Key takeaways
- Guyton and Klinger's 2006 paper concluded that initial withdrawal rates of 5.2%-5.6% were sustainable over a 40-year retirement at their 99% confidence standard, for portfolios holding at least 65% equities. Drop to 50% equities and the figure falls to as low as 4.6%.
- The rule that buys the higher rate is precise: when the current year's withdrawal rate has risen more than 20% above the initial rate, that year's withdrawal is cut by 10%, and the reduced amount becomes the base for the following year.
- Morningstar's 2025 modelling prices the same trade on a 40% equity / 60% bond portfolio: 3.9% starting rate for rigid inflation-adjusted spending, 5.2% for the guardrails. The rigid plan's year-30 withdrawals have a standard deviation of 0.00%; the guardrails plan's is 28.86%.
- In historical backtests run by Derek Tharp and Justin Fitzpatrick on a 60/40 portfolio starting at 4.3%, guardrails retirees who began in 1936, 1965, 1999 and 2007 ended up spending 45%, 54%, 36% and 28% below their original plan.
- In Vanguard's own 2026 worked example, a retiree's $65,000 real spending target splits into $50,000 of essential expenses and $15,000 of discretionary ones. The flexibility a higher rate assumes has to come out of the smaller number.
The answer is worth about 1.3 percentage points, and the invoice arrives later
If you're willing to cut your spending in bad years, how much more can you withdraw at the start? The studies that have actually modelled it put the gap at roughly one to one and a half percentage points.
Morningstar's State of Retirement Income: 2025, published on 3 December 2025, runs both cases through the same simulation engine. A retiree who wants a fixed inflation-adjusted income from a 40% equity / 60% bond portfolio, with a 90% chance of money still being there after 30 years, can start at 3.9%. The same retiree following Guyton and Klinger's guardrails can start at 5.2%. On a $1 million portfolio that is $39,000 against $52,000 in the first year.
The extra $13,000 is real. It is not, however, extra money. It is a claim against your own later spending, and the terms are written into the rule. In the same simulation, the rigid plan's year-30 withdrawal has a standard deviation across successful trials of 0.00% — by construction it never moves. The guardrails plan's is 28.86%.
What Guyton and Klinger actually wrote: four rules, one 20% trigger, one 10% step
Jonathan Guyton and William Klinger published Decision Rules and Maximum Initial Withdrawal Rates in the Journal of Financial Planning in March 2006, building on Guyton's 2004 paper. That earlier paper had already pushed the maximum safe initial rate to 5.8%-6.2% over 1973-2003, depending on the equity weight. Four rules do the work, and only two of them are the guardrails.
The portfolio management rule governs where the cash comes from. After a year in which an asset class gained enough to exceed its target weight, the excess is sold into cash. No withdrawal is taken from any equity following a year with a negative return if cash and bonds can cover it.
The withdrawal rule freezes raises. In the 2006 version, withdrawals rise with inflation except after a year of negative total return and when that year's withdrawal rate would exceed the initial rate. There is no make-up for a missed increase — the money is gone, permanently, from every year that follows. Guyton and Klinger note that this modified version produces about 60% fewer freezes than the 2004 original, which froze after any losing year.
Then the two guardrails. The capital preservation rule applies when the current year's withdrawal rate has risen more than 20% above the initial withdrawal rate. A 5% start puts the trigger at 6%. When it fires, that year's withdrawal is reduced by 10%, and the reduced figure becomes the base for the next year's calculation. The rule expires 15 years before the maximum age the retiree is planning to: someone planning to 100 stops applying it at 85.
The prosperity rule is its mirror. In any year the withdrawal rate sits more than 20% below the initial rate, the withdrawal is increased by 10%, and that higher figure becomes the new base.
Guyton and Klinger tested other trigger points and found tighter ones simply produced more cuts without improving survival, while looser ones let portfolios drift past saving. Cuts deeper than 10% — they tried 15% and 20% — didn't noticeably improve the probability of success.
The 5.2%-5.6% headline is attached to a 40-year horizon and a definition of "confidence" most people have never read
The paper's conclusion is quotable and almost always quoted without its conditions: "For portfolios containing at least 65 percent equities, initial withdrawal rates of 5.2-5.6 percent are sustainable over a 40-year period at the 99 percent confidence standard and rise to 5.7-6.2 percent at the 95 percent confidence standard."
Four conditions are doing heavy lifting there. The horizon is 40 years, not 30, and it is fixed rather than modelled from mortality. The equity weight is at least 65%. At 50% equities the maximum rate drops to as low as 4.6%, because a bond-heavy portfolio struggles to keep pace with inflation over four decades. The engine is Monte Carlo, 14,000 simulated lifetimes per scenario, with returns drawn from two historical windows — 1973-2004 for a six-asset-class equity sleeve, and 1928-2004 for an S&P 500 sleeve.
And the "99 percent confidence standard" is not just a success rate. Guyton and Klinger define it as an initial rate where both the probability of success and the median purchasing power maintained are at least 99%. That second half matters, because the rules can always keep a portfolio alive by shrinking what it pays out.
Withdrawal reductions from the capital preservation rule occurred "no more than 10 percent of the time at very high probabilities of success." That is a ceiling rather than a typical case: at most about four cuts in a 40-year retirement.
Morningstar prices the same trade every year, and the 2025 table is unusually blunt about it
Guyton and Klinger's paper is 20 years old and runs on 20th-century return assumptions. Morningstar's annual study is the closest thing to a refreshed version: forward-looking capital markets assumptions, 1,000 Monte Carlo paths, 2.46% expected inflation, a 30-year horizon and a 90% success threshold. In 2025 it tested nine spending systems side by side on a 40/60 portfolio.
| Spending system | Starting safe rate (%) | What a bad year does to spending | Year-30 cash flow SD (%) | Median lifetime spending ($m) | Median year-30 balance ($m) |
|---|---|---|---|---|---|
| Fixed real withdrawals (base case) | 3.90 | Nothing. The rule forbids a cut. | 0.00 | 1.17 | 1.42 |
| Skip the inflation rise after a losing year | 4.30 | Loses one year's inflation, permanently | 5.48 | 1.20 | 1.28 |
| Scheduled 2% real decline each year | 5.00 | Nothing extra. The decline is planned, not market-linked. | 0.00 | 1.14 | 1.34 |
| Probability-based guardrails | 5.10 | -10% when modelled success falls to 75% | 16.21 | 1.55 | 0.23 |
| Vanguard dynamic spending | 5.10 | Capped at -2.5% in any one year | 36.38 | 1.28 | 0.88 |
| Guyton-Klinger guardrails | 5.20 | -10% when the withdrawal rate runs 20% above its starting level | 28.86 | 1.36 | 0.70 |
| RMD (balance divided by life expectancy) | 4.72 | Falls with the portfolio, no floor | 43.90 | 1.50 | 0.12 |
| Constant percentage of balance | 5.70 | Falls with the portfolio, floored at 90% of year one | 35.01 | 1.25 | 0.92 |
| Endowment method (10-year average balance) | 5.70 | Same floor, smoothed over 10 years | 38.78 | 1.24 | 0.89 |
Read down the first two columns of that table, or across the chart above, and the pattern is monotonous: every method that raises the starting rate raises cash flow volatility with it, and the two that reach 5.70% are the two that hand your standard of living directly to the market. Morningstar's own summary is that "there's no free lunch."
The cuts in the historical record are far deeper than one 10% step
A 10% cut sounds survivable. The trouble is that the rule can fire repeatedly, and each cut resets the base, so the reductions compound.
Derek Tharp and Justin Fitzpatrick ran the Guyton-Klinger rules through historical retirements in March 2024, using a $1 million 60/40 portfolio with a $43,000 starting withdrawal. That is a 4.3% rate, calibrated to a 90% historical success probability, with the cut trigger therefore sitting at 5.16%. Their results: a 1936 retiree was driven 45% below the original real spending plan, a 1965 retiree 54% below, a 1999 retiree 36% below, and a 2007 retiree 28% below. They point separately to Karsten Jeske of EarlyRetirementNow, who ran a 1966 retiree starting at 4%, or $40,000 a year on $1 million. Jeske's withdrawals bottomed at $16,400, a 59% reduction.
They also cite Wade Pfau's 2015 Monte Carlo assessment of the same rules. At a 4.8% starting rate, even the median path involved a 20% pay cut by year 10, a 32% pay cut by year 20, and spending 36% below plan by year 30. At a 5.3% starting rate there was a 10% chance of an 84% pay cut within 30 years.
Their conclusion is the strongest criticism in the literature: the rules carry "much higher risk of dramatic cuts in spending than many advisors realize." Nobody ran out of money in those scenarios — the rules did exactly what they were designed to do. They kept the portfolio alive by shrinking the retiree.
The objection that matters: a cut you cannot make is not flexibility
Here is the case against the whole idea, and it is a good one. Every study above treats a 10% reduction in portfolio withdrawals as a lever the retiree can simply pull. For a large number of households it is not.
In Vanguard's Principles for Retirement Income, published in 2026 with capital markets simulations run on 28 February 2026, the illustrative retiree is a 67-year-old with $1 million across a 401(k) and a taxable account, $21,080 a year of Social Security, and a real spending target of $65,000 — of which $50,000 is essential and $15,000 discretionary. Her portfolio is funding $43,920 of that. A 10% guardrails cut takes about $4,392 out of a discretionary budget of $15,000. That is survivable, and it's roughly the best case: a retiree with a $1 million portfolio, a Social Security cheque, and 23% of her target spending classified as optional.
Move down the wealth distribution and the arithmetic inverts. Housing, food, insurance and medical care don't respond to a Monte Carlo simulation. David Blanchett's analysis of the 2011 Consumer Expenditure Survey found that medical costs alone run from roughly 10% of total expenditures for a 65-year-old household to roughly 20% by age 85. The share of the budget that is least negotiable grows precisely as the retiree ages into the years when the guardrails have already done their compounding.
Guyton and Klinger saw this coming and said so. Their paper explicitly refuses to assume real income needs decline: "real-life client experiences, coupled with dramatically rising health care costs and uncertainty about possible reductions in Social Security benefits, make us unwilling to consider the possibility of such natural reductions in a retiree's income needs."
Morningstar makes the same concession from the other direction. Its report states that dynamic strategies "will be most agreeable for retirees who have a healthy share of their necessary living expenses coming from nonportfolio sources of income like Social Security and a pension." The report then quantifies it: layering the $36,000 Social Security benefit from its own examples on top of a guardrails portfolio cuts the cash flow standard deviation to roughly half of the portfolio-only figure. In its worked example, a 30% portfolio crash forces a 10.0% cut to portfolio withdrawals but only a 3.5% fall in total household income, because Social Security does not take a haircut.
Real retiree spending does fall, but at about 1% a year, not 10% on demand
There's a tempting move at this point: if spending naturally declines in retirement, the cuts take care of themselves.
Blanchett's Exploring the Retirement Consumption Puzzle matched RAND Health and Retirement Study data to five waves of the Consumption and Activities Mail Survey from 2001 to 2009. After filtering, 591 households remained. The average real change in retiree spending from age 60 to age 90 was -0.96% a year, statistically significant with a t statistic of -4.31. Blanchett also documented the shape: a "retirement spending smile", with real spending falling fastest in the middle of retirement and flattening at both ends, the late-life flattening driven by medical costs.
About 1% a year of gentle real decline is a genuine tailwind. But it is a slow drift you can plan around, arriving on a schedule set by ageing. A guardrails cut is a 10% step, arriving on a schedule set by the market, most likely in the years when your portfolio has just fallen. The two are not substitutes, and a household that has already banked the natural decline into its plan has less room left to give.
The upside rule is half the system and almost nobody quotes it
Guyton and Klinger were explicit that the prosperity rule matters as much as the cuts. In their tests it "generates significantly more raises than the cuts from the capital preservation rule". At several initial withdrawal rates the raises pushed purchasing power maintained above 100% of the original plan. A system that only cuts is not the system they published.
Michael Kitces built a version that only ratchets upward. His 2015 rule starts at 4% on a 60/40 portfolio over 30 years, raises real spending by 10% whenever the portfolio has grown 50% above its starting value. It permits that raise no more than once every three years. Spending never falls. In his historical testing no scenario ran out of money and none required a cut, while the most favourable paths supported roughly double the inflation-adjusted spending of a static 4%.
A ratchet captures the upside without asking the retiree to absorb the downside, which is why its starting rate stays at 4% rather than 5.2%.
What this evidence cannot tell you
None of it is a forecast. Guyton and Klinger's returns are drawn from 1973-2004 and 1928-2004 US data. Their own "what if" section asks what happens if future real returns don't resemble those distributions. It answers only that the key trade-off would shift to purchasing power rather than success rate. Morningstar's numbers are not history at all — they are 1,000 simulated paths built on Morningstar's own capital markets assumptions, which are a forecast like anyone else's.
Everything here is US-centric: US equities, US bonds, US inflation, US Social Security. Whether these rates travel is a separate question, taken up in our piece on safe withdrawal rates outside the US. The guardrails were also designed around a specific portfolio — at least 65% equities in the 2006 paper. Morningstar found the guardrails methods worked best at higher equity weights than the other systems it tested. Applying them to a bond-heavy portfolio is not the tested case.
Fees and taxes are absent throughout. Morningstar states plainly that its research "doesn't incorporate the impact of expenses or taxes". It shows that 1% of annual expenses on a tax-deferred 60/40 portfolio drops its base case from 3.9% to 3.4%. There is no reason to think the flexible rates are immune to the same drag.
The largest gap is behavioural. Every figure in this piece assumes the retiree makes the cut when the rule says to. No study cited here tested whether people actually do. The rules are a model of a disciplined household, and the discipline is the untested part.
What would change the conclusion
If the flexibility isn't real, the higher rate isn't either. This is the load-bearing assumption. A household whose portfolio withdrawals fund rent, food, insurance and medical care cannot execute a 10% cut on demand, and the 5.2% figure quietly reverts to something closer to 3.9%. The test is not your risk tolerance. It is the size of your discretionary budget after guaranteed income has covered the floor.
If bond yields fall back. Morningstar's base case has moved from 3.3% in 2021 to 3.8%, 4.0%, 3.7% and now 3.9%, tracking valuations and yields. The gap between rigid and flexible is more stable than either endpoint, but both endpoints move.
If the horizon isn't 30 or 40 years. Morningstar's base case falls from 3.9% to 3.5% when the horizon stretches from 30 years to 35, and to 3.3% at 40. Guyton and Klinger's headline is a 40-year number, which is why it isn't directly comparable to a 30-year 4% rule.
If you're prepared to schedule the decline instead of reacting to it. The 5.00% row in the table gets most of the benefit of flexibility with none of the market-linked volatility, at the cost of committing in advance to spending less in real terms each year. Whether that is a better trade depends on something no simulation can price: whether you would rather your spending be decided by your age or by the market.
The number worth watching isn't the headline rate. It's your current withdrawal rate against the one you started with. That is the ratio that every guardrail in the literature is actually built on, and the one that tells you how close to a trigger you already are. A tracker such as LedgerTouch will show you that ratio continuously; so will a spreadsheet updated once a year. What neither can tell you is whether the cut it implies is one your household could absorb.