pillar · fundamentals · withdrawal

Sequence-of-returns risk: the math that breaks retirement plans

Two retirees with identical 30-year average returns can have wildly different outcomes if the order of returns is different. Here's why, and what to do about it.

By The Yearfold team · April 24, 2026 · Last reviewed May 4, 2026

If you remember one piece of math about retirement, make it this:

The order in which you receive returns matters more than the average return.

This is sequence-of-returns risk, and it's the single most under-appreciated concept in retirement planning. The math is short and decisive — once you see it, you can't unsee it.

The thought experiment

Two retirees, A and B. Both retire on the same day with $1,000,000. Both withdraw $50,000/year (5% withdrawal). Both experience the same set of annual returns over 5 years — just in opposite order.

Year	Retiree A return	Retiree B return
1	-20%	+25%
2	-10%	+15%
3	+5%	+5%
4	+15%	-10%
5	+25%	-20%

Both have the same arithmetic average return: +3% per year over 5 years. Both withdraw the same dollar amounts.

But after 5 years:

Retiree A has approximately $651,000 remaining.
Retiree B has approximately $867,000 remaining.

Same average return. Same withdrawals. $216,000 difference in ending balance — over five years. Over 30 years, the gap can be the difference between a successful plan and complete depletion.

Withdrawals during a market drop compound the loss. When you withdraw $50K from an $800K portfolio (after a 20% loss) you've taken 6.25% of remaining assets. When the market recovers later, you have less capital exposed to the recovery — so the same percentage gain produces fewer dollars.

Mathematically, the worst case for a retiree is a deep early drawdown. Withdrawals during the drawdown lock in losses that compounding can't fully recover.

The opposite is also true: a strong start is huge. Retirees who started in 1995 saw five years of stock-market boom and were essentially indemnified against most of what came after. Retirees who started in 1969 or 2000 saw their first decade eat their portfolio.

The historical record

The worst US retirement starting years in modern history:

1969 — stagflation through the mid-1970s. A 4% withdrawer survived but barely.
1973 — oil shock + inflation spike. Similar to 1969 trajectory.
2000 — dotcom bust, then 2008 financial crisis hit before recovery. This window is what kept the safe-withdrawal-rate research conservative for the next decade.

The best:

1982 — start of an 18-year bull market.
1995 — late 1990s boom.
2010 — 11-year bull run before COVID volatility.

Same 4% rule, wildly different lived experience. The Bengen 4% number is calibrated against the WORST historical sequence (1969). In typical sequences, retirees could have withdrawn far more.

Why "average return" lies

If you ask a 401(k) projection what your portfolio looks like in 10 years assuming a 7% return, it draws a smooth line. The line is wrong — not in the average, but in the variance.

Real returns aren't smooth. They're a noisy signal with the long-run mean buried in volatility. A 30-year retirement that experiences:

5 years of -10% to +30% volatility
Followed by 25 years near the long-run mean

ends up looking very different from one that experiences:

25 years near the long-run mean
Followed by 5 years of high volatility

Same long-run mean. Different ending balance. The single-line projection captures none of this.

The 5-10 year danger window

Sequence-of-returns risk concentrates in the first 5 to 10 years of retirement. This is when:

Your portfolio is at maximum size — so a percentage drawdown is the biggest absolute dollar loss.
You're withdrawing — so you're locking in losses that compounding can't recover.
You have the longest remaining horizon to weather any subsequent recovery.

After year 10, the math gets meaningfully more forgiving. Returns matter less because you have less compounding runway and the portfolio has had time to either grow significantly or be drawn down to a more sequence-resilient size.

The implication: the years 60-70 (for typical retirees) are the most dangerous. The years 75+ are the most forgiving.

The four hedges

Four strategies that empirically reduce sequence-of-returns risk:

1. The bond tent

Be more conservative in years 60-65, then RE-RISK back to growth in your 70s. (Wade Pfau and Michael Kitces, 2014) The bond tent reduces drawdown exposure during the danger window without sacrificing long-term growth.

2. A cash buffer

Hold 1-3 years of spending in cash or short-duration bonds, separate from the long-term portfolio. In a market drop, spend from the buffer instead of selling stocks at lows. Refill the buffer in good years.

3. Variable spending

Cut discretionary spending in down years. Real retirees do this spontaneously — research suggests retirees naturally reduce non-housing spending by 5-10% in years following large portfolio drops. Formalising this with a spending guardrail (Guyton-Klinger or similar) makes the plan substantially more robust.

4. Part-time income in early retirement

A few years of part-time consulting or low-stress work in your early 60s materially reduces portfolio withdrawal pressure during the danger window. $20K/year of side income from age 62-67 is roughly equivalent to having $300K more saved at retirement.

What Monte Carlo actually tells you

A single-line projection asks "what's the ending balance assuming average returns?" The answer is one number.

Monte Carlo asks "across thousands of plausible return sequences, what fraction of them succeed?" The answer is a probability distribution.

The percentile band that matters most for sequence-of-returns analysis is the 10th percentile — the bottom 10% of outcomes. If your 10th percentile balance is positive at age 95, your plan survives the rough sequences. If it isn't, you're betting on getting lucky with the order of returns.

Yearfold's calculator reports the 10th-percentile balance alongside the median because the 10th percentile is the answer that actually defends your plan against bad sequences. The median is just where the average case lands.

Related reading: