The 4% Rule in 2026: Safe Withdrawal Rates and Bengen’s 4.7% Update

This analysis draws on William Bengen’s research, published in the Journal of Financial Planning beginning in 1994 and in his books, research reports, and interviews since, and on Michael Kitces’s work on safe withdrawal rates at Kitces.com. We’d recommend reading them in full.

What Bengen Actually Tested

Bengen’s question was practical. How much could a client withdraw in the first year of retirement, increase each year by inflation, and still not run out for at least 30 years, no matter when retirement began? He used U.S. market data from 1926 on, with large-company stocks and intermediate-term Treasury bonds, and tested starting years from 1926 into the 1970s. With half in stocks and half in intermediate-term Treasuries, a 4% first-year withdrawal had lasted at least 33 years in every period he examined, and with 75% in stocks the shortest was 32 years. His 1996 follow-up put the 30-year figure at about 4.1%. In this research, a safe withdrawal rate is the highest starting rate that survived every period tested; Bengen later named it the SAFEMAX. He has since put the precise figure from that early two-asset work at 4.15%, set by a single retiree who started in October 1968, and he has said the 1994 paper never mentioned a rule: the name came later, from the press, and he has since adopted it as his own.

The rule is 4% only once. In the first year, a retiree with $1,000,000 withdraws $40,000. After that the withdrawal follows inflation, not the portfolio: at 3% inflation it would be $41,200 in the second year, about $53,757 in the eleventh, about $72,244 in the twenty-first, and about $94,263 in the thirtieth, whatever the markets have done to the balance in between. That is what makes the rule simple to follow and what makes the early years so important, since a falling balance does not lower the withdrawal.

Bengen also found that the mix mattered. In 1994 he advised holding as close to 75% in stocks as possible and never less than 50%, because portfolios with less in stocks ran out sooner, while more than 75% in stocks was to be avoided at the start of retirement.

The number is set by the worst periods, not the typical ones. Kitces has shown that across all the 30-year periods in U.S. history, the safe starting rate averaged about 6.5%, and in the best periods topped 10%. The low points came from retirements that began just before the Panic of 1907, the Great Depression, and the stagflation of the late 1960s and 1970s, and across those periods a balanced portfolio earned, on average, less than 1% a year after inflation over the first 15 years. A starting rate around 4% to 4.5% was low enough to survive them, depending on the data set. Bengen’s own current database, with several hundred retirement start dates since 1926 and a broader mix of investments, tells the same story: the average safe rate has been slightly higher than 7%.

We ran the same test ourselves on Robert Shiller’s public market data, which reaches back to 1871, for a 60/40 portfolio (60% stocks and 40% intermediate Treasury bonds) rebalanced each year, with each year’s withdrawals spread evenly through the year, as in Bengen’s book (we use quarterly installments). Across the 123 start years from 1871 to 1993, a 4% withdrawal survived every one. The average safe starting rate was about 6.7%, and more than half of the start years supported 6% or more. The years under 4.5% were 1906 and the retirements that began from 1964 through 1969, and the worst, a 1966 retirement, supported about 4.0% before any fees, right at the rule itself, a margin thin enough that details such as when in the year withdrawals are taken, or a cost of even a few hundredths of a percent a year, decide which side of 4% it lands on, which is what a number built for the worst case should look like. Our data and method differ a little from Bengen’s (a 60/40 mix with one start date a year, Shiller’s monthly average prices, and bond returns estimated from long-term yields), so our worst case lands just under his 4.15%, but the picture is the same: 4% sits at the floor of the historical record.

Two features of that design are easy to miss. “Safe” meant that a rate survived the worst sequence in this particular data set, not that it could not fail. And the answer depended on its inputs: a particular mix of investments, a 30-year horizon, a raise for inflation every single year, savings held in tax-deferred accounts with no tax counted on the withdrawals, no fees, and nothing required to be left at the end.

The Trinity Study and the Language of Success Rates

In 1998, three professors at Trinity University, Philip Cooley, Carl Hubbard, and Daniel Walz, approached the question another way. Using the S&P 500 and long-term high-grade corporate bonds from 1926 to 1995, they reported how often each withdrawal rate had lasted across rolling historical periods, rather than finding the single rate that survived them all. A 4% inflation-adjusted withdrawal from a portfolio split evenly between stocks and bonds lasted 30 years in 95% of those periods.

That shift, from the worst case to a success rate, shaped much of what followed. The percentage that planning software reports today follows the same idea, run on simulated markets instead of historical ones, as we explain in our article on the Monte Carlo probability of success. Bengen’s version asks what survived every period on record; the Trinity version asks how often a given rate survived. Both describe the past. Neither promises the next 30 years. Bengen has said as much himself: historical and simulated methods alike have “no predictive value whatsoever,” because the future can be completely different from what came before.

How the Number Has Moved: 4%, 4.5%, 4.7%

Bengen has revisited his own number more than once. His 2006 book put the figure at about 4.5% when taxes are left out and about 4.1% for a taxable account, counting the tax on the account’s growth. His 2025 book, A Richer Retirement, set the baseline at 4.7% by using seven asset classes rather than the two in his first work: five kinds of stocks (large, mid-size, small, and very small U.S. companies, plus international), intermediate-term government bonds, and Treasury bills, held at 55% stocks, 40% bonds, and 5% cash. Most of that rise came from the broader mix rather than from newer data: over the same start dates through 1993, the book shows two asset classes topping out at about 4.2% and seven at about 4.7%. In the book he describes techniques that effectively lift the figure to about 5%, and he has written that he does not expect much further expansion from there, since the gains are reaching the point of diminishing returns.

He treats 4.7% as the worst case in the historical record and has cautioned that it could fail if inflation stayed very high for a long stretch. In a December 2022 podcast interview, with inflation still running high, he suggested retirees might consider starting closer to 4.5% for the time being; since the book, he has said his method points to higher starting rates for many of today’s retirees.

What has not changed is the core of the method: find the starting withdrawal that survived the worst historical sequence, for a spending path that rises with inflation every year and never adjusts. His newer work adds a step on top of that, building on Kitces’s 2008 research linking stock valuations to safe withdrawal rates. Bengen sorts retirement start dates by inflation regime, meaning the average inflation over the first ten years (low, under 2.5%; moderate, 2.5% to 5%; high, above 5%), and within each regime fits a curve relating the safe rate to the Shiller CAPE, a measure of stock valuations. For a retirement starting now, he suggests estimating the regime from the past five years of inflation, while noting there is no scientific way to do it.

His tables, revised in early 2026 and built from fitted curves rather than historical minimums, put the starting rate at roughly 5.8% to 6.1% for CAPE readings of 30 to 40 when inflation is low or moderate; the CAPE stood near 41 in October 2026 (about 5.8% to 5.9% in those tables). Those figures keep every other assumption behind his 4.7%: 30 years, a tax-advantaged account, nothing left at the end, and spending that never adjusts. For a high inflation regime at those valuations he gives no figure at all, because no high-inflation start date in his data had a CAPE above about 22. He has written that he hopes retirees never face that combination of high valuations and returning inflation, since it “would be extremely uncomfortable for retirees.” Every caveat that applied to 4% in 1994 applies to 4.7% in 2026, and to the tables built on it.

What the Rule Assumes, and What Each Assumption Is Worth

The rule’s simplicity comes from holding many things fixed. Each of them moves the answer when it changes, and the effects stack. In one worked example in Bengen’s book, a retiree with a taxable account, a 40-year horizon, a goal of leaving $25,000 in nominal dollars for every $100,000 at the start, and 45% in stocks got a base starting rate of about 4.1%, even assuming 2% inflation and valuations near their long-run average.

A 30-year horizon. Kitces has summarized the research on other horizons. Bengen’s 1996 work found that stretching the horizon from 30 years to 45 lowered the safe rate from about 4.1% to about 3.5%, and that a 20-year horizon raised it to about 5%. David Blanchett’s 2007 research and Wade Pfau’s 2012 research found similar results, with about 3.3% to 3.5% for 40 years and about 5.3% to 5.5% for 20. Beyond about 40 years the rate stops falling much, because a withdrawal low enough to survive a bad first decade or two lasts long enough for the good years to arrive. Our own calculation on the same Shiller data traces the whole curve: the worst start year at each horizon supported about 4.9% for 20 years, 4.0% for 30, 3.7% for 40, and 3.5% for 50, so stretching the horizon from 30 years to 50 cost only about half a point, while the median start year supported about 8.0% for 20 years, 6.5% for 30, 5.3% for 40, and 4.9% for 50. The worst cases are not the typical ones: in our own calculation, a 4% withdrawal lasted 40 years in about 96% of start years and 50 years in about 86%, and in about two-thirds of start years it still left at least the starting amount, after inflation, even after 50 years.

For a couple retiring at 60, or one with long-lived parents, 30 years may be too short, which is why we lean toward setting the horizon at a fixed age adjusted for health and family history, and revisiting it as health changes.

Taxes. Bengen’s original research assumed all savings sat in tax-deferred accounts such as an IRA or 401(k), so 4% is a withdrawal before income tax. Those withdrawals are generally taxed as ordinary income, so if each were taxed at 22%, a 4% withdrawal would leave about 3.1% of the portfolio to spend. Kitces has pointed out that the drag is smaller in a taxable brokerage account, where tax falls only on the portfolio’s growth (interest, dividends, and realized gains), and that it is largest in a pre-tax account, where net spending falls by the full tax on every dollar. Which account pays first, and how much is converted to Roth along the way, changes the after-tax result, as we cover in our article on which account to draw from first and in our article on the gap years before Social Security and required distributions.

No costs. The research assumed no advisory or fund costs. Costs come out of the same returns that fund the withdrawals, so they lower the sustainable rate. In one 2010 analysis by Pfau, a portfolio of 60% stocks and 40% bonds paying 1% a year in costs had a worst historical 30-year rate of about 3.56%, from the retiree who started in 1966, compared with about 4.03% with no costs. How much costs matter depends on their size and on what they buy.

Spending that never changes shape. The rule raises spending with inflation every year for 30 years. Real spending tends to fall about 1% a year on average, as we explain in our article on how real spending changes through retirement, and Derek Tharp’s 2017 research at Kitces.com found that declining spending patterns lifted the historical safe starting rate by 0.32 to 0.75 percentage points. Bengen himself has taken the other side with his own clients, telling them not to count on spending falling late in life, because health costs rise and because gifts to family and charity belong in the plan as a regular expense. Which path fits depends on how much of a household’s budget is core and what it plans to give.

Spending on autopilot. The rule assumes spending holds its course through every market, without a review. Kitces has argued that at least some of the criticism of the rule’s inefficiency treats it as “a blind autopilot program,” which it was likely never intended to be, since Bengen was himself a planner who met regularly with his retired clients. His own practice bears that out. He has described starting clients near 4.2% and moving from there, closer to 5.5% for one expecting a large inheritance and closer to 3% for one whose pension had no inflation adjustment, and checking each plan against its targets at every annual review. A plan that agrees in advance on when spending and giving would rise or fall is the approach we lean toward, as we explain below.

One country’s history. Every figure above comes from the United States, which is the subject of the next section.

Outside the United States

The United States had an exceptional century for investors, and a rule built only on its history inherits that luck. Wade Pfau’s 2010 study of 17 developed countries since 1900 found that, even with the best mix of stocks and bonds chosen in hindsight, the SAFEMAX was above 4% in only 4 of them (the United States, at 4.02%, barely cleared it), and under 3% in Spain, Italy, Belgium, France, Germany, and Japan. A 2025 study by Aizhan Anarkulova, Scott Cederburg, Michael O’Doherty, and Richard Sias, drawing on a broad sample of developed markets and allowing for how long a couple might live, found that a 65-year-old couple willing to bear a 5% chance of running out could withdraw just 2.31% a year.

The counterarguments deserve a fair hearing. Those samples include countries that lived through wars fought on their own soil and bouts of hyperinflation, and a retiree in any of them could have softened the blow by owning other markets. A U.S. household today invests in a different world, with access to global diversification and inflation-protected bonds. Reasonable researchers disagree about how much weight U.S. history deserves. Even on the more hopeful reading, 4% is a product of one country’s past, not a law of finance.

What Is Different in 2026

Real yields. Treasury Inflation-Protected Securities, or TIPS, pay a fixed return above inflation, backed by the U.S. government. On October 28, 2026, real yields ran from 2.36% for five years to 3.10% for 30 years. That is a sharp change from late 2021, when a 10-year TIPS auction priced at a real yield of about negative 1.1%. One way to see what those yields mean for retirement income is a TIPS ladder: a set of bonds maturing one per year, so that each year’s principal and interest fund that year’s spending, rising with inflation, and the money runs out on schedule at the end.

We ran the numbers. At the October 28, 2026 yield curve, a 30-year ladder bought with $1,000,000 would pay about $48,711 a year in today’s dollars, starting immediately and rising with inflation, before taxes and costs. That is about 4.9% of the starting amount. At a flat 2% real yield the same ladder would pay about $43,774; at 1%, about $38,364; and at 0%, about $33,333. The 4% rule’s first-year withdrawal on the same $1,000,000 is $40,000.

That does not make the 4% rule obsolete, and a ladder is not a plan by itself. It ends at zero after 30 years, so it does nothing for a retirement that runs longer. Its payments rise only with inflation, and it leaves nothing for heirs or charity at the end. It tracks the consumer price index, which may rise more slowly than a retiree’s own basket of costs. In a taxable account, the inflation adjustments are taxed as they accrue, before any cash arrives. Bonds sold before they mature can lose value if real yields rise. And Treasury does not issue TIPS maturing in every year, so a real ladder bridges a few gap years in the late 2030s with nearby maturities, and its payout would differ somewhat from this calculation. What the comparison does show is the trade each approach makes. A withdrawal from a stock and bond portfolio takes market risk in exchange for the chance of more spending and giving in good decades and money left at the end; a ladder gives up that chance for a fixed schedule that runs out on a known date.

Valuations. Some research, Bengen’s newer work among it, suggests that, for a given inflation outlook, the starting rate should be lower when stock valuations are high than when they are low. We lean toward keeping valuations out of the withdrawal rate a plan starts from; at most, they inform some investment decisions at the margins, and a weak decade is handled through reviews and adjustments set in advance. The counterargument deserves a fair hearing: high starting valuations have tended to be followed by weaker returns over the next decade.

Inflation. The rule’s worst cases were inflationary ones. In Bengen’s book, the October 1968 retiree behind the 4.7% figure earned 10.9% a year over 30 years, more than a retiree who started in July 1989 (8.0%), but lived through inflation averaging 5.3% a year against 2.5%. After inflation the two earned nearly the same, yet the 1968 retiree’s safe rate was 4.7% and the 1989 retiree’s 7.1%; the difference lay in when the returns and the inflation arrived. Across 259 start dates from 1926 to 1993 in his book, the 30-year annualized return explained less than 2% of the differences in safe rates. We lean toward weighting inflation risk more heavily than many plans do, most of all for decisions that lock in fixed dollar amounts, and the inflation that matters to a household is its own basket of costs, not only the index. The case on the other side is real: the Federal Reserve targets 2%, inflation has stayed low for long stretches, and at positive real yields, inflation-protected bonds offer a direct hedge.

The Cost of a Number Built for the Worst Case

A number designed to survive the worst historical sequence is, in most other sequences, too low. As we explain in our guardrails article, in Kitces’s analysis of rolling 30-year periods back to the 1870s, a retiree following the 4% rule with a portfolio of 60% stocks and 40% bonds usually finished with more than the starting balance. Bengen has made the same point in his own words: a retiree who follows the rule should not be surprised to “accumulate considerable wealth,” because money that could have been spent earlier “will pile up and compound into a massive legacy.” For a household that wants that legacy, he adds, the rule fits. The trouble comes when the legacy is an accident rather than a choice.

For a household, that surplus has a cost that does not show up on a statement. It is the family trip that stayed on the someday list, or the gift that would have meant the most years before it finally arrived. In our experience, spending and giving too little is the more common problem.

An accidental legacy also tends to be a less efficient one. Pre-tax retirement accounts left to heirs are taxed as income when the money comes out, generally within ten years for most heirs other than a spouse and often during the heirs’ own peak earning years, as we explain in our article on how the 10-year rule works for inherited IRAs. And without current beneficiary designations and a will or trust, a probate court and state law can end up settling what the household never decided, the gap we describe in our article on why a will alone is not enough, and family members left to guess at intentions can end up at odds with one another.

So we lean toward giving that is intentional rather than incidental, and it starts with reflection and conversation. We want clients to think it through themselves and, if they are married, with each other; we help facilitate that thinking and turn it into a plan; and when it fits, the conversation extends to children or others they may want to give to. The questions are when, why, and to whom, both now and after they are gone. We like to see giving while living when it fits, because some gifts carry more value during life: a trip the household pays for and gets to take with its children and grandchildren; college help for a grandchild now, which can take real stress off the parents, rather than an inheritance they cannot plan around; or giving to a church or other causes as the household is able, rather than only through an estate. Others belong in the estate plan.

The conversation matters even for a household with no wish to leave money to family. Picture a retiree whose pension and Social Security already cover the bills, who is working to enjoy more of what the portfolio could provide, and who plans to leave nothing to children. Without forethought, more of what remains can be lost to taxes, probate costs, and disputes among relatives; with it, the money can go where the household chooses. A pre-tax account left to a charity, for example, owes no income tax on the balance. What matters is that each choice is purposeful rather than settled by default.

How We Think About It

We lean toward treating the 4% rule as a benchmark worth knowing well rather than a number to plan around. It is one of the most durable findings in retirement research: a 4% withdrawal, raised with inflation every year, came through the Great Depression, two world wars, and the inflation of the 1970s, and it still frames most conversations about retirement income. But a household’s sustainable spending depends on things the rule holds fixed: its other income, its taxes, its horizon, the way its spending is likely to change, and its willingness to adjust.

When someone asks for a ballpark before any modeling, the rough sketch we lean toward is about 4% of the portfolio for a plan that will never adjust and about 5.5% for one that will follow guardrails. That sketch starts a conversation; it is not a rule of thumb and never a plan. The higher figure is bought with a willingness to adjust: it assumes spending and giving will come down in a bad decade and go up in a good one. The real number for a household depends on its other income, its spending path, its legacy goals, and how much of its budget is core. The sketch sits below the figures in Bengen’s current tables because those figures come from fitted curves rather than the worst historical cases and assume inflation over the first decade stays low or moderate; the sketch makes no forecast of inflation, in keeping with the lean described above.

The plan itself starts somewhere else: with what the money is for, which is why every plan we build begins with a Statement of Financial Purpose, and with actual spending mapped against actual income sources, the approach we describe in our article on why no single number answers how much is enough. We lean toward basing the risk analysis on historical returns, treating history as a record of what has happened rather than a promise of what will, and testing the plan against bad starting years like the ones that set the 4% rule, a risk we explain in our article on why the order of returns matters.

What we lean toward, building on that history, is a plan that agrees in advance on when spending and giving would rise or fall, an approach built on the risk-based guardrails research Derek Tharp and Justin Fitzpatrick published at Kitces.com. The reasoning follows from the history above. A rule that never adjusts has to be set for the worst start year on record, about 4% in our data, so in most start years it leaves much of what could have been spent and given unused; the average start year supported about 6.7%. A plan that will adjust does not have to start at the worst case: its margin of safety sits in the adjustments it has agreed to in advance, in starting somewhat below what the plan expects the money can support, and in regular reviews and a planning horizon set with health and family history in mind.

In our plans, the guardrails come from that risk analysis rather than from withdrawal-rate bands. We ask how often, in the historical record, a given level of spending would have needed a cut or allowed a raise, and we show clients what they can spend now and the portfolio values at which a cut or a raise would happen, in dollars, before anything happens. We lean toward small, temporary cuts and full raises, toward protecting the core part of the budget as a floor (the core and adaptive split Michael Kitces has described), and toward revisiting the plan about twice a year. Our article on how retirement spending guardrails work explains the settings, and why risk-based lines called for much smaller cuts than withdrawal-rate rules in some of the hardest historical markets.

We lean toward this approach because we believe it serves most households better than a fixed rule: it spends the margin of safety only when a bad decade calls for it, instead of holding all of it back in every decade just in case. It is not free. It asks a household to accept that spending and giving may come down for a while in a bad stretch, and the more of a budget that is core spending, the more conservatively we lean toward starting. For a household that wants no adjustments at all, a lower fixed starting rate, or more income that does not depend on markets, can be the better fit, and that is a decision we make together, case by case.

Common Questions

Is the 4% rule still valid in 2026?

As a description of U.S. history, yes: a 4% starting withdrawal, raised with inflation every year, survived every 30-year period William Bengen tested in his 1994 research, and our own calculation on market data back to 1871 agrees, with the worst start year, 1966, supporting just about 4% before fees. Bengen now puts the worst-case figure at 4.7% with a broader mix of investments. As a plan, it was never meant to stand alone. It leaves out taxes, costs, horizons other than 30 years, changing spending, and any adjustment along the way, and research outside the United States has found lower safe rates. At October 2026 real yields, a 30-year ladder of inflation-protected Treasury bonds could pay about 4.9% of the starting amount a year, rising with inflation, though it ends at zero after 30 years and leaves nothing for heirs or charity. We lean toward a plan that agrees in advance on when spending and giving would rise or fall, built on risk-based guardrails, rather than a fixed rule.

What is Bill Bengen’s new safe withdrawal rate?

In his 2025 book, A Richer Retirement, William Bengen set his baseline at 4.7% for a 30-year retirement, using seven asset classes, including small company U.S. stocks, rather than the few in his earlier work. He treats 4.7% as the worst case in the historical record, describes techniques in the book that effectively lift it to about 5%, and has cautioned that it could fail if inflation stayed very high for a long stretch. His current tables, revised in early 2026 and built from fitted curves rather than historical minimums, put the starting rate at about 5.8% to 5.9% at stock valuations like October 2026’s, for the same 30-year setup, if inflation over the first ten years stays under 5%. They give no figure if inflation runs higher, and no one knows in advance which it will be. His 2006 book had put the figure at about 4.5% when taxes are left out and about 4.1% for a taxable account.

Does the 4% rule account for taxes?

No. The 4% is a withdrawal before income tax. Bengen’s 1994 research assumed all savings sat in tax-deferred accounts and counted no tax on what came out. His later work, which counted the tax on a taxable account’s growth, put that account at about 4.1%, against about 4.5% with taxes left out. From a traditional IRA or 401(k), withdrawals are generally taxed as ordinary income: if each were taxed at 22%, a 4% withdrawal would leave about 3.1% of the portfolio to spend. In a taxable brokerage account the drag is smaller, because tax falls only on the portfolio’s growth, and the order in which accounts are drawn down changes the result.

What withdrawal rate works for a 40-year retirement?

Research on U.S. markets puts the safe starting rate for 40 years at about 3.3% to 3.5%, compared with about 4% to 4.1% for 30 years. Bengen’s 1996 research found about 3.5% for 45 years, and our own calculation on Robert Shiller’s public market data puts the worst historical 40-year case at about 3.7% for a portfolio of 60% stocks and 40% bonds, with the median start year at about 5.3%. The 3.7% is a worst case: in the same calculation, a 4% withdrawal lasted 40 years in about 96% of start years. Michael Kitces has noted that the rate stops falling much beyond about 40 years, because a withdrawal low enough to survive a bad first decade or two tends to last. A horizon that long also makes flexibility more valuable, since a plan that can adjust has more years in which to do it.