Why the Order of Returns Matters Once You Start Withdrawing

What is sequence of returns risk?

Two people can experience exactly the same market — the same years, the same returns, the same average — and end up in completely different positions. Not because one of them chose better investments, and not because one was luckier about what the market did. Only because of the order the years arrived in.

That is a strong claim and it can be settled without any forecasting, because it can be tested inside a closed set of numbers. Take the 26 complete calendar years of Indian equity total returns from 2000 to 2025. Rearrange them. Nothing is added, nothing is removed, and the compound return of the whole stretch — 13.2% a year — is identical in every rearrangement. Then see what happens to three different people living through those same years.

The third of them is the control, and it is what makes the result mean something.

How much does the order of returns change the outcome?

	Worst years first	Best years first
Nobody touches the money	₹25.17 crore	₹25.17 crore
Taking an income out each year	₹0.00 lakh	₹20.92 crore
Paying an amount in each year	₹12.22 crore	₹70.73 lakh

Assumptions: ₹1 crore starting corpus; ₹6.00 lakh annual withdrawal taken at the beginning of each year (flat in nominal terms); ₹1.00 lakh annual contribution paid at the beginning of each year. Returns are the 26 complete calendar years of Indian equity total return index, 2000–2025. The hold-only control and the two extreme orderings use that fixed return set. The random-ordering statistics use 20,000 shuffles with a fixed seed so every published figure regenerates identically. No fund costs, no tax, no inflation adjustment applied.

Read the first row first, because it is the reference point everything else is measured against. Somebody who invests and then simply leaves the money alone ends at exactly the same value under both orders. Not approximately — identically, to the limit of the arithmetic. This is not an empirical finding but a necessary one: with no money moving in or out, the years are just numbers being multiplied together, and multiplication does not care about order.

That row is the control, and it is why the other two rows cannot be explained away. Whatever is producing the differences below is not the returns, because the returns are the same. It is the cash flow.

Now the second row. Someone drawing an income from the portfolio is wiped out when the bad years come first, and finishes with ₹20.92 crore when they come last. Same returns. Same average. One of those people ran out of money and the other did not.

And the third row is the one worth sitting with. For someone paying money in rather than taking it out, the ordering that destroyed the retiree is the best thing that could have happened: ₹12.22 crore with the bad years first, against ₹70.73 lakh with the good years first. That is 17.3 times better from having the crashes at the beginning.

Why does sequence risk hit savers and retirees differently?

The mechanism is simple once separated from the arithmetic, and it is worth having in plain terms because it is what makes the finding usable.

What matters is not what the market did. It is how much money you had in the market when it did it.

A retiree starts with everything and spends it down. Their largest balance is at the beginning, so a fall in the early years hits the biggest pile of money they will ever have — and, crucially, they are selling into it. Every withdrawal during a downturn permanently removes units that would otherwise have participated in the recovery. The portfolio can recover; the money already spent cannot come back. A bad first five years is therefore not a bad patch to be waited out. It is permanent damage, compounded for the rest of the retirement.

Somebody still contributing has the mirror-image position. Their largest balance is at the end, so an early fall hits a small pile, and every contribution made during it buys at reduced prices. A crash in year two is a discount on twenty years of future buying. What genuinely hurts them is a crash in the final years, when the balance is at its largest and there is no time left to recover.

Measured across 20,000 random orderings of these years, the relationship between the first five years' returns and the final outcome is 0.74 for the person withdrawing and -0.71 for the person contributing. Almost equal in strength and opposite in sign, which is what the mechanism predicts.

When does sequence risk reverse?

This gives a precise answer to a question that is usually answered vaguely.

For most of a working life, volatility is not the enemy. Somebody contributing steadily is, in a real sense, buying more when prices fall, and the years of ugly returns in the middle of that period are doing them a favour they cannot feel at the time. The advice to stay invested through a crash is not stoicism; for a net contributor it is arithmetic.

That protection fades as the balance shifts from contributions still to come towards corpus already built. The sign of the early-return correlation is opposite for a pure net-contributor and a pure retiree — that is what the correlation figures above show. What changes continuously is the proportion of each that applies to a particular person: someone with a large corpus and only a few contributions left already has most of the sequence-risk exposure of a retiree, even while they are still working. The same person, holding the same portfolio, increasingly faces a risk that runs the other way, and nothing about the portfolio signals the shift.

This is why Ramesh, a few years from retiring, is right to feel that a bad stretch would matter more now than it did a decade ago. It genuinely would, and the feeling is not a loss of nerve. It is the most dangerous window of his entire financial life: the balance is near its maximum, and the contributions that used to buy the dips are about to end.

Does sequence risk affect people still saving?

There is a second finding in the third row that is easy to skip past.

Across those same 20,000 orderings, the contributor's outcome ranges by a factor of 3.2 between the poor and the good cases — and again, that is with the returns held fixed. Someone paying in the same amount every year, into the same market, over the same 26 years, can end up with several times as much or several times as little purely according to when the good years happened to fall.

So the order of returns is not only a retirement problem. It sets a floor on how precise any projection of a future corpus can be. A plan that produces a single number for what a portfolio will be worth in twenty years is not wrong by a small margin; it is answering a question that has a range of answers even when the returns are known in advance.

What does this test not show?

Being clear about the limits matters here, because this experiment is stronger than most and it would be easy to over-read.

It is a rearrangement, not a forecast. Shuffling destroys whatever pattern real markets have — any tendency for bad years to cluster, or for a crash to be followed by a rebound. So the orderings tested include many that markets would be unlikely to produce. That is acceptable for demonstrating that order matters, and it is not acceptable as an estimate of how often a real retiree runs out of money.

It is one market over one stretch. 26 years of Indian equity, containing one very large crash. A different market or a different period would give different amounts. The direction of the effect, and the fact that it vanishes without cash flows, would not change.

The withdrawals are held flat rather than rising with prices. A real retiree needs more rupees each year to buy the same things, so the second row is kinder than reality — deliberately, because inflating the withdrawals honestly would need an inflation series reaching back to 2000 and we hold one that begins in 2013.

It is shorter than a retirement. 26 years is not thirty, and the record cannot supply thirty. That limitation is the entire subject of the limits of a safe withdrawal rate, and it is the reason this page reports a mechanism rather than a rate.

How can sequence risk be made survivable?

The risk cannot be removed. It can be made survivable, and the responses follow directly from the mechanism rather than from anybody's forecast.

The first is to hold enough non-volatile money to fund the early withdrawals, so that a fall in the first years is met by spending from the stable part rather than by selling equity into it. That is what the bucket approach is for, and this experiment explains why the argument exists — the damage in the first row came entirely from selling during declines. Whether a finite buffer outlasts a specific downturn is a separate question that rearranging a fixed return set cannot answer.

The second is to reduce risk before the transition rather than after, on a schedule, which is the glide path. The window that matters is the few years either side of the last pay cheque, and the reduction has to be in place before it opens.

The third is flexibility in the withdrawal itself. A retiree who can take less in a bad year — by deferring something, or by having some income that is not from the portfolio — addresses the mechanism at its source, because the damage comes from selling during the fall. How much this helps depends on how much flexibility actually exists.

And the fourth is simply to know which side of the line you are on. The same market event is a discount to one person and permanent damage to another, and the only difference is the direction the money is flowing.

What to take away

Order does not matter at all while nobody touches the money — the control row proves it exactly. It matters enormously once money is moving, and it matters in opposite directions depending on which way.

For someone still paying in, early crashes are a benefit and the dangerous years are the last ones. For someone drawing an income, early crashes are permanent damage, and in this test the worst ordering emptied the portfolio while the best left it larger than it started. Where any particular person sits between those two positions depends on how much corpus is already built relative to how much future contribution remains — and for someone with a large corpus and few contributions left, the balance can already resemble a retiree's well before the last pay cheque arrives.

Several years of spending held outside equity before the withdrawals begin, risk lowered on a schedule rather than in response to markets, and some ability kept to spend less in a bad year — none of that improves the returns. It changes how much damage a bad order can do, which — as the first row shows — is the only thing that was ever in play.