Sequence risk: when the order of returns decides your retirement

In depth8 min read · 3 sources

Sequence risk (sequence of returns risk) describes a trap in the withdrawal phase: as long as you only invest and withdraw nothing, it does not matter in which order good and bad years come. As soon as you withdraw money regularly, that changes: a stock market slump in the first years of retirement can shrink your wealth permanently, even though the average return over the whole period stays the same.

Without withdrawals, the order does not matter

With a lump-sum investment and no inflows or outflows, the annual returns multiply. And in multiplication, the order makes no difference: 0.8 × 1.2 gives the same as 1.2 × 0.8. A portfolio that first loses 20% and then gains 20% ends up in exactly the same place as one that first gains and then loses.

With withdrawals, the calculation looks different. Each year, a fixed amount is deducted first and only the remainder earns a return. A loss then hits a portfolio that is already smaller because of the withdrawal. And the money that was sold at the low can no longer recover.

Worked example: same returns, different order

The reason: in order A, wealth falls sharply in the first two years, and the withdrawals of €25,000 each make up an ever larger share of it. The good years that follow then work on a smaller base. In order B, wealth rises first; the later losses do hit a large portfolio, but until then the withdrawals came largely out of gains.

Formally, this is the same logic as with the money-weighted return: with withdrawals, a return counts for more the larger the wealth is at the time. In retirement, wealth is largest at the beginning. That is why the first years carry the most weight.

The mirror image in the accumulation phase

With regular contributions, the effect is reversed: now losses at the end are dangerous, because they hit the largest portfolio. The same ten returns as above, this time with a €10,000 contribution at the start of each year:

That is why the period shortly before retirement is so delicate: wealth is at its largest, and hardly any contributions are left to make up for a slump. A glide path that reduces risk during this time targets exactly this point.

The risk zone around retirement

Sequence risk is not limited to the withdrawal phase. In the accumulation phase, it works as a mirror image: a slump shortly before retirement hits the largest portfolio you have ever had and leaves little time to recover. Planners therefore speak of a retirement risk zone of about five to ten years before and after the start of retirement.

• Young and saving: a crash is actually rather favourable, because the next contributions buy cheaply. The portfolio is small relative to future contributions.
• Shortly before retirement: the portfolio is large and the remaining contributions are small. A loss of 30% can hardly be made up.
• Early in retirement: withdrawals sell at the low and make losses permanent. This is where the risk is greatest.
• Late in retirement: the remaining time is short, and the impact of a slump on the question “Will the money last?” becomes smaller again.

What the research shows

Milevsky and Robinson (2000) calculate the probability that a portfolio from which a fixed amount is withdrawn each year runs out before the end of life (probability of ruin). They assume lognormally distributed returns and compare this self-annuitisation with buying a lifetime annuity: the annuity protects against running out of money but costs flexibility. Sequence risk is at the heart of this probability of ruin: with the same average return, the money lasts on some paths and not on others.

Pfau and Kitces (2013) propose not lowering the equity allocation in retirement but letting it rise again from a low level (a rising equity glide path). The idea: a low equity allocation is most valuable precisely in the risk zone; later, when sequence risk declines, it can rise again. Their simulations show that such a path can reduce the risk of running out of money early. More on this in the article equity allocation by age.

Because historical data contain only a few independent retirement periods, sequence risk is usually studied with simulations. Glasserman (2003) describes in detail how Monte Carlo methods work and where their sources of error lie; the article Monte Carlo simulation explains how investors can apply them.

What helps against sequence risk

1. Flexible withdrawals: if you withdraw 10 to 20% less in bad years, you sell less at the low. Rules with guardrails (cutting the withdrawal when the withdrawal rate exceeds a limit) reduce the risk considerably, but require room in the budget.
2. Cash buffer: one to three years of withdrawals in an instant-access savings account mean you do not have to sell for a while after a slump. In good years, the buffer costs return.
3. Less risk in the risk zone: a lower equity allocation around the start of retirement, possibly rising again later.
4. Secure basic income: the state pension, a company pension or a private annuity cover fixed expenses; the portfolio only finances the flexible part.
5. Lower initial withdrawal: the safe withdrawal rate is precisely the rate that would have survived even the worst historical sequence.

None of these measures is free: a buffer or a lower equity allocation reduces the expected return. It is a trade-off between return and security that everyone has to weigh up for their own situation.

For advanced readers: measuring sequence risk

A simple metric is the range of final values when you play the same historical returns through every possible starting year (rolling retirement periods). More meaningful is the distribution from many simulated paths: the success rate (share of paths in which the money lasts), the median and the poor percentiles of final wealth, and the year in which the money runs out on the poor paths.

In the simulation, the method matters. Drawing individual months with replacement (bootstrap) breaks up phases in which bad years follow one another; a block bootstrap with contiguous segments is more likely to preserve such runs. Normally distributed returns underestimate extreme losses; a Student's t distribution with fat tails is more cautious (see skewness and kurtosis).

Mathematically, sequence risk arises because the final value with withdrawals is a sum of products: each withdrawal is only compounded with the returns after it is made. Early returns affect all later withdrawals, late returns only a few. That is why the return of a withdrawal plan cannot be described by a single average figure.

An intuitive measure is the wealth-weighted average return: each annual return is weighted by the wealth invested in that year. In the example with withdrawals, wealth in the loss years is large in order A and small in order B. Accordingly, the weighted averages are far apart, even though the simple mean is the same. In essence, this measure corresponds to the money-weighted return of the withdrawal plan.

How Portfolio Lab calculates it

The Monte Carlo simulation draws monthly returns by bootstrap, block bootstrap, normal distribution or Student's t distribution and calculates withdrawals with inflation adjustment; it shows the success rate and percentile fan in nominal and real terms. From the historical series, the Backtest calculates the safe withdrawal rate (the highest inflation-adjusted monthly withdrawal at which the money lasts until the end) and the perpetual withdrawal rate (at which the real starting value is preserved). In Financial goals, you can combine withdrawals, income and a glide path. Details are in the Monte Carlo methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.