Sequence-of-returns risk, in plain English

Here is the idea, stated as plainly as possible. During your working years, you are putting money in. The order in which the market goes up and down barely matters, because you're buying the dips on the way. During retirement, you are taking money out. The order in which the market goes up and down matters enormously, because a bad stretch early in retirement forces you to sell more shares at lower prices to meet the same withdrawal — and those shares are not coming back.

The academic name is sequence-of-returns risk. The everyday version is: the first five years of retirement are doing most of the work of the next thirty.

Consider two retirees, both 65, both starting with $1 million, both withdrawing 4% a year — $40,000 — adjusted for inflation. Both portfolios average 7% over the next 25 years. But one retires in year zero of a bad decade — two down years, then a flat year, then recovery — while the other retires into a strong first five years. Same average. Same withdrawal. At year 25, the first retiree has run out. The second has more than she started with. The returns were identical; the order wasn't.

This is why the year you retire feels, honestly, like luck — and in an important sense, is. A 62-year-old who took an early package in 2008 had a very different thirty years than a 62-year-old who retired in 1994. Neither made a better decision. They made the same decision in a different market.

The right response isn't to predict markets. Nobody has ever done that for a living without eventually being wrong about it. The right response is to build a withdrawal plan that bends under a bad sequence without breaking. A few of the things that tend to bend well:

Hold two to three years of expected withdrawals in short-dated bonds or cash instruments. In a bad stretch, you spend from that bucket and leave the equity portfolio alone until it recovers. This is unglamorous and works. It costs you some expected return in average years. It buys you enormous resilience in bad ones.

Use a dynamic withdrawal rule rather than a fixed percentage. Guardrails — spending goes up a little after a good year, down a little after a bad one, within a defined band — preserve roughly the same lifetime spending while substantially lowering the probability of the worst outcomes. The math on this is well-established; the psychology is the hard part. Clients mostly don't want to cut spending. The small cuts early are almost always smaller than the large cuts later.

Fill the low-bracket years. The runway between retirement and the first required minimum distribution is usually the cheapest tax real estate in a life. Roth conversions and capital-gain realizations done in that window don't increase returns, but they meaningfully change the order in which tax gets paid — which is, again, a sequence problem.

Finally, and unfashionably: spend from the right account first. Pulling from the taxable brokerage while the pre-tax IRA compounds untouched feels safer than the reverse, but it leaves a tax torpedo in year 73 when RMDs hit. The correct order is household-specific, and it is worth building once and revisiting every few years.

The bumper-sticker version of all of this: in accumulation, averages win. In distribution, order wins. A retirement plan that only talks about average return is telling you half the story.

There is one planning decision that interacts with sequence risk in a way most people don't consider until it's too late: when to claim Social Security. A 63-year-old who retires early and begins drawing from her portfolio immediately faces maximum sequence risk — every dollar of withdrawals comes from a pool of assets that may be compressing in a bad stretch of markets. The same person, if she can delay Social Security to 70, essentially replaces a portion of those portfolio withdrawals with a guaranteed income stream that doesn't decline with markets and adjusts for inflation automatically. The result is that she draws less from the portfolio in the years it might be most vulnerable. Sequence risk is partly a portfolio problem and partly a cash-flow architecture problem — and the two levers are worth pulling together.

None of this eliminates the uncertainty of markets. What it does is reduce the dependence of the retirement outcome on the specific timing of the bad decade — which is the part you cannot control. A plan designed around sequence risk is not a plan that predicts returns. It is a plan that survives the wrong ones.

Part-time work in the first few years of retirement is underrated as a sequence-risk tool, and not for the reasons usually cited. The conventional argument is that it provides income. The more important benefit is that it reduces the portfolio withdrawal rate during the highest-risk period. A retired engineer who earns $25,000 doing consulting two days a week for three years has effectively shifted roughly $75,000 of withdrawals off the portfolio during what may be its most vulnerable window. Whether that matters depends on what the market does — but the option to work part-time is one of the few sequence-risk hedges a retiree controls entirely, without needing to predict anything. The plan that accounts for this option is more durable than the one that doesn't, even if the part-time work never materializes.