Retirement risk

Why the first 10 years of retirement matter most: sequence-of-returns risk

Imagine two retirees with the same savings, the same spending, and the same average market return over thirty years. One finishes comfortably; the other runs out of money in her early nineties. The only difference is the order the returns arrived in. That is sequence-of-returns risk, and it's the reason Runway's Historical Backtest replays your plan from every real starting year instead of using one average. (The simple savings ÷ spending runway formula ignores exactly this.)

Quick answer. When you're withdrawing money, bad returns early do far more damage than the same bad returns late, because withdrawals made during a downturn sell more shares at low prices and leave less to recover. In Runway's engine, two historical 31-year windows, starting in 1920 and in 1970, both averaged about 6.3% a year in real terms, but the first ten years averaged 12.73% in one and −1.05% in the other. A household spending $95,000 a year ends with $4,124,948 after the good start and $1,498,510 after the bad one; at $110,000 a year, the good start ends with $2,991,579 and the bad start runs out of money at 87.

What is sequence-of-returns risk?

While you're saving, the order of returns doesn't change your final balance; ten good years then ten bad years multiplies out the same as the reverse. While you're withdrawing, it does change it, because you take money out between the returns. A withdrawal during a downturn locks in the loss, and the shares you sold can't recover. Researchers have known this since the 1990s. In his 1994 study of historical returns, William Bengen found that a 4% first-year withdrawal followed by inflation-adjusted withdrawals: "In no past case has it caused a portfolio to be exhausted before 33 years" (Journal of Financial Planning). The starting year that produced the shortest portfolio life in his data was 1966.

Same average return, different retirement

To isolate the order, we searched Runway's 154 years of real historical returns for two 31-year windows (the length of the sample couple's plan) with the same average return but very different first decades: a 60% stock / 40% bond mix, real returns, dividends reinvested. The best match is a retirement starting in 1920 versus one starting in 1970:

Same average return, different retirement
Start 1920 (good first)Start 1970 (bad first)
Average real return, all 31 years6.27% a year6.25% a year
Average real return, first 10 years12.73% a year−1.05% a year
Ending balance at $95,000 of yearly spending$4,124,948$1,498,510
Ending balance at $110,000 of yearly spending$2,991,579$0, runs out at age 87

Same household (the sample couple, $1,460,000 saved, Social Security at 69 and 68), same average return, a difference of more than $2.6 million at $95,000 of spending, and the difference between a comfortable finish and a shortfall at $110,000. Only the order changed.

How bad can a bad start get?

Across the 124 real 31-year starting years in the data (1872 through 2024), the sample household at $95,000 of spending survives 122 of the 124 (98.4%), and the worst of them started in 1966, the same year Bengen identified, with a first decade averaging −1.73% a year in real terms despite a 4.48% average over the full 31 years. The first ten years are where the plan is most exposed.

You can also test the timing directly. Take one crash (markets down 20% a year for two years) and start it at different points in the same household's retirement ($95,000 of spending; Pat is 65 today):

How bad can a bad start get?
Crash begins at Pat's ageEnding balance at 95Money runs short at
No crash$1,806,248never
65$0age 92
70$0age 94
75$0age 95
80$181,191never

The same crash that leaves $181,191 when it comes at 80 exhausts the money by 92 when it comes at 65 (and by 94 or 95 when it comes at 70 or 75). And with a more stretched plan ($120,000 of yearly spending) that identical two-year crash causes the first shortfall at age 77 if it starts at 65, at 87 if it starts at 75, and at 92 if it starts at 85.

What can I do about it?

For the broader picture of how simulations, crash tests and backtests fit together, read Will my money last?

Replay your plan through 1929 and 1966. Runway's Historical Backtest shows how your plan would have fared from every real starting year.

Try the planner

How were these numbers computed?

Returns are annual, real (inflation-adjusted), dividend-reinvested total returns for U.S. stocks and 10-year government bonds, 1872–2025, from Robert Shiller's public U.S. Stock Markets 1871–Present dataset. The 60/40 blend is the engine's default allocation. The two windows were chosen by searching all 31-year windows for the pair whose average real return differs by less than 0.04 percentage points and whose first-decade returns differ most. Backtest balances are Runway's walk-forward of the sample household with default assumptions, married filing jointly, plan through age 95, with the long-term-care cost left out; the crash table is a deterministic projection of the same household that does include Runway's long-term-care episodes (one for each person, $129,575 a year for three years in each person's last three years). Everything is in today's dollars. The Runway app now shows these same plans in future dollars (each year raised by the inflation assumption), so its figures run larger than the ones here.

Runway content is educational only and is not financial, tax, or legal advice. Consult a qualified professional before making financial decisions.

About the author

Lei Huang is a former professor, turned founder and developer. He builds Runway, a DIY retirement income planner whose planning engine computes the figures in these posts. He is not a financial advisor.