A cash bucket doesn't beat a rebalanced portfolio. In the longest test anyone has run, it loses to one, and the loss grows with the size of the bucket.
The numbers come from Javier Estrada at IESE Business School, who ran bucket strategies against plain static allocations across 21 countries and 115 years of data. In the US sample, a static 60/40 split never once failed. The most popular bucket rule, holding two years of spending in bills, failed 4.7% of the time. Stretch the cash reserve to five years and the failure rate hits 27.9%.
That's the finding. What follows is why it happens, why a second study on different data agrees, and why the behavioural case for buckets still isn't dead.
What a bucket strategy promises
The structure is simple, and simplicity is most of its appeal. You split a retirement portfolio by time horizon rather than by asset class alone. Bucket one holds one to five years of spending in cash or short-dated government bills. Bucket two holds bonds. Bucket three holds equities and anything else volatile.
When shares fall, you spend from the cash bucket instead of selling equities at a depressed price. When shares recover, you refill the cash. The promise is that you never have to liquidate a fallen asset to eat.
It maps neatly onto how people already think about money. Mental accounting is one of the best-documented habits in household finance, and the bucket structure works with it rather than against it. The open question isn't whether it feels better. It's whether the comfort costs anything.
The 115-year test
Estrada's 2019 study used the Dimson-Marsh-Staunton database of real, local-currency stock and bill returns for 21 countries from 1900 to 2014. Each simulated retiree starts with $1,000, withdraws 4% in the first year, adjusts that amount for inflation each year after, and spends for 30 years. That yields 86 overlapping retirement periods per country, running from 1900-1929 through to 1985-2014.
He tested 11 static allocations, from 100% stocks down to 100% bills, each rebalanced annually. Against those he ran three bucket rules. All three park two years of inflation-adjusted withdrawals in bills. They differ only in when they judge equities safe to sell. The first spends from shares whenever last year's real return was positive. The second spends from shares only when last year beat the long-run mean. The third compares the trailing five-year mean to the long-run mean.
In the US, the static allocations won on every measure. A 60/40 portfolio never failed across the 86 periods. The 70/30 and 50/50 mixes failed 1.2% of the time, and in those rare failures they fell short by an average of one year. The best bucket rule failed 3.5% of the time, the other two 4.7%. When bucket strategies failed, they fell short by 3.3 to 5.8 years.
The size of the cash reserve mattered far more than the rule used to refill it. Under the simplest and most widely recommended rule, one year of cash produced a 2.3% failure rate, two years 4.7%, three years 9.3% and five years 27.9%. Estrada found the same pattern in all three rules: moving from a small reserve to five years of withdrawals pushed failure from under 10% to over 25%.
One honest qualification. Across the full 21-country average, the gap narrows a lot. Static 60/40 averaged a 30.7% failure rate, static 50/50 averaged 33.5%, and the three bucket rules averaged 30.1% to 31.6%. On raw failure counts, the buckets sat comfortably inside the range of static allocations. The static advantage in the global sample showed up in the softer measures instead: shorter shortfalls when things went wrong, and better risk-adjusted outcomes. The headline result is real, but it's sharpest in US data and mildest elsewhere.
Why the drag shows up at all
Estrada's explanation is mechanical rather than psychological. Bucket strategies move money in one direction only, from the aggressive bucket into the safe one. They avoid selling low. They never buy low.
A static allocation with periodic rebalancing does both halves of the job. It sells whatever has become relatively expensive and buys whatever has become relatively cheap, automatically, without a rule about when equities look safe. That's the whole mechanism, and it's the reason the bucket approach can't quite keep up. The comparison of calendar-based and threshold-based rules in annual rebalancing versus 5% threshold bands covers how much the timing of that trade actually matters.
There's a second cost, and it's the cash itself. Over the 1900-2014 window, US Treasury bills returned 0.9% a year after inflation against 6.5% for US equities. UK bills also returned 0.9% real, against 5.3% for UK equities. A five-year reserve on a 4% withdrawal rate parks roughly a fifth of the starting portfolio in the lowest-returning asset available, permanently. That isn't a rounding error over 30 years.
A different dataset, the same answer
Walter Woerheide and David Nanigian tested the same idea from a different angle in the Journal of Financial Planning in 2012. They used US returns from 1926 to 2009, overlapping horizons of 15 to 30 years, withdrawal rates from 3% to 9%, and money-market buffers of one, two, three and four years.
Their tally was lopsided. Across 560 comparisons, the buffer strategy produced a better success rate 35 times. The static allocation produced a better one 226 times. With a portfolio held entirely in stocks, the buffer never once beat the static approach, while the static approach beat it 51 times.
Buffer size drove the result again. A one-year buffer won 1 comparison out of 140 and lost 32. A four-year buffer won 13 and lost 77. Withdrawal rate mattered too: at a 3% rate the buffer won three comparisons and never lost, while at 8% and 9% it never produced a better result at all.
Their conclusion is blunt. Once equities enter the portfolio, or the withdrawal rate rises above 3%, a buffer of any size is more likely than not to leave the retiree worse off than a plain static allocation. What counts as a sustainable starting rate in the first place is a separate argument, and the 4% rule tested across 17 countries shows how much of it is US-specific.
Rebalancing was doing the work all along
Michael Kitces ran a narrower but sharper test in 2014. He took a 60/30/10 portfolio of equities, intermediate government bonds and Treasury bills, applied a 4% inflation-adjusted withdrawal, and started the retirement in 1966 — one of the worst US start dates on record.
He then applied a decision-rules bucket approach: spend from equities if they're up, from bonds if equities are down but bonds are up, from cash if both are down. Run alongside an ordinary total-return portfolio that simply rebalances to target each year, the two produced identical outcomes. Not similar. Identical.
The reason is that rebalancing already does what the bucket rules are trying to do. If equities have risen, rebalancing sells equities. If they've fallen, rebalancing buys them and funds withdrawals from whatever held up. The decision rules are a longhand version of an instruction the portfolio was already following.
Kitces also ran the bucket approach without any rebalancing, which is how most retirees actually implement it. That version ended the period with almost nothing left. The buckets weren't the active ingredient. The rebalancing was.
The counter-argument, which is stronger than it looks
Advocates don't really dispute the arithmetic. Their case is that the arithmetic assumes a retiree who holds the plan through a 40% drawdown, and that this retiree is partly a fiction. If a visible cash reserve is what lets someone own 60% equities instead of 25%, the drag buys an allocation they'd otherwise never have held.
Woerheide and Nanigian concede exactly this point in their own conclusion. They note that a buffer may be merited if it provides a psychological mechanism to induce clients to accept stock exposure. Their own data supports the size of that prize: buffer strategies beat a long-term bond portfolio 337 times to 30.
And capitulation isn't hypothetical. Elkind, Kaminski, Lo, Siah and Wong studied 653,455 brokerage accounts belonging to 298,556 US households, defining a panic sale as a 90% drop in a household's equity assets within a month, at least half of it from trades. Panic sales are rare in normal conditions, running at about 0.1% of investors, but they spike to roughly three times that baseline during sharp market falls. Of the households that panic sold, 30.9% never returned to risky assets at all. The median panic seller earned a zero to negative return afterwards.
The demographic detail is the part that should give bucket critics pause. Investors over 45, married, with more dependents, and those who rated their own investment knowledge as excellent were more likely to freak out, not less. That's a fair description of the median retiree.
Where the case runs thin is the link in the middle. Nobody has measured whether retirees using buckets actually hold more equities, or sell less in a crash, than comparable retirees using a total-return portfolio. The behavioural benefit is asserted from plausibility, not from a controlled comparison. So the cost side of this trade has been quantified across two datasets and a century of returns. The benefit side hasn't.
How big is the behaviour gap, really?
The standard number quoted in support of behavioural framing is Morningstar's investor return gap. Its 2025 study found the average dollar in US funds earned 7.0% a year over the decade to 31 December 2024, against 8.2% for the funds themselves — a 1.2 percentage point annual gap, about 15% of total returns.
Two things complicate that. First, Morningstar's own breakdown shows allocation funds had the narrowest gap of any category, with investors capturing 6.3% of a 6.5% total return, roughly 97%. Pre-mixed, automatically rebalanced portfolios already produce close to no gap, without any bucket structure.
Second, the figure itself is contested. Fulkerson, Jordan, Riley and Yan re-examined the same dataset in a 2026 Financial Analysts Journal paper and put the cost of bad timing at 0.10% a year — roughly a twelfth of Morningstar's estimate. The wider evidence on what mistimed trades cost is covered in what bad timing really costs investors. If the true behavioural cost is a tenth of a percent, a bucket structure that gives up a percentage point or more of return to prevent it is a bad trade.
What did beat sequence risk in the data
The problem buckets are built to solve is genuine. Vanguard's Khang and Clarke examined cohorts retiring into six US bear markets since 1926 with $500,000 in a 50/50 portfolio and a 5% fixed spending rule over 35 years. Retirees hit by the worst sequences were 31% more likely to outlive their wealth, had 11% lower income and left 37% smaller bequests. Among those who retired straight into a full bear market, 81% ran out of money.
What fixed it wasn't a cash reserve. It was adjusting spending. Applying a dynamic rule that capped annual increases at 5% real and allowed cuts of up to 2% eliminated portfolio depletion entirely for both groups. Even the retirees with the worst sequence of returns ended up with bequests 20% larger. The cost to the average full bear market retiree was an income shortfall of 4.5% across the first five years. The average ending balance for that group rose from 9% of the starting portfolio to 38%.
That's a much larger effect than anything the bucket literature produces, and it comes from the spending rule rather than the portfolio structure. The mechanics of why early losses hurt disproportionately are set out in how sequence risk turns identical returns into opposite outcomes, and the trade-off in flexible rules is quantified in what a 5.2% starting withdrawal rate costs under guardrails.
What would change the conclusion
Three findings would move this.
A study showing that bucket users hold materially more equity than matched total-return investors would flip the trade. If the structure reliably converts a 30% equity retiree into a 60% equity retiree, that swamps a drag of one or two percentage points. The evidence would have to be a real comparison of behaviour, not a survey of stated comfort.
A prolonged period of positive real cash returns would shrink the cost. Estrada's 0.9% real return on bills is a 115-year average that includes decades of financial repression. If short rates stayed meaningfully above inflation, the opportunity cost of a two-year reserve would be small enough to buy without argument.
A replication finding that Morningstar's 1.2 point gap survives the Financial Analysts Journal critique would strengthen the behavioural case considerably. That argument is live, and the 0.10% figure is new enough that it hasn't been tested by others yet.
Absent those, the evidence points one way. The bucket structure is a presentation layer. It's a useful one for anyone who genuinely won't hold equities without seeing next year's grocery money sitting in cash, and a one-year reserve costs almost nothing — 2.3% failure against 1.2% for the best static mix. Beyond that, each extra year of cash is paid for in outcomes. Whatever holdings report you use, including how far a 60/40 portfolio drifts in a single year or a tool like LedgerTouch, the number that matters is total equity exposure and how consistently it gets rebalanced, not which labelled pot the money sits in.