Sequence of Returns Risk: Why Order Beats Average

Here are two retirees. Both start with $1 million. Both hold the same fund. Both take out 4% in year one and raise it 3% a year for inflation. Over thirty years they earn the exact same compound annual return, down to the decimal: 10.26%.

One finishes with $7,793,981. The other finishes with $10,469,341.

The only difference is the order the returns arrived in. That is a $2,675,360 spread produced by nothing but scheduling, and it is the single most under-appreciated number in retirement planning. The first time I ran this I assumed I had a bug in the loop. I did not. Order is a real, enormous, uncompensated risk, and it has a name: sequence of returns risk.

The experiment: run thirty real years forward, then backward

I used the S&P 500 annual total returns from 1996 through 2025, the series NYU Stern maintains from market data and updates every January.^[1] Thirty years, dividends reinvested, no fees. The arithmetic average is 11.80% a year and the compound rate is 10.26%. Retiree A gets them in calendar order, so 1996 first. Retiree B gets the identical thirty numbers in reverse, so 2025 first and 1996 last. Both start with $1 million, both withdraw $40,000 at the start of year one, and both raise the withdrawal 3% every year after.

Both retirees pull out exactly $1,903,017 across the thirty years. Same money in, same money out, same average. Watch what happens in between.

And this sequence was kind. Ten percent a year compounded is a genuinely good thirty years, which is why nobody actually goes broke in the chart above. Raise the withdrawal to 7% and the mercy runs out: the calendar-order retiree is scraping the account for a final $21,378 in year 29 and is done, while the reversed retiree closes year 30 with $4,283,473 still in the account. Same returns. Same withdrawal rule. One of them ran out of money.

Why the order is free while you are saving and fatal while you are spending

Okay, here is the part that took me a while to get comfortable with, because on its face it sounds like the market treats savers and spenders differently. It does not. The math is the same. The cash flows are what change.

Leave $1 million alone across those thirty years, no deposits, no withdrawals, and it becomes $18,717,165. Reverse the returns and it becomes $18,717,165. Not approximately. Exactly. You are multiplying the same thirty growth factors together and multiplication does not care what order you do it in. That is the whole reason compounding behaves so predictably over long holding periods when nothing is moving in or out.

Add a withdrawal and the commutative property dies, because now you are selling shares at whatever price the market happens to be quoting that year. Say your fund trades at $1,000 a unit and you need $50,000. In a normal year you sell 50 units. In a year like 2008, when the S&P 500 total return was -36.55%, that unit is worth $634.50 and the same $50,000 costs you 78.8 units. You just sold 57.6% more of your portfolio to buy the same groceries, and those 28.8 extra units are gone. They do not come back in 2009. A paper loss becomes a permanent one at the exact moment you convert it into a sale.

Now, the version of this you usually hear is that order does not matter at all during accumulation. That is only true for a lump sum. Run those same thirty years with someone contributing $20,000 annually and they end at $4,004,562 in calendar order against $3,145,559 reversed, a 27.3% gap. Order still matters. What is different is that it can never ruin them, because a saver hitting a crash is buying units cheap rather than liquidating them cheap. Sequence risk does not appear at retirement. It flips sign. That is also why the lump-sum versus dollar-cost-averaging argument lands so differently before and after you stop earning.

The two dates that made this famous: 1966 and 2000

William Bengen, a financial planner in Southern California, ran the original version of this study for the Journal of Financial Planning in October 1994. He tested rolling 30-year retirements against actual US stock, bond and inflation history and concluded that a 4% first-year withdrawal “should be safe,” because “in no past case has it caused a portfolio to be exhausted before 33 years.” He also found that a 5% withdrawal “survived only about 20 years” for people who retired in the late 1960s and early 1970s.^[2] That gap between 4% and 5% is the whole conversation. Later work refined the worst case to a 4.15% starting rate for a 1966 retiree holding 50/50 stocks and bonds.^[3]

Why 1966? Because the Dow sat at 983 in January 1966 and closed at 991 in October 1982, going absolutely nowhere for sixteen years while monthly CPI climbed from 0.9% to 13.6% at its 1980 peak.^[4] Flat nominal markets are survivable. Flat nominal markets plus double-digit inflation are not, because the withdrawal keeps ratcheting up while the portfolio does not. The 1966 retiree was not killed by a crash. They were killed by sixteen years of nothing.

The modern version is January 2000, and I ran it rather than quoting it. Same rules as before: $1 million, all in the S&P 500, $40,000 in year one, 3% annual inflation adjustment.

What actually defends against it, honestly ranked

Most of the advice here is soft. Some of it is measurable. Start with the measurable part, because it is also the least glamorous.

Diversify and rebalance. Take that identical January 2000 retiree and put them in a 60/40 stock and bond portfolio rebalanced annually instead of 100% equities. They finish 2025 with $965,495, essentially their starting balance, instead of zero. Ten-year Treasuries returned +16.66%, +5.57% and +15.12% in the three years the S&P 500 was falling apart, and +20.10% in 2008.^[1] That is what a second asset is for. It is not about return, it is about having something to sell that is not down. If you take one thing from this piece, make it an actual rebalancing policy you follow on a schedule, because rebalancing is the mechanism that forces you to sell the winner and hold the loser through the exact years you will not want to.

Hold a cash or short-bond bucket sized in years, not dollars. Two to three years of spending in something that does not move means you never have to fund a grocery bill by selling equities down 37%. It is the same instinct as an emergency fund measured in months of expenses, scaled up for a household with no paycheck. The honest cost is drag: three years of spending parked in T-bills is roughly 10% of a portfolio earning a low return for decades. I think that is a fair price for removing forced selling, but it is a price, not a free lunch.

Consider a rising equity glidepath. Michael Kitces and Wade Pfau tested portfolios that start conservative and get more aggressive through retirement, which is the opposite of what target date funds do. A path from 30% to 70% equities produced a 95.1% success rate against 93.2% for a static 60/40, and at the 5th percentile it lasted about 30 years against 27.7.^[5] The logic is exactly the mechanism above: hold the least equity when the balance is largest and the sequence exposure is worst. I find this genuinely persuasive and I also notice it asks retirees to buy more stock at 75 than at 65, which very few humans will actually do.

Spend flexibly. The fixed inflation-adjusted withdrawal in every simulation above is a fiction. Real people cut back after a bad year. Guardrail rules formalize that by trimming the withdrawal when the portfolio falls past a threshold and raising it when it runs hot. Morningstar's The State of Retirement Income, 2025 edition, published December 3, 2025, puts the safe starting rate at 3.9% for a 30-year horizon at 90% success with 30% to 50% in equities, and notes that simply acknowledging that real spending declines with age is worth roughly another percentage point.^[6] Flexibility is the cheapest defense on this list and the one people resist hardest, because it means the plan can tell you no.

Delay Social Security. For anyone born in 1960 or later, full retirement age is 67, and waiting until 70 raises the benefit to 24% above the full amount while claiming at 62 cuts it by 30%.^[7] That is an inflation-linked, government-backed income stream you cannot outlive, bought with three years of patience. Every dollar it covers is a dollar you are not selling shares to produce. It is the best longevity hedge available to a normal household and it costs nothing but the willingness to spend down other assets first.

The part that changes how you plan

Sequence risk is usually described as crash risk, and that framing is wrong in a way that leads people to the wrong defense. Kitces measured it: the correlation between a retiree's first-year return and their eventual safe withdrawal rate is only 0.21. The correlation with first-decade real returns is 0.79.^[8] A single bad year is noise. A mediocre decade is the thing that kills you, and a mediocre decade does not announce itself. There is no headline, no circuit breaker, no moment where anyone rings a bell.

Which means the defense cannot be a reaction. You cannot see it coming and you cannot time your way out of it. The 2000 retiree above did not make a single mistake in 26 years; they held a diversified index fund through everything and followed the most famous rule in retirement planning. They just happened to start on a bad Tuesday.

So build the defenses before you need them. Get the allocation right while you still have a paycheck, fund the cash bucket in the last year of work rather than in the first bad quarter, write the spending rule down while you are calm, and treat the 4% rule as a starting estimate rather than a promise. The order of returns is the one variable in your plan you have absolutely no say over. Everything else on this page, you do.