Once you are drawing on your savings, the order of returns can matter more than their average.

Sequence of returns risk is the danger that poor returns arrive early, just as you begin living on your savings. A saver who meets a crash buys more cheaply and has years to recover. Somebody drawing down must sell holdings at low prices to pay for the year, and those holdings are not there when prices come back. Michael Kitces showed the effect in US history: an American who stopped work in 1966 earned 9.56% a year on a balanced portfolio over the next 30 years, yet could only sustain a starting withdrawal of barely over 4% — because the bad years came first¹.

Same returns, different order

With nothing going in or out, order makes no difference: multiplying the same yearly returns in any order gives the same result. Add a withdrawal every year and that stops being true. The table below takes one illustrative set of ten yearly returns — two bad years, then recovery — and runs it forwards and backwards.

With no withdrawals the two orders finish level. With the same withdrawal every year, the household that met the losses first ends roughly 261,000 behind — over a quarter of what it started with — despite earning exactly the same returns. Early losses shrink the base that every later good year compounds on, and each withdrawal taken from a fallen portfolio sells more of it.

The danger window is the first five to ten years

Morningstar’s 2025 retirement income research looked at the simulated cases in which an all-share portfolio ran out within 30 years, and found that nearly 70% of them had already lost value by the end of year five². Turned the other way round, portfolios that came through the first five years with a gain had only about a 4% chance of running out later, and even one year of gains at the start cut the risk of failure in half².

Bengen saw the same thing in history in 1994. The 1973–74 slump, which combined falling markets with high inflation, shortened the life of portfolios whose withdrawals had begun as much as 20 or more years earlier³. His warning to recent retirees was not to raise withdrawals just because the first years went well: early gains may be needed later³.

What history shows: 1966, 1982 and Japan

• 1966. The worst year in the last century to start drawing on a US balanced portfolio. After ten years its compound return was negative after inflation, and the portfolio had gone backwards before withdrawals were counted⁴. Over 30 years it still averaged 9.56% a year — the order, not the average, set the safe withdrawal rate¹.
• 1912. The mirror image: a 30-year return of only 5.50% a year, yet the sequence allowed a higher starting withdrawal of 4.6%¹.
• 1982. A retiree who started at the beginning of a long bull market could have raised spending every three years throughout the following decades under a rule that allowed raises only when the portfolio grew well beyond its starting value⁵.
• Japan, 1990. The Nikkei 225 peaked at 38,915.87 on 29 December 1989 and did not close above that level again until February 2024 — more than 34 years later⁶. In Pfau’s 17-country study, Japan also produced the single worst 30-year start: a retiree in 1940 could have sustained only 0.47% a year⁷.

Bengen’s paper also carries a cautionary tale about the lucky ones. A client who stopped in 1958 enjoyed a decade of strong markets, raised her withdrawals as her balance passed double its starting value, and then met the 1973–74 slump — by which point she was drawing 8% a year from a portfolio that had lost more than half its purchasing power³. A good sequence is only good if it is not spent as though it will last.

What reduces sequence risk

• A modest starting withdrawal rate. The lower the first-year draw, the more a bad start can be absorbed. Morningstar’s 2025 estimate for 30 years is 3.9%, falling for longer horizons (the 4% rule by country)².
• A runway of steady money. Holding the next few years of spending in cash and bonds means nothing has to be sold in a downturn. Morningstar found bond holdings acted as a shock absorber, with portfolios holding 50% to 70% in bonds supporting the highest starting rates in its study². This is the logic of the bucket strategy.
• Spending rules decided in advance. Jonathan Guyton and William Klinger’s “guardrails” cut spending after markets fall and raise it after they rise. With those rules, they found initial withdrawal rates of 5.2% to 5.6% sustainable at a 99% confidence level over at least 40 years for portfolios holding at least 65% in shares⁸. In Morningstar’s version, spending is cut by 10% when the withdrawal rate climbs 20% above where it started².
• Some income in the early years. Part-time work, a partner still earning, rent or a pension that starts early all reduce what has to be sold while prices are low.
• Flexibility about the date. Stopping a year or two later after a crash, while markets recover, can be worth more than any portfolio change.

Testing your own plan against a bad start

No one can predict when a fall will come. What you can know is what one would do to you. Nivritee’s stress tests place market falls after you stop working, when there is no salary to ride them out: A dot-com style slump applies a 25% fall one year after you stop, and A 2008-style crash a 35% fall five years after — each milder than the episode it is named after¹⁰. Tick several and they are drawn as one combined line against your plan, with what changes listed beneath.