ‘prolonged’ dying (individuals surviving for more years in an unhealthy state through, say, technological improvements in extending the life of the sick). Strikingly, Figure 1 shows that the 2.4-year increase in remaining life expectancy consists entirely of increases in healthy life-years (i.e. morbidity group 1), that is, expected years free of any of the physical or cognitive functional limitations that commonly accrue with advanced age. Moreover, expected time spent in very bad health – in which individuals have severe physical or cognitive limitations (morbidity group 5) – actually declined by 0.6 years (about 30%).

Figure 1 Changes in life expectancy at age 66, overall and by morbidity group

What are the implications for Social Security and Medicare spending? Since Social Security is an annuity that pays out as long as the individual is alive, calculating the impact of demographic change on expected lifetime Social Security payments is relatively straightforward: they increase at the same rate as life expectancy, by about 14%.

For healthcare spending, however, the impact of demographic change is much less clear, and depends critically on how health conditional on age has evolved, and how health maps into medical spending. We therefore used the rich data on health measures as inputs to train a standard machine-learning algorithm to predict annual spending (both overall and by spending category). The resultant predictive spending model allows us to analyse how, holding the mapping from health and age to healthcare spending fixed, observed changes in age-specific morbidity and mortality transitions have affected expected elderly lifetime healthcare spending. This exercise allows us to isolate the impact of changes in morbidity and mortality on lifetime spending, while abstracting from the many changes in the practice and pricing of medicine that occurred over the same time period, for example because of changes in technology and healthcare policy.

As shown in Figure 2, the combination of increased life expectancy and improved health conditional on age between 1993 and 2017 results in an increase in lifetime healthcare spending per elderly individual of about 6%. Notably, this increase in expected lifetime healthcare spending is less than a half of the 14% increase in life expectancy (and hence expected lifetime Social Security spending).

Figure 2 Changes in expected lifetime healthcare spending at 66

An important driver of the slower healthcare spending growth is the pronounced decline in time spent in the worst of the five morbidity states; this state is associated with particularly high healthcare spending, both overall and especially for nursing-home and home-healthcare. Indeed, because we find that poor health, rather than age, is the key predictor of nursing-home and home-healthcare, Figure 2 shows that the net impact of changes in mortality and morbidity is, in fact, to (slightly) decrease expected lifetime spending on nursing-home and home-healthcare over our sample period, despite the substantial increase in overall remaining life expectancy. We also find that the increase in expected lifetime healthcare spending is about five times higher for men than women and about three times higher for higher-income individuals than lower-income individuals.

Taken together, the demographic changes in mortality and morbidity increased lifetime public spending on the average elderly individual between 1993 and 2017 by about 11%, with expected Social Security spending growing more than twice as fast as expected Medicare spending (14% compared to 6%).

These increases reflect the ‘mechanical’ impact of demographic changes under passive, status-quo programme rules. In the final part of our analysis, we analyse forces that affect the optimal allocation of a given public budget across these two social insurance programmes. Specifically, we write down and calibrate a stylised dynamic consumption-saving model of a risk-averse retiree facing stochastic mortality and health and use this model to analyse consumer welfare under alternative budget-neutral policy parameters that re-allocate public spending across Social Security and Medicare.

The model clarifies two advantages of an incremental expected dollar of Medicare coverage relative to Social Security. First, Social Security annuity payments – by design – only insure individuals against the risk associated with the number of years they are alive. Medicare coverage serves a similar role, but also helps reduce consumption volatility across states of differential health realisations, and therefore provides a more complete insurance product. 

Second, as we showed in Einav and Finkelstein (2026b), the optimal public annuity is back-loaded, although current Social Security payments are constant with respect to age. This turns out to create a second advantage for Medicare: because health deteriorates with age, Medicare coverage naturally delivers more resources to older individuals, providing an indirect form of back-loaded payments. As a result, absent moral hazard, the optimum is to allocate the government budget entirely to Medicare until Medicare covers all medical expenses, and only then to allocate any remaining funds to Social Security. To arrive at an interior solution, we also allow for moral hazard effects of health insurance. This creates a disadvantage for Medicare relative to Social Security, since it induces inefficient medical spending. We illustrate via stylised calibrations how the demographic changes we estimate can affect this interior solution.

While the modelling exercise is highly stylised and illustrative in nature, it highlights the importance of studying these two programmes jointly and the extent to which programme design should adjust to the non-trivial changes in morbidity and mortality that occurred over the last several decades.

Source : VOXeu