Sequence of returns risk, explained plainly

Two retirees start with identical £1,000,000 portfolios. Both withdraw £50,000 a year, adjusted for inflation. Both earn the same average return over thirty years, 7% a year. One runs out of money at year 19. The other finishes with nearly £2.8 million.

The difference is not skill, luck in the conventional sense, or market outperformance. It is the order in which the returns arrived. This is sequence of returns risk, and it is the single most-ignored variable in early retirement planning.

Why averages lie

The 4% rule, Bengen's 1994 research, and almost every FIRE spreadsheet on the internet work by applying an average return assumption to a portfolio over time. If the average real return is 5%, and you withdraw 4%, you grow by 1% per year. Simple.

This works perfectly if returns actually arrive at their average. They never do. Real market returns cluster, oscillate, and occasionally deliver decade-long periods that bear no resemblance to the long-run average. The US market returned roughly 10% nominal a year from 1900 to 2025, but within that, there were stretches of 0% real returns lasting 15 years (1966 to 1981), stretches of 20% real returns lasting a decade (the 1990s), and crashes of 50% in under two years (2000 to 2002, 2007 to 2009).

When you are accumulating money, the order doesn't matter. A 40% crash in year 1 versus year 20 produces the same ending balance, because you are adding money during the down periods and buying cheaper shares. Dollar-cost averaging works in your favour.

When you are withdrawing money, the order matters enormously. A 40% crash in year 1 of retirement forces you to sell a larger percentage of your portfolio to fund the same withdrawal. You are now depleting your capital base at the worst possible moment, leaving less to participate in the recovery. A 40% crash in year 20 lands on a portfolio that has already weathered most of its withdrawals, and the compounding damage is far smaller.

This asymmetry is what sequence of returns risk actually means: the identical average return can produce radically different outcomes depending on when the good and bad years fall.

A concrete illustration

Consider two 30-year retirement paths, both starting at £1,000,000 and both withdrawing £40,000 in year one (rising 2% a year with inflation), both averaging 7% nominal returns across all 30 years.

Path A (bad sequence): returns of -15%, -12%, +5%, -8%, then steady 9% for 26 years. This produces a 7% average.

Path B (good sequence): 9% for 26 years, then -8%, +5%, -12%, -15%. Same returns, reversed.

Path A runs out of money around year 23. Path B finishes with £2.2M remaining.

The numbers used here are illustrative — the specific sequence matters less than the principle. What matters is that the first decade of retirement effectively sets the ceiling on how much damage can be done to the plan. A portfolio that survives its first decade intact is extremely likely to last another twenty. A portfolio that loses 25-30% in its first three years, while also funding withdrawals, is very likely to fail regardless of what happens afterwards.

Why this is worse for early retirees

A traditional retiree at 65 has a 20-year expected horizon. An early retiree at 45 has a 45-year horizon, and if they're lucky, longer.

Longer horizons sound like they should help — more time for markets to recover, more time for compounding. But longer horizons also mean more potential sequences you could retire into. And crucially, early retirees are usually fully invested when they stop working, because their accumulation strategy required aggressive equity exposure to hit the target.

This creates the precise conditions for sequence risk to bite. Maximum equity allocation, maximum withdrawal duration, no new capital coming in to buy dips. The worst historical sequences for a 45-year-old retiree — retiring into 1906, 1929, 1966, or 2000 — were not characterised by a single crash. They were characterised by a long first decade of poor real returns. The 1966 retiree experienced 0% real returns for 15 years while inflation consumed purchasing power from the withdrawal stream.

Bengen's 4% rule (actually 4.15% SAFEMAX in the original research) is specifically the rate at which the worst historical US sequence produced a portfolio that survived 30 years. Not the average sequence. The worst one. This is the entire point of the research, and it's the part that FIRE spreadsheets using 5% or 6% withdrawal rates are quietly ignoring.

For 40+ year horizons, Morningstar's 2025 State of Retirement Income report puts the base-case safe withdrawal rate at 3.3%, not 4%. Pfau's international research shows that only five countries — Sweden, Canada, New Zealand, Denmark, and the US — have historical sequences that supported 4% over 30 years. The UK's SAFEMAX was 3.05% for a 50/50 portfolio. Italy's 4% rule failed in 80.5% of historical periods.

These numbers are all downstream of sequence risk. They are the mathematical expression of "what withdrawal rate survives the worst thing that has ever happened to this market."

What actually protects against it

The plain reading of sequence risk is depressing, because most of what it reflects is outside your control. You cannot choose when to retire. You cannot choose the market conditions of your first decade. You can only choose how to structure your plan so that a bad sequence doesn't end it.

Four strategies meaningfully help.

Lower starting withdrawal rate. The simplest and most effective. A 3.5% initial withdrawal rate survives worse sequences than 4%. A 3% initial withdrawal rate survives almost any historical sequence, but requires a 33% larger portfolio to produce the same income. This is the trade-off at the heart of honest FIRE planning.

Variable withdrawal rules. Guyton-Klinger guardrails, Vanguard's dynamic spending, or simple floor-and-ceiling rules (withdraw 4% of current portfolio with a 10% floor and 20% ceiling on annual changes) reduce withdrawals after bad years and increase them after good ones. This cushions sequence risk directly by refusing to compound capital losses with matching withdrawals. The trade-off is income variability, which is psychologically harder than it sounds.

Cash buffer. Holding 2 to 3 years of expenses in cash or short-duration bonds, drawn from during equity drawdowns instead of selling equities at a loss. This is not a panacea — the buffer itself drags returns during normal periods — but it provides the specific ammunition to avoid the worst compounding behaviour of a bad first decade.

Flexibility to earn. The single largest reducer of sequence risk is not a portfolio adjustment. It is willingness to earn some income in the first five to ten years of retirement if the market is poor. £10,000 of part-time income during a bad sequence is worth far more than £10,000 five years later, because it directly reduces the withdrawals being made at the worst moment. This is why "Barista FIRE" and "Coast FIRE" variations exist, and it's why the purist "never work again" version of FIRE is the most fragile.

The stress test most people skip

Retirement planning conventionally uses Monte Carlo simulation to estimate "probability of success" — the percentage of simulated sequences that survive the plan. A 90% success rate is typically treated as acceptable.

This framing has a problem. It tells you that 10% of sequences fail, but not what those failures look like or when they happen. A plan that fails in year 28 of a 30-year retirement is very different from a plan that fails in year 12. The first is survivable with modest adjustments. The second is catastrophic, because by the time you notice you're in a bad sequence, you're already too deep to recover.

A better framing is to look at the worst-case portfolio value in the first decade across your Monte Carlo runs. If the 5th percentile shows your portfolio down 40% by year 7, you have a bridge-problem-shaped version of sequence risk: not "will I run out" but "can I psychologically survive the drawdown without abandoning the plan." Most people cannot. They sell at the bottom, crystallise losses, and guarantee the worst outcome their plan was designed to avoid.

An honest FIRE plan looks at both. Survival probability is necessary but insufficient. The shape of the failures matters.

What this means for your plan

The practical takeaways are uncomfortable but not complicated.

Assume your starting withdrawal rate will be tested by a bad sequence. Plan as if you are retiring into 1966. If your plan survives that, it will likely survive anything.

Treat the first decade as load-bearing. Every pound of withdrawal during a bad first decade does disproportionate damage. Every pound of earned income during a bad first decade does disproportionate good. The accumulation phase does not end on your last working day; it ends roughly a decade later, once the sequence has revealed itself.

Do not confuse long-run averages with your experience. Your experience is one sequence, not the average of all sequences. You will not get the textbook 7% real return evenly distributed over 40 years. You will get something jagged, and the shape of the jagged will matter more than the average.

Build flexibility into the plan before you need it. It is far easier to reduce spending by £5,000 a year in year one when it's a choice, than to do it in year five when it's a necessity. Variable withdrawal rules force this choice in advance, which is a feature, not a bug.

Where Endute fits

Most FIRE calculators show a single projected path with a smooth return assumption. This produces reassuring charts and misleading confidence. Endute's FIRE simulator runs thousands of historical and Monte Carlo sequences and shows you the distribution, including the 5th and 10th percentile outcomes — the ones that matter for sequence risk. It also lets you model variable withdrawal rules against the same sequence set, so you can see what a guardrail rule actually does to the worst cases rather than just the average.

None of this removes sequence risk. Nothing removes it. But modelling it honestly is the first step to building a plan that can survive it.

Sources and further reading

• Bengen, W.P. (1994). "Determining Withdrawal Rates Using Historical Data." Journal of Financial Planning 7(4):171-180. The original SAFEMAX research, 4.15% for US historical data.
• Cooley, Hubbard, Walz (1998). "Retirement Savings: Choosing a Withdrawal Rate That is Sustainable." AAII Journal. The Trinity Study, extending Bengen's work.
• Pfau, W.D. (2010). "An International Perspective on Safe Withdrawal Rates: The Demise of the 4 Percent Rule?" Journal of Financial Planning, December 2010. SAFEMAX across 17 developed markets, 1900-2008.
• Morningstar (2025). The State of Retirement Income 2025. 3.3% base case for 40-year horizons.
• Guyton, J. and Klinger, W. (2006). "Decision Rules and Maximum Initial Withdrawal Rates." Journal of Financial Planning. The foundational work on variable withdrawal guardrails.

This is the fourth post in a cluster on early retirement planning. The preceding posts cover the total portfolio target, the tax-advantaged wrappers used to build it, and the bridge problem of accessing the pot before pension age. The next post in the series examines why the 4% rule performs poorly outside the US.