Key takeaways
- A UK man of 65 and a woman of 60 have a 50% chance that one of them is still alive in 31 years, against 22 years for him alone on the same ONS projections.
- Lengthening a fixed planning period from 20 years to 30 years cut the safe starting withdrawal rate from 4.70% to 3.00% in Blanchett's simulations, a 36.2% cut in starting income.
- The two-lives premium shrinks as the age gap widens: 3 years when both partners are 65, 1 year at a five-year gap, and nothing at all at ten years.
- A single person's Moderate retirement budget is £32,700 a year against £45,400 for a couple, so a survivor needs 72.0% of the joint figure rather than half of it.
- Counting mortality instead of a fixed end date raised the sustainable rate for a 65-year-old couple from 2.6% to 3.2% in the same study, which Blanchett puts at 23.1% higher.
Life expectancy for couples measures the second death, not the first
If you're 65 and your partner is 60, neither of your individual life expectancies is the number the money has to clear. Life expectancy for couples, or joint life expectancy, asks a different question. The portfolio has to last until the second of you has died, and that horizon is longer than either one on its own.
Here's the size of it. On the Office for National Statistics 2024-based cohort projections, published on 15 May 2026, a UK man aged 65 in 2026 has a 50% chance of living another 22 years. A woman aged 60 has a 50% chance of another 30 years. The chance that at least one of them is alive after 31 years is still close to half. That's nine years beyond the man's own median, and it is the number that sets the sustainable withdrawal rate.
Nine years is not a detail. In David Blanchett's 2021 study in the Journal of Financial Planning, stretching a fixed retirement period from 20 years to 30 years cut the safe starting withdrawal rate at a 90% target success rate from 4.70% to 3.00%. On every £100,000 of portfolio, that is £3,000 of first-year income instead of £4,700. Nothing about the portfolio changed. Only the length of time it has to survive.
What the ONS tables say for a man of 65 and a woman of 60
The figures below come from the ONS 2024-based period mortality rates read along the diagonal, which is how the ONS itself describes building a cohort measure: the rates "for age 26 years in 2026, age 27 years in 2027, age 28 years in 2028, and so on." Taking the man's and the woman's survival curves and combining them gives the chance that at least one of the pair is alive at each horizon.
| Horizon from 2026 | Man alive (from age 65) | Woman alive (from age 60) | At least one alive | Both alive |
|---|---|---|---|---|
| 20 years | 54.3% | 78.8% | 90.3% | 42.8% |
| 25 years | 33.6% | 65.7% | 77.2% | 22.1% |
| 30 years | 14.1% | 46.9% | 54.4% | 6.6% |
| 35 years | 2.9% | 24.3% | 26.5% | 0.7% |
Read the fourth column against the second and you can see the problem with a retirement planning horizon built on one life. At 30 years out, the man has a 14.1% chance of being there and the household has a 54.4% chance. Morningstar calls 30 years "the default time horizon used in most research on retirement planning." For this couple that default is a coin toss, not a conservative assumption.
Two things are worth separating here. Cohort rates are the right basis for this, because they allow for mortality improving over the decades the couple is actually living through. The ONS puts it plainly: cohort life expectancy "is seen as a more realistic measure of how long a person can expect to live for because it accounts for future projected improvements in mortality." On the same 2024-based projection, the period figure for a man of 65 in 2026 is 19.06 years against a cohort figure of 20.17, and for a woman of 60 it is 25.82 against 27.53. Reading the period row rather than the diagonal costs him 1.11 years and her 1.71. The National Life Tables most people meet first are shorter again, because they are built on recorded deaths rather than projections: for 2022 to 2024 they give a man of 65 18.73 years and a woman of 60 25.45 years.
The second point is that life expectancy is a mean and the horizon you plan on is a percentile. The cohort mean for the man is 20.17 years. His median is 22. For half of this distribution, the mean is the wrong side of the answer.
A wider age gap lengthens the horizon and shrinks the two-lives premium
The intuition most people start with is that an age gap and joint survival compound each other. They don't. The age gap does lengthen the absolute horizon, but the extra years you get purely from having two lives rather than one get smaller as the gap widens.
| Couple in 2026 | Man's median | Woman's median | At least one alive, median | Two-lives premium |
|---|---|---|---|---|
| Both 60 | 26 years | 30 years | 33 years | 3 years |
| Both 65 | 22 years | 25 years | 28 years | 3 years |
| Man 65, woman 60 | 22 years | 30 years | 31 years | 1 year |
| Man 70, woman 60 | 17 years | 30 years | 30 years | 0 years |
| Man 65, woman 55 | 22 years | 35 years | 35 years | 0 years |
The last column is the finding. When both partners are the same age, the pair of them buys about 3 years of horizon over the longer-lived one alone. At a five-year gap it buys 1 year. At ten years it buys nothing you could measure in whole years, because by the time the younger partner reaches her own median the older one is very unlikely to be there. For a couple with a wide age gap, life expectancy for couples collapses into the younger partner's own horizon.
That has a practical edge to it. If the gap is large, the planning horizon is already close to one person's number, and getting that one person's mortality right matters more than the joint arithmetic. If the ages are close, the joint effect is doing real work and a single-life life expectancy table understates the horizon by about three years.
Wide gaps are also not the common case. In 2023 the median age for men entering an opposite-sex marriage in England and Wales was 34.8 years, against 33.0 years for women, a difference of 1.8 years between the two medians. That isn't the same statistic as the gap inside any given couple, which the ONS bulletin doesn't report. But read the table above again with that in mind: for most couples the work is being done by the gap between male and female mortality, not by an unusual difference in age.
Each extra decade of horizon takes about a percentage point off the rate
Blanchett's Monte Carlo analysis ran a 40% equity and 60% fixed income portfolio through 10,000 simulations, with equities at an assumed 8% return and 20% standard deviation, fixed income at 4% and 7%, a correlation of 0.1, a 1% annual fee and 2% inflation. The resulting safe initial withdrawal rates, with the initial amount rising annually with inflation, look like this.
| Fixed retirement period | Rate at 90% success | Rate at 95% success |
|---|---|---|
| 20 years | 4.70% | 3.80% |
| 25 years | 3.60% | 2.90% |
| 30 years | 3.00% | 2.50% |
| 35 years | 2.60% | 2.10% |
| 40 years | 2.30% | 1.90% |
Lay the horizons from the previous section over this table. A man of 65 planning alone to his own median of 22 years sits near the 20-year row and 4.70%. The same man with a partner of 60, planning to the couple's median of 31 years, sits near the 30-year row and 3.00%. That's a 36.2% cut in starting income for the same pot.
Plan to the point where the chance one of them is still alive has fallen to a quarter, which is 36 years for this pair, and the 35-year row gives 2.60% at a 90% target. Against the single man's 4.70%, that's 44.7% less to spend in year one, or £2,600 a year for every £100,000 instead of £4,700.
Morningstar's forward-looking work gets to a similar shape from different inputs. Using capital markets assumptions as of 30 September 2025 and 1,000 simulated return paths on a 40% equity portfolio at a 90% probability of success, it estimated 9.7% for a 10-year horizon, 6.7% for 15 years, 5.3% for 20 years, 4.4% for 25 years and 3.9% for 30 years. The slope is the same. The level is not, and that gap is discussed below.
One structural point falls out of comparing the two. Morningstar's published range stops at 30 years, which is where this couple's median sits rather than where its tail does. Blanchett's table runs to 50, and it is the more useful document for a joint plan for exactly that reason. The horizons a couple needs are at or past the edge of most published work, including our own read of withdrawal rate success rates from the historical record.
The strongest objection: a fixed end date overfunds the plan
There is a serious case that everything above is too pessimistic, and it comes from the same paper.
Picking a long horizon and then demanding a high success rate over it counts the same caution twice. You're funding years you probably won't see, and then insisting on a 90% chance of funding them. Blanchett modelled the alternative, where failure means the portfolio runs dry while someone in the household is still alive, rather than before an arbitrary end date.
The difference is large. For a 65-year-old couple, the 20% survival probability period is 34 years. A planner rounding to 35 years at a 90% target success rate lands on 2.6%. Blanchett's finding is that "a 2.6 percent initial withdrawal rate has a 95 percent probability of success when incorporating mortality rates." If the couple actually wanted 90% once mortality is counted, the rate "should be 3.2 percent, which is 23.1 percent higher."
So the joint horizon raises the question and mortality-weighting softens the answer. Both are true at once. The age gap genuinely lengthens the period the money must cover, and planning to the far tail of that period as though survival were certain costs roughly 0.6 percentage points of income you didn't need to give up. Blanchett's own shortcut for planners who can only model fixed periods is to add five years to a single person's life expectancy and eight years to the longer of a couple's, the difference between the two reflecting the tail risk a two-person household carries.
There is also a reason to think planners are nowhere near this careful in either direction. Across 31,211 Canadian financial plans written between 2009 and 2020, roughly 70% of the retirement end ages used were 90 and roughly 20% were 95. Age 90 is 25 years for a 65-year-old, which is short of this couple's median by six years.
When one partner dies the budget falls to about 72%, not 50%
How long will our money last is only half of the arithmetic. The other half is what the survivor spends, and it isn't half of what the couple spent.
Pensions UK's Retirement Living Standards, updated in June 2026, cost a Moderate standard of living at £32,700 a year for one person and £45,400 for two. The single figure is 72.0% of the joint one. At the Minimum standard it's £13,900 against £22,500, or 61.8%. At Comfortable, £45,400 against £62,700, or 72.4%. Housing costs sit outside all of these, and they assume the home is owned outright.
Rent, council tax, standing charges, the car and the broadband do not halve when a household goes from two people to one. The standards say so directly: living in a two-person household "can reduce costs by sharing expenses, but this won't apply to everyone." So a plan that assumes spending halves on the first death understates the survivor's need at all three standards, on exactly the stretch of the horizon that the joint arithmetic made longer. The two errors point the same way.
What UK drawdown plans are actually withdrawing
Against rates of 2.6% to 3.9%, here is what people are doing. In 2024/25, of 578,491 UK pension plans making regular partial withdrawals, 259,507 withdrew at an annual rate of 8% or more. That is 44.9% of them. Only 156,463, or 27.0%, withdrew less than 4%. Among plan holders aged 65 to 74, 117,096 of 283,124 plans, or 41.4%, were running at 8% or above.
Those numbers need reading carefully before anyone draws a conclusion from them. The FCA counts plans, not people, and many holders have several. Pot size explains a great deal: of plans holding less than £10,000, 32,037 of 38,354 were running at 8% or more, which is 83.5%. Of plans holding £250,000 or above, 20,490 of 144,757 were, which is 14.2%. Emptying a small pot over a few years is a reasonable thing to do and not a failed retirement plan. The 14.2% is the figure closest to a household living off one substantial pot, and it is still well above anything the joint horizon supports. The shape of the risk is set out in our piece on running out of money in retirement, where the failure is usually a few years short rather than destitution.
What these numbers cannot tell you
Start with the joint survival calculation behind every life expectancy for couples figure here. It treats the two deaths as independent events, and they aren't. A couple shares a house, a diet, an income, a postcode and a care burden, and bereavement raises mortality in the period that follows it. Independence almost certainly overstates the joint horizon somewhat. How much is not something these tables can say, because the ONS publishes population mortality by age and sex, not by household.
The published rates also stop at age 100, so survival past 101 is treated as zero here. That only bites in the deep tail. The 31-year median for this couple is reached while both are under 100, so it is untouched; the quarter-point at 36 years loses a little, and the effect is to shorten the horizon rather than lengthen it. And the ONS figures are projections, built on assumed future improvements in mortality. The 2024-based set came out close to its predecessor, with "differences of less than 0.1 years for cohort life expectancy at birth in 2049," but a projection that has been revised before can be revised again.
Then there is the withdrawal-rate side, where the uncertainty is larger than anything mortality contributes. Blanchett's 30-year, 90% figure is 3.00%. Morningstar's, on the same horizon and the same success target and a similar 40% equity allocation, is 3.9%. That is a 0.9 percentage point spread between two credible Monte Carlo studies, and it is bigger than the effect of moving the horizon from 25 years to 30. Part of it is the 1% annual fee Blanchett charges the portfolio; the Morningstar article never mentions one. Part of it is different return assumptions. The honest reading is that the horizon changes the answer by less than the choice of model does.
Both studies are also simulations, not history. They sample from assumed distributions rather than from what markets actually did. Blanchett's returns "were assumed to follow a normal distribution", which is an assumption about how often very bad years happen rather than a measurement of it. The historical record gives different answers again, as our look at the safe withdrawal rate outside the United States found. Neither study is UK-specific: the mortality is British here, the market assumptions are not.
What would change the conclusion
If mortality improves faster than the principal projection assumes, every horizon in this piece lengthens and the rates attached to them fall. The projection already builds in improvement. A faster path would push this couple's median past 31 years and the quarter-point past 36, and the 2.60% row would stop being a cautious reading.
If the plan has a guaranteed income floor underneath it, most of this stops mattering. A joint-life annuity or a defined benefit pension with a survivor's pension transfers the longevity tail to someone else, and what the portfolio has to cover is the gap above the floor rather than the whole budget. The joint horizon is a problem specific to money you are drawing down yourself.
And if the spending plan is flexible rather than fixed in real terms, the horizon loses much of its force. Every rate quoted here funds an inflation-linked amount for a fixed number of years, which is the assumption that makes a long horizon expensive. A rule that cuts spending after a bad run breaks the link, which is what our work on a fixed percentage withdrawal measures.
The number to keep an eye on isn't your life expectancy. It's the gap between it and your partner's, and whether the plan on the table was built around one life or two. LedgerTouch shows the drawdown against the horizon you set; the horizon itself is the assumption worth arguing about.