Longevity Risk and Outliving Your Savings: What the Evidence Shows
Key takeaways · 14 min read
- Headline life expectancy figures usually assume today’s death rates never improve. Projections that allow for improvement give a UK 65-year-old man 19.8 more years, against 18.6 on current rates.
- On the US period life table, about one in four 65-year-old men and one in three women reach 90; for a couple of that age, the chance that at least one does is about half.
- In England, people in their 50s and 60s expected to live 6.1 (men) to 7.7 (women) fewer years than life tables suggested.
- The opposite risk is real too: US retirees who started with at least $500,000 had spent only 11.8% of it, at the median, after 20 years.
Most financial risks are about markets. Longevity risk is about you: the chance of living longer than your money lasts. It is an odd risk, because the bad outcome is also the one most people hope for. And it is hard to plan for, because nobody knows their own date of death, and the averages that are supposed to help are widely misread.
The research points two ways at once. People tend to underestimate how long they will live, which makes them spend and insure as if retirement will be short. Yet many retirees with savings barely touch them, and seem to be guarding against a long life with the most expensive tool available: not spending.
This article covers how life expectancy is measured, how people judge their own survival, why so few buy longevity insurance, what research says about claiming Social Security, and the evidence on underspending. It is general information, not financial advice.
Two kinds of life expectancy
The life expectancy figures that make headlines are usually period figures. They answer a hypothetical question: how long would someone live if the death rates of one year applied for the rest of their life? For a 65-year-old, that means assuming that at 85 they will face the death rates of today’s 85-year-olds, who were born two decades earlier and grew up with different medicine.
Cohort figures instead project how death rates for that generation are likely to change. The UK’s Office for National Statistics says plainly that a lower period figure does not mean a baby born that year will live a shorter life. For planning a retirement, the cohort figure is the more relevant one, although it rests on assumptions about the future.
How many more years at 65?
Remaining life expectancy at age 65 on current death rates (period) and allowing for projected improvement.
| Measure | Men | Women |
|---|---|---|
| England and Wales, period, 2021–2023 | 18.6 | 21.1 |
| UK, cohort, aged 65 in 2023 | 19.8 | 22.5 |
| United States, period, 2023 | 18.1 | 20.7 |
| United States, SSA estimate for those turning 65 in 2026 | 19.2 | 21.8 |
Sources: ONS National life tables, England and Wales, 2021 to 2023 (2024); ONS Past and projected period and cohort life tables, 2022-based (2025); SSA actuarial period life table 2023; SSA retirement planner (ages 84.2 and 86.8).
The gap at 65 is about a year. It sounds small, but a year of spending is a large share of most people’s savings, and the cohort figures are themselves averages around a wide spread. The ONS projects that 11.5% of boys and 17.9% of girls born in 2023 will reach 100.
The odds of a long life
Averages hide the part that matters for longevity risk: the tail. Planning for the average age at death means planning to run out of money, at around that age, in roughly half of cases. A more useful number is the chance of reaching a given age. The US Social Security actuaries’ 2023 period table can be used to work this out for people who have already reached 65.
Chance that a 65-year-old reaches 90 or 95
Calculated from the US period life table for 2023. Current death rates; actual odds for today’s 65-year-olds are probably a little higher.
Source: SSA actuarial life table, period 2023 (as used in the 2026 Trustees Report). ROR Labs calculation; the couple figure assumes the two lifespans are independent, which they are not quite.
These are population averages. Health, income, education and smoking all shift individual odds, often by years. People with money to invest also tend to live longer than the population as a whole. In the English study described below, the researchers estimated that their sample of people living at home had death rates about 70% of those in national tables.
People expect to die sooner
The economists Cormac O’Dea and David Sturrock used the English Longitudinal Study of Ageing, in which 16,345 people aged 50 and over answered questions such as “What are the chances that you will live to be 75 or more?” over seven survey waves. They compared the answers with official life tables adjusted to the sample’s actual death rates.
People in their early 60s underestimated their chance of reaching 75 by around 25 to 30 percentage points. Those in their late 60s and early 70s underestimated their chance of reaching 85 by 15 to 20 points. Only people in their late 80s were about right on average, and the very old became optimistic. Overall, men expected to live 6.1 fewer years than the tables suggested, and women 7.7 fewer.
Survival pessimism in England
How subjective expectations compared with adjusted life tables among people aged 50 and over.
Source: O’Dea C, Sturrock D, Review of Economics and Statistics (2023), doi:10.1162/rest_a_01048. English Longitudinal Study of Ageing, waves 1 to 7.
US data show a milder version. In the Health and Retirement Study, women approaching retirement understated their chance of reaching 75 by about 10 percentage points, while men were about right. Survival questions are imperfect: in the English data, about one in seven answers showed clear signs of misunderstanding. But the direction matters: someone who expects to die at 80 will judge a retirement plan, and any form of lifetime income, against the wrong horizon.
The annuity puzzle
In 1965 the economist Menahem Yaari showed that, under idealised conditions, a retiree with no wish to leave money to heirs should convert all of their savings into a life annuity, an income that lasts as long as they do. Annuities pool longevity risk: those who die early subsidise those who live long. Yet few people buy them voluntarily, a gap known as the annuity puzzle.
There are rational reasons. Social Security and state pensions already provide an inflation-protected annuity for most people. Private annuities carry insurers’ costs and are priced for buyers who tend to live longer than average. Many retirees also want cash for emergencies, care or heirs. O’Dea and Sturrock found that 88% of their sample would see a fairly priced annuity as a poor deal, because they expected to die sooner. In one version of their model, the share of wealth annuitised fell from 69% to 43% once people’s own beliefs were used. That drop was about as large as the effect of typical pricing loads, which cut it to 38%. These are model results, not observed purchases.
Framing matters too. In a 2008 survey experiment by Jeffrey Brown and colleagues, 72% of respondents preferred a life annuity over a savings account when the choice was described in terms of what they could spend. Described in investment terms, as returns and the risk of dying early and “losing”, only 21% did.
Same product, different frame
Share of survey respondents choosing a life annuity over a savings account.
Source: Brown JR, Kling JR, Mullainathan S, Wrobel MV, American Economic Review 98(2):304–309 (2008). Hypothetical choices, not purchases.
The Social Security decision
For most Americans the largest longevity decision is not about a private product at all. Delaying Social Security is, in effect, buying more of an inflation-linked annuity from the government. For anyone born in 1960 or later, claiming at 62 pays 70% of the full benefit, while waiting until 70 pays 124% of it, a monthly cheque about 77% larger for life.
Monthly benefit by claiming age
Share of the full retirement benefit paid for life, for people with a full retirement age of 67.
Source: Social Security Administration, early retirement reduction and delayed retirement credit (8% a year for those born in 1943 or later). Bars scaled to the age-70 amount.
Early claiming has been falling. Citing SSA data, the economist Sita Slavov reports that 40% of eligible 62-year-old men and 45% of women claimed at 62 in 2000–2004, against 30% and 33% in 2010–2014. A modelling study by David Altig, Laurence Kotlikoff and Victor Ye estimated that more than 90% of workers aged 45 to 62 would do best to wait until 70, but only about one in ten appear to, at a median lifetime cost of $182,370. That is a model result, sensitive to assumptions about interest rates, health and spending needs.
Early claiming is not always a mistake. People in poor health, with no other savings or with physically demanding jobs often have good reasons. A matched comparison by the AARP Public Policy Institute found that age-62 claimants had lower income into their 70s, but no difference in mortality or self-reported financial hardship.
The other risk: spending too little
If people expect short lives, the standard economic model predicts they will spend too fast. Many do not. Using the US Health and Retirement Study, the Employee Benefit Research Institute tracked how much of their non-housing assets retirees had spent down. Those who started retirement with at least $500,000 had spent 11.8% of it within 20 years, at the median. About a third of retirees had more assets than when they started.
How much of their savings retirees spent
Median share of non-housing assets spent down, by assets just before retirement and by whether the household had a pension.
Source: Banerjee S, EBRI Issue Brief (3 April 2018), Health and Retirement Study and Consumption and Activities Mail Survey, 2015 dollars. The under-$200,000 figure is given as “about one-quarter”.
Households with a traditional pension spent down the least: 4% at the median over 18 years, against 34% for those without one. The median retiree spent about as much as their regular income and left savings largely alone. Guaranteed income seems to give people permission to spend; a pot of savings, it seems, mostly gets guarded.
Some of this is sensible. Long-term care can be very expensive, and savings are a buffer against it. Some people want to leave money to their children. But some underspending looks like fear rather than plan. In the Prudential survey in the news card above, about 44% of people with substantial assets said they did not know how long their money needed to last.
What a plan cannot know
The well-known 4% rule, from a 1994 study by the planner William Bengen, used historical US market returns to ask what starting withdrawal would have lasted 30 years. It is a test of market risk over a fixed horizon. It does not tell anyone how long that horizon should be, and a 65-year-old couple has about even odds that one of them will still be alive at 90, 25 years later. The order in which returns arrive is a separate risk, covered in another article.
Cohort life expectancy is also a model. The ONS publishes high and low variants alongside its main projection; the high one assumes death rates eventually improve by 1.5% a year. Any figure for how long a particular person will live is a probability distribution, not a date. The useful question is not whether a plan survives the average lifespan, but how it behaves if you live to 95, and whether you are giving up a good life now to insure against an outcome that is far from certain.
What the evidence suggests
Retirement can last much longer than headline averages suggest: on current US death rates, about a third of 65-year-old women reach 90. People in their 50s and 60s systematically underestimate their own survival, and that pessimism helps explain why few buy longevity insurance. At the same time, many retirees with savings spend very little of them and keep a cushion that is never used. Both risks are real. The research describes the trade-off well; it cannot say where an individual should sit on it, and most of the striking figures here come from models.
What to do
These are planning checks, not investment recommendations. Steps sorted by when they matter. Each one says how strong the evidence is: Measured Official advice Untested
Before
- Test how your plan behaves if you live to 95, not just to average life expectancy. About a third of 65-year-old US women reach 90. Measured see Sources
- For a couple, plan for the survivor. A 65-year-old couple has about even odds that one of them reaches 90. Measured see Sources
- Count state and Social Security pensions as the lifetime income they are before deciding whether to buy more. Official advice see Sources
What not to do
- Do not treat the 4% rule as an answer to how long you will live. It tested market risk over a fixed 30 years. Measured see Sources
Not financial advice. Decisions about annuities and drawdown belong with a regulated adviser.
Questions people ask
What is longevity risk?
The risk of living longer than your savings and income can support. It grows as fewer people have traditional pensions that pay for life.
What is the difference between period and cohort life expectancy?
Period life expectancy assumes today’s death rates never change. Cohort life expectancy allows for projected improvements, so it is usually higher and more relevant for planning.
How likely is a 65-year-old to reach 90?
On the 2023 US period table, about 24% of men and 35% of women. For a couple of that age, the chance that at least one reaches 90 is about half.
Why do so few people buy annuities?
Research points to existing state pensions, pricing costs, wanting cash for emergencies or heirs, framing, and people underestimating how long they will live.
Do retirees really spend too little?
Many with substantial savings spend slowly: in EBRI’s analysis, those starting with $500,000 or more had spent 11.8% at the median after 20 years. Some of that is deliberate saving for care or bequests.
The short version
- Headline life expectancy figures usually assume today’s death rates never improve. Projections that allow for improvement give a UK 65-year-old man 19.8 more years, against 18.6 on current rates.
- On the US period life table, about one in four 65-year-old men and one in three women reach 90; for a couple of that age, the chance that at least one does is about half.
- In England, people in their 50s and 60s expected to live 6.1 (men) to 7.7 (women) fewer years than life tables suggested.
- The opposite risk is real too: US retirees who started with at least $500,000 had spent only 11.8% of it, at the median, after 20 years.
- Claiming Social Security at 70 instead of 62 raises the monthly benefit by about 77% for life; models suggest most would gain from waiting, but early claiming can be reasonable for people in poor health or with no other savings.
This article summarises published research and official statistics for general information. It is not financial, tax or retirement advice, and it does not recommend any product, provider or strategy. Life expectancy figures are averages and projections, not predictions for any individual. Decisions about pensions, annuities and benefit claiming depend on personal circumstances; consider speaking to a regulated, independent adviser and checking official sources such as your national pension or social security agency.
Further reading: O’Dea and Sturrock, Review of Economics and Statistics (2023), on survival pessimism and annuities. The ONS bulletin on past and projected period and cohort life tables (2025), for how the two measures differ. Banerjee, EBRI Issue Brief (2018), for how slowly retirees spend their savings.
- Die with Zero, Bill Perkins (2020). A case for spending savings on experiences while you can enjoy them. A useful counterweight to underspending; it is written by a wealthy author and says little about the risk of running short.
- Retirement Planning Guidebook, Wade Pfau (2021). A retirement income researcher sets out the risks of retirement, longevity among them, and the main approaches to them. Thorough and evidence-based; it is US-focused and dense in places.
- How to Retire, Christine Benz (2020). Interviews with 20 retirement experts on spending, claiming and living well. A good range of views; as a collection of interviews it offers perspectives rather than a single tested framework.
Sources
- Office for National Statistics. Past and projected period and cohort life tables: 2022-based, UK, 1981 to 2072 (14 February 2025).
- Office for National Statistics. National life tables, life expectancy in England and Wales: 2021 to 2023 (23 October 2024).
- Social Security Administration, Office of the Chief Actuary. Actuarial life table, period 2023.
- Social Security Administration. Retirement planner: life expectancy; early retirement reduction; delayed retirement credits.
- O’Dea C, Sturrock D. Survival pessimism and the demand for annuities. Review of Economics and Statistics (2023), doi:10.1162/rest_a_01048.
- Yaari ME. Uncertain lifetime, life insurance, and the theory of the consumer. Review of Economic Studies 32(2) (1965).
- O’Dea C, Sturrock D. Subjective expectations of survival and economic behaviour. IFS Working Paper W18/14 (2018).
- Sun W, Webb A. How do subjective mortality beliefs affect the value of Social Security and the optimal claiming age? Center for Retirement Research at Boston College working paper (2011).
- Brown JR, Kling JR, Mullainathan S, Wrobel MV. American Economic Review 98(2):304–309 (2008), doi:10.1257/aer.98.2.304.
- Slavov SN. Two decades of Social Security claiming. NBER Working Paper 30843 (2023), doi:10.3386/w30843.
- Altig D, Kotlikoff LJ, Ye VY. Tax Policy and the Economy 37 (2023), doi:10.1086/724355.
- Armour P, Knapp D. The consequences of claiming Social Security benefits at age 62. AARP Public Policy Institute (2021), doi:10.26419/ppi.00134.001.
- Banerjee S. Asset decumulation or asset preservation? What guides retirement spending? EBRI Issue Brief (3 April 2018).
- Bengen WP. Determining withdrawal rates using historical data. Journal of Financial Planning (1994).
- Business Insider via Yahoo Finance, 3 October 2026, on a Prudential Financial survey.
