Start with the arithmetic, because it's the whole trick. If you withdraw a fixed real percentage of a portfolio each year, the capital required is annual spending divided by that percentage. Divide by 4% and you get 25 times spending. Divide by 3.5% and you get 28.6 times. Divide by 3% and you get 33.3 times.
That's the entire relationship. The multiple isn't a separate rule sitting beside the withdrawal rate. It's the withdrawal rate turned upside down.
So every dispute about safe withdrawal rates is also a dispute about how many years of spending someone is meant to accumulate. And because a multiple is a reciprocal, it doesn't move in a straight line when the rate moves. That curve is where most of the confusion lives.
Small changes in the rate, large changes in the capital
Cut the withdrawal rate by 0.1 percentage points and the multiple rises by roughly one divided by the rate squared. At 4%, that's an extra 0.64 years of spending. At 3%, the same 0.1-point cut adds 1.15 years. At 2.5% it adds 1.67 years.
Put money against it. On spending of £32,700 a year, moving from 4.0% to 3.9% adds about £21,000 to the target. The same 0.1-point step from 3.0% to 2.9% adds about £37,600. Identical change in the rate, nearly double the change in capital.
That's why the argument feels so charged at the pessimistic end. Debating 4% against 3.9% is debating a rounding error. Debating 3% against 2.75% is debating roughly three years of spending. The cross-country work on the 4% rule tested across 17 developed markets lands at the pessimistic end, which is exactly where each fraction of a point costs the most.
One research house moved its own number by 0.7 points
Morningstar publishes an annual estimate of the highest safe starting withdrawal rate. Its report dated 3 December 2025 puts the base case at 3.9% over a 30-year horizon, for portfolios holding between 30% and 50% in equities, at a 90% probability of success.
The same report lists its own back catalogue: 3.3% in 2021, 3.8% in 2022, 4.0% in 2023 and 3.7% in 2024. Turn those into multiples and the range runs from 25.0 times spending to 30.3 times. On £32,700 of annual spending that's a swing of about £173,000, from one firm using one method across five consecutive years.
Nothing about the hypothetical retiree changed across those estimates. What changed were the capital-market assumptions feeding the model. Morningstar runs 1,000 simulated return paths per asset mix and defines a 90% success rate as at least 900 trials ending with a positive balance. The report is blunt about how thin that bar is: if at least a penny remains at the end of year 30, the trial counts as a success.
Two caveats travel with the 3.9%. The model excludes fees and taxes, and the report says so plainly — both would push the sustainable rate down and the multiple up. It also assumes flat inflation-adjusted spending for 30 years, the convention Morningstar attributes to William Bengen's original work on withdrawal rates. Fees are not a rounding error either, as the gap between 0.2% and 1% compounded over 30 years shows.
Horizon does more work than asset allocation
The same research shows what the horizon costs. A 40-year retirement supports a base-case starting rate of just 3.3%, which is 30.3 times spending. A 15-year horizon supports nearly 7%, or roughly 14 times.
Retiring five years early doesn't simply add five years of withdrawals to the target. It also lifts the multiple. The gap between 25.6 times at 30 years and 30.3 times at 40 years is 4.7 years of spending, applied before any of the extra withdrawals are counted.
How long is a UK retirement, then? ONS national life tables for 2022 to 2024 give life expectancy at age 65 of 18.7 years for men and 21.2 years for women. Those are period figures. The ONS states that they reflect current mortality rates at each age and do not give a true estimate of how long someone can expect to live, because future mortality will improve or worsen. Cohort measures, which allow for improvement, run higher. And a horizon set at average life expectancy runs out early for about half of people by construction.
Guaranteed income comes off the spending, not off the multiple
This is the part the multiple framing hides, and it changes the answer more than any argument about 3.5% versus 4%. The multiple applies only to the slice of spending the portfolio has to fund. Guaranteed income is subtracted first.
Required capital equals annual spending minus guaranteed income, divided by the withdrawal rate. Nothing else.
The full new State Pension for 2026/27 is £241.30 a week, up from £230.25 the year before, according to the DWP rates table. That's £12,547.60 a year on a 52-week basis. The old basic State Pension rose to £184.90 a week over the same uprating.
At a 3.5% withdrawal rate, £12,547.60 of inflation-linked lifetime income is worth £358,500 of capital. At 3.9% it's worth £321,700. A second full entitlement in a couple doubles it. That figure isn't a projection or a forecast — it's the same reciprocal, run backwards.
What that does to the UK targets
The Retirement Living Standards, produced by the PLSA with Loughborough University's Centre for Research in Social Policy, put annual expenditure for a one-person household at £13,900 at the minimum level, £32,700 at moderate and £45,400 at comfortable. Two-person households are £22,500, £45,400 and £62,700. The site states the latest research was carried out during 2025 with modelling completed in 2026.
Three caveats matter before the arithmetic. The figures assume the home is owned outright, so rent and mortgage payments are excluded. They're expenditure, not gross income, so income tax sits on top. And the State Pension itself is taxable, which eats into the personal allowance before any drawdown starts.
| Household and standard | Annual spending | State Pension | Portfolio must fund | Capital at 3.5% |
|---|---|---|---|---|
| One person, minimum | £13,900 | £12,548 | £1,352 | £38,600 |
| One person, moderate | £32,700 | £12,548 | £20,152 | £575,800 |
| One person, comfortable | £45,400 | £12,548 | £32,852 | £938,600 |
| Two people, moderate | £45,400 | £25,095 | £20,305 | £580,100 |
| Two people, comfortable | £62,700 | £25,095 | £37,605 | £1,074,400 |
Look at the moderate single household. Ignore the State Pension and the target is £934,300. Include it and the target is £575,800. The same spending, the same rate, and a £358,500 difference produced entirely by subtracting a guaranteed income before dividing.
The minimum standard is starker. A one-person household needs £38,600 of capital once the full new State Pension is counted, against £397,100 without it. Two full entitlements at £25,095 already exceed the £22,500 two-person minimum, with £2,595 to spare. Whether an individual has a full entitlement depends on National Insurance record, so these are the ceiling cases, not the typical ones.
How much the State Pension is actually doing
DWP's Pensioners' Incomes statistics for the financial year ending 2025 show benefit income, State Pension included, made up 58% of gross income for single pensioners and 40% for pensioner couples. Occupational pensions supplied 24% and 29%. The series is accredited official statistics, but it comes from the Family Resources Survey with around 6,300 pensioner units, and the response rate was 31% in FYE 2025. DWP itself warns that single-year comparisons are rarely statistically significant.
The Institute for Fiscal Studies put a figure on what a more generous flat-rate state pension removes from the saving problem. Its September 2024 report on retirement income adequacy was funded by the abrdn Financial Fairness Trust and the ESRC. It takes a stylised worker earning close to full-time median pay of £38,500, targeting a 67% gross replacement rate worth about £25,800 of retirement income. Post-2004 changes to the state pension cut the private saving required from around 9% of earnings to around 6% — roughly a third — assuming earnings indexation. Under the triple lock, the IFS says the required rate falls further.
The same report shows how fragile any of these projections are. It finds 57% of private-sector employees saving into a defined contribution pension are on track for an adequate replacement rate, and 68% projected to clear a minimum living standard of £14,400 a year. Lower the assumed investment return by one percentage point and over half are undersaving. Raise it by a point, add a 10% higher state pension and count expected inheritances, and undersaving falls to just under two in ten.
The UK modelling convention lands on 3.5%
The Pensions Policy Institute's June 2025 technical report, produced for the PLSA using round 8 of the Wealth and Assets Survey, models income from capital at an initial 3.5% of starting capital, rising with inflation thereafter. It estimates the chance of exhausting that capital before death at approximately 5%. The survey round covered about 15,100 households and more than 25,000 individuals aged 25 to 64 across Great Britain. The PLSA sponsored the work, which the report discloses on its cover.
3.5% is 28.6 times spending. Morningstar's 3.9% is 25.6 times. Two credible institutions, the same arithmetic, three years of spending apart. Neither is wrong; they're answering slightly different questions with different assumptions about returns, horizons and what failure means.
Flexibility moves the number further than either. Morningstar found a flexible approach supported up to 5.7% of the starting balance, which is 17.5 times spending. A 30-year US inflation-linked bond ladder supported 4.5% as at 30 September 2025, or 22.2 times, though such a ladder self-liquidates and leaves nothing behind. The price of that flexibility is variable income, which is the trade-off examined in the piece on withdrawal guardrails and the cost of a 5.2% start.
The case against a single multiple
The strongest objection is that spending isn't flat through retirement, so applying one multiple to one spending figure is a fiction. Critics of the 25x convention argue that real spending falls as people age, and that flat-real modelling therefore over-funds retirement by a wide margin.
The evidence behind that objection is real and quantified. David Blanchett's 2014 paper for the Society of Actuaries' Living to 100 symposium built actual spending curves from the RAND Health and Retirement Study and US Consumer Expenditure Survey data, rather than assuming flat real withdrawals. Modelling the cost over a household's expected lifetime while using those curves produced a required balance up to 25% lower than traditional models.
The effect on success rates is large. In Blanchett's simulations, a 4.0% initial withdrawal rate had a 73.3% probability of surviving 30 years under a constant-real strategy. Using the spending curve for a household starting at $100,000 of annual spending, the same 4.0% start had a 91.1% chance.
Here's why that doesn't kill the multiple. A multiple is a conversion between a yearly flow and a stock of capital. If the flow declines, the right multiple is smaller — Blanchett's 25% reduction is itself a multiple, closer to 19 times than 25. The arithmetic survives intact. What changes is the number fed into it.
There's also a catch inside the decline, and Blanchett flags it directly. The direction isn't uniform. Households with lower consumption and higher funding ratios tended to raise real spending through retirement. Households with higher consumption and lower funding ratios cut it significantly. Some of the observed decline is people adapting to what they have rather than choosing to spend less. Planning on a fall that partly reflects running short is circular reasoning.
And the data is American. RAND HRS and the Consumer Expenditure Survey describe US households, US medical costs and US benefit structures. Blanchett also notes that the Consumer Expenditure Survey is cross-sectional, which is why he turned to a panel dataset for the changes over time. Whether the same curve holds for a UK retiree is a separate empirical question, and none of the sources read here answer it.
Order of operations beats precision
If the multiple is sensitive to the rate, and the rate is contested, the reciprocal is a weak place to spend effort. The bigger lever is what gets subtracted before the division.
For the moderate one-person standard, the State Pension removes 38% of the capital requirement. For the two-person minimum standard, it removes all of it. No plausible argument about 3.5% versus 3.9% moves the answer that far. Adding a defined benefit entitlement of even £6,000 a year removes another £171,400 at 3.5%.
The multiple also says nothing about the order of returns. Two portfolios with identical average returns and identical starting multiples can end in opposite places, which is the point made in the analysis of sequence risk and why identical returns diverge. A multiple is a starting condition, not a guarantee.
What would change the conclusion
Four things would move these numbers materially.
First, long real yields. Morningstar's 3.3% estimate in 2021 came in a low-yield world, and its 4.0% in 2023 came after yields rose. A sustained fall in real yields would push the multiple back toward 30 times without anything about the retiree changing.
Second, state pension policy. Every figure in the table above assumes a full new State Pension of £241.30 a week is received and continues to be uprated. Weaken the triple lock, means-test it, or push the state pension age out, and the subtraction shrinks. The IFS analysis is explicit that its required saving rates depend on the indexation assumption.
Third, UK evidence on retirement spending. If a British dataset reproduced the US spending curve, the honest multiple would drop toward Blanchett's 19 times. If it showed real spending holding flat or rising with care costs, the 28.6 times convention would look about right, or generous.
Fourth, taxes and fees. Morningstar's model excludes both. A pot drawn down through the income tax system, minus platform and fund charges, supports less spending than the headline rate implies. Every one of those figures is a gross number.
What wouldn't change is the shape of the relationship. The multiple is the reciprocal of the rate, it steepens as the rate falls, and guaranteed income is subtracted before the division rather than after. Those three facts do more work than any single estimate of what's safe.