What Is Sequence Risk, and Why Does It Matter in Retirement?

An average return does not tell the whole story. Nor does a chart showing that some capital remains. The useful test follows both the investment account and the cash available for spending.

In this article

Why do withdrawals change the effect of return order?

With no contributions or withdrawals, multiplying the same set of investment returns in a different order produces the same final value. The path still matters for interim liquidity and risk, but the multiplication is unchanged. Once money leaves, the amount exposed to each later return changes.

Suppose a household needs a particular rand income after a market fall. Paying that amount removes a larger proportion of the reduced account than it would have removed from a higher balance. A later recovery applies only to what remains invested. It does not earn a return on the money already spent.

This is not exclusively an early-retirement phenomenon. Any substantial cash flow can make order relevant, including contributions before retirement. The separate investor return-gap article addresses that measurement problem. Here the question is whether a retirement withdrawal policy remains workable under a difficult sequence.

The same five returns, with the weak years first or last

Use an invented starting portfolio of R5 million. One constructed record has returns of −20%, −10%, 0%, +10% and +20%. The other reverses them. There are no separately modelled fees or taxes, and the figures are nominal. These are hypothetical observations for explaining the mechanism, not a forecast or actual investment performance.

First take no income. Both records end at R4,752,000. The arithmetic average of the five annual returns is zero, but the compounded result is a 4.96% loss. Even before adding withdrawals, a simple average is not the compound return.

Now take R250,000 at the start of the first year, increasing that rand withdrawal by an assumed 5% each year. Total withdrawals over the five years are R1,381,408 in either order, rounded to the nearest rand.

The difference is R564,519, despite the same starting capital, annual withdrawal amounts and set of investment returns. Each annual row is calculated as (opening capital − withdrawal) × (1 + return). Full precision is retained in the calculation file; the displayed amounts are rounded.

The example does not establish that R250,000 is affordable for a real retiree or that a 5% annual increase matches that household's inflation. It isolates the interaction. A longer period, different returns, fees, tax and other income can materially change the result.

What changes if you always withdraw 5% of the current balance?

Replace the rising rand amount with a withdrawal of 5% of each year's opening balance. Keep the same two return records. Both now finish with R3,677,007.

That equality is mathematical. Each year multiplies the account by 95% and then by that year's return factor. Reordering those factors does not change their product. Yet the income paid during the period is very different.

In the weak-first record, income falls from R250,000 in year one to R154,328 in year four. That is a nominal reduction of about 38.3%, before considering higher living costs. The account has adapted by paying less. A household whose essential expenses cannot fall with it has not had its funding problem solved.

This is why “the portfolio does not run out” is an incomplete success measure. It can remain technically true while income becomes inadequate. Assess the resources left and the spending delivered, not one metric in isolation.

What does South African research add?

A 2022 University of Cape Town master's study by Kilian Petika Schabort examined sequence risk in South African retirement portfolios using local and other asset-index data, simulated returns and inflation, and several allocation strategies. Its repository abstract identifies the model and historical-data basis. That supports examining the issue in a South African setting; it is not evidence that a particular buffer or allocation will succeed for every retiree. The strategy rankings are not used here as a personal prescription.

The ASISA Standard on Living Annuities makes the practical risk allocation clear: living-annuity customers bear investment and longevity risk, and the selected income is not assured for life. It also distinguishes indicative guidance from an assessment using the annuitant's circumstances and other income. An industry illustration cannot replace the household budget.

Our examples establish an arithmetic mechanism, not a local probability of retirement failure. Estimating such a probability requires assumptions about returns, inflation, lifespan, costs, taxes and the withdrawal response. Different assumptions or definitions of failure can produce different results.

Can a living-annuity income simply change whenever spending changes?

Product terms matter. In Allan Gray's Living Annuity terms, effective 26 August 2026, the selected income and frequency normally change at the policy anniversary, subject to permitted exceptions. The document states current annual limits of 2.5% to 17.5% and explains that annuity income is taxable. This is a concrete provider example; check the actual contract and current law for another arrangement.

A household can reduce discretionary spending during the year without assuming its annuity instruction changes immediately. Conversely, a fall in the balance may make the next anniversary's permitted or chosen rand income very different. Align the budget decision, instruction deadline, gross payment and tax deduction rather than treating them as one number.

A rate within the permitted range is not automatically sustainable. Our two illustrated withdrawal rules remain inside those bounds over the constructed period, but that does not make either a recommendation. Their purpose is to show why the income policy and return path have to be read together.

Does a cash reserve remove the problem?

A reserve can fund spending without requiring an immediate sale from a depressed holding. Its usefulness depends on where the cash comes from, how much it must pay and what happens when it is used.

A comparison that adds a reserve to one portfolio without reducing its other assets gives that household more starting capital. It has not isolated the effect of the reserve. Similarly, a rule that replenishes cash only after favourable returns needs an answer for a long period without those returns.

The reserve also has costs and risks: its after-tax return, inflation exposure and the investment exposure forgone. Holding more cash can reduce one source of volatility without resolving a long retirement's purchasing-power need. The assessment needs the total household resources and a replenishment rule, not an assumed universal number of years in cash.

Which response can the household actually make?

Separate expenses that can change temporarily from commitments that cannot readily be reduced. Then identify other income, accessible resources and the dates at which investment income can be revised. A modelled spending cut is not a credible response if it falls on rent, necessary care or a contractual payment the household cannot alter.

Contractual lifetime income can change which risks the household retains, but price, escalation, survivor terms, liquidity and insurer obligations matter. The living-versus-guaranteed-annuity article examines that distinction. Combining arrangements is a separate planning decision; a difficult return sequence does not establish a universal annuity split.

Ask the review to show what actually happens after the weak period: the next gross withdrawal, net spending, remaining accessible capital, changed percentage draw and any action the household can carry out. A stress test becomes useful when the response is feasible, not merely when a spreadsheet contains a lower number.

ClearGauge's judgement is that sequence-risk planning has to protect a spending purpose, not just an account balance. The fixed-rand example exposes capital pressure; the percentage example exposes income pressure. Neither risk can be assessed properly without the household's commitments, other resources and tolerance for change.

This article has not calculated an appropriate withdrawal, asset allocation, reserve or annuity purchase for you. For the broader income target, read how much retirement capital is needed. The worked files can support a discussion of the mechanism, but cannot determine whether a personal retirement plan is funded.

Sources and as-of note: researched 6 September 2026. ASISA's linked standard displays an internal update date of 30 October 2024; the provider policy is effective 26 August 2026. The UCT reference is limited to its accessible repository abstract, not an independently reproduced simulation. Original five-year examples use constructed returns, annual withdrawals before returns, no separate fees or tax, and no prediction of future performance. General education, not personal financial or investment advice.