The Order of Returns

A fixed withdrawal against a balance that is not fixed

Sequence risk is not a market phenomenon, it is an interaction. A retiree who withdraws a fixed real amount from a portfolio that moves up and down is selling a **number of shares** that changes with the price: at $741,000 the same $50,000 buys more shares than it did at $1,000,000. Those extra shares are sold permanently, and the recovery then happens to a smaller position. Everything about the risk follows from that one sentence. The averaging intuition fails here, and it fails in a specific way. Over a long horizon the average return of a portfolio is a reasonable thing to plan with, but a withdrawal schedule converts the arithmetic mean into a path-dependent result: the same returns in a different order produce different outcomes, because the withdrawals are taken at different prices along the way. Two retirees with identical returns, identical spending and identical portfolios end a decade apart. In the accumulation phase the mechanism runs the other way, which is worth knowing because it changes how a bad market should feel. A saver contributing into a fall is buying more shares with each contribution — the same fixed-contribution effect, working in their favour. That is not a reason to hope for a crash; it is a reason not to stop contributing during one. One year, two orders — Bad-first: ($1,000,000 − $50,000) × 0.78: $741,000 · Bad-last: ($1,000,000 − $50,000) × 1.07: $1,016,500 · Gap after one year: $275,500 · Recovery the bad-first retiree needs: 34.95% ← The bad year is not larger on one side; it is only earlier. That is the whole lesson, and no forecast appears anywhere in it.

Why the first decade carries the weight

A retirement is not uniformly exposed to sequence risk. The early years do the damage, and the reasons compound: the balance is at its largest, so a percentage fall costs the most dollars; the withdrawals are being taken from the most vulnerable position; and the recovery needed is a larger percentage than the fall. By the time a portfolio is a decade into retirement, the sequence of the remaining years matters far less, because there is less capital at risk and fewer withdrawals left to take. That is why the “4% rule” is a worst-case-over-history number rather than a prediction. It comes from asking what withdrawal rate survived the bad starting points in the historical record — high valuations, poor first decades — and it is a useful anchor precisely because it was built from the bad sequences rather than the average ones. It is not a promise, and its own author has been clear that it depends on the assumptions behind it. The practical implication is that the starting point matters more than the average market. Two retirees with the same portfolio and the same spending can have very different odds depending on whether they began after a long bull market or after a drawdown — which is an argument for flexibility in the first decade rather than for a more precise forecast, because nobody knows in advance which decade they are in. Where the exposure is — Years 1–5: the largest balance and the largest dollar cost of a fall · Years 5–15: still decisive — the recovery has to happen with fewer shares · Years 20+: much less sensitive; the sequence has already done its work ← A withdrawal rate is a starting rule, not a guarantee. Its safety depends on the valuation you start from, the fees you pay, the flexibility you keep and how long the money has to last — four things a single percentage cannot carry.

Four mitigations, one mechanism

Every mitigation breaks the link between a bad market and a forced sale. **A cash sleeve** of one to three years of withdrawals means a bad year is funded from cash, so the shares stay invested to participate in the recovery — the buffer of PF3, extended into retirement. **Guardrails** are pre-agreed spending rules: if the portfolio falls 20%, spending is trimmed 10%; if it rises well past the plan, spending can rise a little. The rules are written in a good year, when they are easy, so that they exist in a bad one, when they are not. A **slightly lower equity weight at the start** reduces the size of an early shock, at the cost of some long-run return — a deliberate trade of expected return for sequence protection during the window where the exposure is highest. And **flexibility about timing** is the most powerful of the four and the least glamorous: working one more year adds a contribution and removes a withdrawal, which improves the arithmetic of both ends of the sequence at once. Part-time work in the first years of retirement does something similar. What the four share is worth stating plainly, because it is the exam answer: none of them requires knowing what the market will do. They change the structure of the plan — where the money sits, how rigid the spending is, how much risk is carried at the start, and when the withdrawals begin — and that structure is what decides whether a bad decade is survivable. The four, and what each breaks — One to three years of withdrawals in cash: the bad year is not funded by selling shares · Guardrails agreed in advance: spending drops before the portfolio is forced to · A lower equity weight at the start: the early shock is smaller · One more year of work: adds a contribution and removes a withdrawal ← These are the same ideas as the buffer, the band rule and the horizon-first allocation, applied to the decade where all three matter most.

The withdrawal rule, and why a fixed percentage is a choice

Sequence risk is usually introduced with a spending rule that never changes: withdraw a fixed real amount each year whatever the portfolio does. That assumption is where most of the drama comes from, and it is worth separating the risk from the rule — because a household that is willing to adjust its spending has a materially different problem from one that is not. The origin of the fixed-amount convention is the safe-withdrawal-rate research, which asked a narrow and useful question: what constant real withdrawal, over every historical starting point, would have survived a particular horizon in the worst case? The answer that became famous was around four percent of the starting balance, inflation-adjusted, for a thirty-year horizon with a particular stock-and-bond mix. What matters about that number is what it is not. It is a worst-case result from one country’s history, not a probability; it is sensitive to the mix, the horizon, the fees, and the sample period; and it describes a rule nobody actually follows, because it withdraws the same amount in the second year of a crash as in the first year of a boom. That last point is where the practical decisions live. A **dynamic rule** adjusts the withdrawal to the portfolio’s condition — reduce spending after a decline and restore it after a recovery — which is the single most effective mitigation of sequence risk, because it prevents the sale of a large number of shares at depressed prices. The cost is that the household must accept a variable income, which is a real constraint rather than an accounting one. A **banded rule** does the same thing with a lag: withdraw a percentage of the current balance, with guardrails that trigger a modest cut or raise when the rate drifts outside a range. Both keep the portfolio inside the range where recovery is possible; both require the household to have decided in advance that it will accept the adjustment. There is also a **floor** to consider, and it is a different instrument rather than a different rule. A guaranteed income stream covering essential spending converts the variable part of the problem into a discretionary one: with the essentials funded, a portfolio drawdown is a reduction in the discretionary budget rather than a threat to housing and food. That is the honest argument for annuitising part of a balance, and it is behavioural as much as actuarial — the household that is not frightened of its own spending rule is the household that keeps it. The practical conclusion is that the safe-withdrawal-rate number is an input to a decision about how flexible the household is willing to be. A rigid real spending rule demands a low starting rate and a large portfolio. A household prepared to adjust can start higher, provided the adjustment is pre-committed, because the mechanism that causes ruin is not the withdrawal rate itself — it is the unwillingness to change it during the one decade that decides the outcome. • Safe-withdrawal-rate figures come from one history and one fixed-spending rule, not from a probability. • Adjusting spending after a decline is the most effective mitigation of sequence risk. • A guaranteed floor converts a portfolio drawdown into a discretionary-budget question. • The mechanism of ruin is refusing to change the withdrawal during the decisive decade. A useful pre-commitment: write down the spending adjustment rule before retirement, including the trigger and the size of the cut. The rule is worth more than a tenth of a percentage point on the withdrawal rate, and it is the part that is not in any of the published tables.

Sources

• Determining withdrawal rates using historical dataBengen (1994), Journal of Financial Planning; the Trinity study
• Sequence-of-returns risk and the retirement red zoneStandard portfolio research; Vanguard and Morningstar retirement-income studies
• Guardrails and dynamic withdrawal rulesGuyton–Klinger decision rules; Kitces retirement research
• Sequence risk in the accumulation phaseStandard dollar-cost-averaging arithmetic (accumulation is the mirror image)