Is 4% still a safe withdrawal rate? A stress test
The 4% rule is quoted constantly and stress-tested rarely. We ran four withdrawal rates across two planning horizons and three levels of volatility to see where it holds and where it quietly stops working.
Why the rule needs testing
The 4% rule came from historical US data over 30-year retirements with a specific mix of stocks and bonds. Three things people do with it break those conditions: they retire earlier than 65, they hold portfolios with different volatility, and they treat the number as a permanent entitlement rather than a starting point. We tested each.
Assumptions used
- Age today / retiring at / plan to
- 65 / 65 / 95
- Starting portfolio
- £1,000,000
- Annual contributions until retirement
- £0
- Retirement spending from the portfolio
- £40,000 a year, today's money
- Guaranteed income
- None assumed
- Expected return / inflation / volatility
- 6% / 2.5% / 13%
- Simulation
- 1,000 runs, returns drawn at random each year
- No guaranteed income is included, deliberately: this isolates the withdrawal rate itself.
- Withdrawals rise with inflation each year, exactly as the rule specifies.
Taxes, platform charges and changes in spending through retirement are not modelled. Figures are illustrations produced by the Retiris model, not predictions.
Test 1 — the rate itself
| Scenario | Runs that lasted | Median years funded | Typical shortfall age | Steady projection |
|---|---|---|---|---|
| 3.0% withdrawal | 88% | 30 | 91 | £1,119,437 left at 95 |
| 3.5% withdrawal | 78% | 30 | 90 | £849,649 left at 95 |
| 4.0% withdrawal | 65% | 30 | 89 | £579,860 left at 95 |
| 5.0% withdrawal | 42% | 28 | 87 | £40,283 left at 95 |
Median balance by starting withdrawal rate (today's money)
Each percentage point costs real ground, and the losses accelerate: 3% to 3.5% costs about ten points of success, 3.5% to 4% costs a similar amount, and 4% to 5% removes more than twenty. That is characteristic of drawdown maths — as the withdrawal rate approaches the real expected return, the portfolio loses its ability to absorb a bad decade. Note too that these runs carry no State Pension or Social Security; adding a guaranteed income lifts every row substantially, which is why the same 4% can look safe in one plan and marginal in another.
Test 2 — horizon and volatility
| Scenario | Runs that lasted | Median years funded | Typical shortfall age | Steady projection |
|---|---|---|---|---|
| 4% over 30 years (to 95) | 65% | 30 | 89 | £579,860 left at 95 |
| 4% over 40 years (to 105) | 47% | 37 | 92 | £315,775 left at 105 |
| 4%, low volatility (8%) | 83% | 30 | 92 | £579,860 left at 95 |
| 4%, high volatility (18%) | 53% | 30 | 86 | £579,860 left at 95 |
4% withdrawals under different conditions
Horizon matters more than most people assume
A 40-year retirement is not 33% harder than a 30-year one. It is disproportionately harder, because the extra decade sits at the end, funded by whatever survived the first three. Anyone retiring in their fifties should treat 4% as an upper bound rather than a target.
Volatility is the hidden variable
Both volatility runs have the same 6% expected return. The high-volatility version performs materially worse, because withdrawals during drawdowns crystallise losses. This is the single most under-discussed point in withdrawal-rate debates: two portfolios with identical expected returns are not equally safe in retirement.
What we conclude
- 30 years, moderate volatility, plus a decent guaranteed income: 4% is a defensible starting point — but only as a starting point.
- A portfolio doing all the work on its own: 4% funded the full 30 years in about two-thirds of our runs. Treat it as ambitious, not safe.
- Over 35–40 years: plan closer to 3.25%–3.5% unless you have strong flexibility.
- High-volatility portfolios: either lower the rate or lower the volatility. Expected return alone does not settle it.
- In every case: a rule you revisit annually beats a rule you set once. Fixed real spending is the assumption doing most of the damage in the failing runs.
Test your own rate
Your withdrawal rate, stress-tested
Everything else stays as set out in the assumptions above.
In this model, under these assumptions. Not advice, and not a forecast.
Sources
Modelled figures on this page were produced with the Retiris retirement model, described in full on how the model works. They are illustrations of hypothetical scenarios, not research findings about real households.
Related analysis
Retiris publishes educational planning tools and analysis. Nothing on this page is personalised financial advice, and modelled scenarios are illustrations rather than predictions. See our editorial methodology and disclaimer.