Investing Classics: Retirement & Withdrawal StrategyDecumulation — The Nastiest Problem in Finance

The method is what deserves study. Bengen did not divide by an average return; he treated every single starting year since 1926 as a live experiment — a 50/50 stock/bond portfolio debited year after year — and then took the worst starting point of them all, which he named SAFEMAX. That makes 4% a floor meaning "even the unluckiest retiree did not go broke," not a reasonable expectation. And the variable that pushes that floor down is not equity returns but inflation: it raises the amount siphoned out at the very moment the portfolio is shrinking.

The worst starting point was the late 1960s (1968–1969), with a SAFEMAX near 4.1% — that cohort alone is what capped the rule at 4%. The culprit was not a crash but stagflation: the Dow touched 995 in early 1966 and did not durably clear 1000 until late 1982, 16 years of going nowhere in nominal terms, while CPI rose roughly 180% cumulatively (peaking at 13.5% in 1980). In real terms the index was halved, while the inflation-adjusted withdrawal had nearly tripled. Start instead in 1982 and the same portfolio supports 8% and up with room to spare.

Three limits. The data is 20th-century American — using the most successful market in history as your "worst case" carries built-in survivorship bias: plugging in the low bond yields of the day, Pfau, Finke and Blanchett found the 30-year failure rate leaping to roughly 57% (2013). It assumes you withdraw like a machine, whereas flexible rules (the Guyton-Klinger guardrails) lift the sustainable rate materially. And it is a floor, not a plan: across most starting years, someone holding to 4% ends the 30 years with more than they began, with a median above twice the original principal — Bengen himself later revised it up to 4.5%.

• Is my withdrawal rate a stress-test floor, or have I quietly turned it into a return expectation?
• Which of my expenses are rigid, and which could be cut 20% in a bad year?
• Is the denominator pre-tax market value, or what actually lands after tax and fees?

Hold a sum with nothing going in or out and order is irrelevant — multiplication commutes. But withdraw a sum each year and it stops commuting: shares sold at a low have permanently exited the compounding that follows. The risk also has a brutal time distribution — the first decade very nearly decides everything. Kitces has shown repeatedly that the sustainable withdrawal rate correlates strongly with the real returns of those first ten years: once the principal is ground thin early, later good returns act on a far smaller base.

The year 2000 is the textbook bad sequence: the start was the bubble's peak, followed by the halvings of 2000–2002 and 2008, leaving the S&P 500's nominal total return for the decade close to zero and negative in real terms. 2008 compressed the same lesson into months: the S&P 500 fell from 1,565 on October 9, 2007 to 677 on March 9, 2009 (−57%), and did not recover nominally until March 2013. Anyone forced to sell equities at that trough converted the drawdown into a permanent loss.

There is an antidote, and there are boundaries. It is only lethal under rigid withdrawals: taking 10% less in bad years and pausing the inflation adjustment cuts failure rates sharply — flexibility beats allocation here. But it cannot be dodged: sitting in cash to avoid an unknown bad sequence buys uncertain protection with a certain opportunity cost. Two misuses: treating it as proof that "equities are too dangerous" and turning so conservative that market risk is merely swapped for longer-lived inflation risk; and watching equities alone — in 2022 the S&P 500 returned −18.1% and the U.S. Aggregate bond index −13.0%, so the "bond cushion" simply was not there.

• If my first three years open like 1968 or 2000, does the plan still hold?
• Do I have a spending-cut rule written down in advance (trigger, and how much), rather than deciding in the moment?
• Is the order in which I raise cash predetermined, or do I sell whatever has fallen most?

Harold Evensky formalized this as a "cash-flow reserve": roughly two years of net spending in cash and short-duration bonds, the rest invested for the long run. Be honest about the trade: mathematically it usually does not win — Estrada, Kitces and others find bucket approaches end with slightly less wealth than a simple rebalanced total-return portfolio, because the cash bucket is a permanent drag. The value sits on the behavioral side: knowing the next two years of spending is already sitting there demotes a halving from "existential threat" to "ugly quote." Mental accounting is an irrationality in Thaler's work — money has no labels — and the elegance of buckets is that they turn the bias into a tool: accept a mathematically second-best plan in exchange for one you will actually follow to the end.

2008–2009 is the strongest defense of the structure. Anyone holding two or three years of cash had no decision to make at that 677 trough: the cash bucket covered spending and the equities stayed put until the 2013 recovery, while those liquidating stock month by month locked in losses at the low permanently. Same market, same portfolio — the difference was only whether cash flow forced a sale. The output of this structure is not higher return; it is a lower probability of making an irreversible decision at the worst possible moment.

Cash drag is a real cost: an oversized bucket buys psychological comfort with certain inflation erosion. The refill rule is usually missing — most people define the buckets but never define when to sell what to top them up, so every refill becomes improvised market timing. The middle bucket is not always safe: bonds fell alongside equities in 2022, so the cushion was itself shrinking. And buckets are only an accounting device: what actually determines return is the real allocation of all the assets combined; if the aggregate exposure is unchanged, the only thing that changed is how you feel.

• How many years of net spending does my cash bucket cover — calculated, or guessed?
• Is the refill rule written down (trigger + what to sell + how much)?
• Adding all buckets together, is my real stock/bond split what I believe it is?

It is the hardest because three unknowns stack: returns, inflation and horizon — and the third barely exists in other investment problems. On U.S. actuarial tables, for a couple aged 65 the probability that at least one lives past 90 is roughly 50%, and past 95 roughly 20%. The only true hedge is mortality pooling: a life annuity can pay until you die because those who die early subsidize those who live long, and that "mortality credit" is something no portfolio can manufacture for itself. Yaari proved in 1965 that absent a bequest motive a rational agent should fully annuitize; almost nobody does, which is known as the "annuity puzzle." The workable compromise is floor plus upside: lifetime income covers rigid spending — which demotes the worst case from ruin to a cut in flexible spending.

The cautionary tale comes from the 1970s: a fixed nominal annuity is close to useless against inflation — U.S. CPI rose roughly 240% between 1970 and 1990, so a fixed-dollar annuity bought in 1970 retained only about a third of its purchasing power by 1990. It hedged longevity perfectly and inflation not at all. Which is why any floor must be interrogated: is it nominal or real? Japan makes the same point from the other side — decades of low inflation left nominal floors solid, meaning the identical instrument is not the same instrument under a different inflation regime.

Annuities are far from a cure-all. You surrender liquidity and optionality: the principal exchanged does not come back, which makes lump-sum needs such as a health shock awkward — precisely the quite-rational part of the "annuity puzzle." You also take on decades of issuer credit risk; pricing moves with interest rates, so buying lifetime income is brutally expensive when rates are low; and many markets lack inflation-linked products, so the theoretical cure cannot be bought. The inverse misuse exists too: compressing today's spending to guard against living too long — the risk is real, but living every year as a stress test is also a failure.

• Does my planning horizon use median lifespan, or the 90th percentile?
• Which expenses form the floor that must be covered for life, and how large is it?
• Is that floor covered by lifetime income, or sustained by selling down the portfolio year after year?
• Is the floor nominally fixed or inflation-adjusted? What survives ten years of 5% inflation?