The 4% Rule: What It Says and Whether It Still Holds

It has three parts, and the second is the one most often lost. First, in year one, withdraw 4% of the portfolio's value on the day you retire. Second, in every later year, withdraw the same dollar amount as the year before, increased by that year's inflation. The portfolio's value plays no further part. Third, the research behind the rule found that this pattern would have lasted at least 30 years in every period tested.

On $1,000,000 the first withdrawal is $40,000. If inflation runs at 3%, the schedule is fixed from that day:

This is why the rule is sometimes called constant-dollar or constant real spending. It was designed to answer one question: how much can a retiree spend and keep the same standard of living for three decades without knowing in advance what markets will do? The price of that steadiness is that the withdrawal can drift a long way from 4% of what is actually in the account. A retiree of January 1966 following the rule had $493,000 of an original $1,000,000 left after ten years, in real terms, and the same real $40,000 was by then 8.1% of the balance. A retiree of January 1982 had $2.22 million after ten years, and the same $40,000 was 1.8%. The rule told both to carry on unchanged.

Bengen, 1994

William Bengen was a financial planner in El Cajon, California, when he published “Determining Withdrawal Rates Using Historical Data” in the Journal of Financial Planning in October 1994 (reprinted by the journal in 2004). Planners at the time often reasoned from averages: if a balanced portfolio earns about 5% above inflation, a client can spend 5%. Bengen showed that this fails, because actual returns and inflation arrive in an order, and a bad order early in retirement can empty a portfolio that the averages say is safe. That is sequence of returns risk, and his paper is one of the first careful treatments of it.

His method was to replay history. Using Ibbotson Associates' data on common stocks, intermediate-term Treasury notes and inflation, he started a hypothetical retirement on January 1 of each year from 1926 to 1976, took a first-year withdrawal as a percentage of the starting balance, raised it with actual inflation each year, and counted how many years the money lasted. His base portfolio was 50% stocks and 50% Treasuries, rebalanced. He then repeated the exercise for portfolios holding 0%, 25%, 50%, 75% and 100% in stocks at first-year withdrawals from 1% to 8%.

The findings, in his own terms:

• A 3% first-year withdrawal lasted at least 50 years in every case.
• A 4% first-year withdrawal “should be safe”: in no past case had it exhausted a portfolio in fewer than 33 years.
• A 4.25% withdrawal could exhaust a portfolio in as little as 28 years. He called 5% risky and 6% or more “gambling.”
• The retirements that did worst began in the late 1960s and early 1970s, because the 1973 to 1974 market fall came with high inflation.
• Stock allocations below 50% and above 75% were counterproductive. He advised holding as close to 75% in stocks as the client could accept and no less than 50%.

The 1994 paper does not use the phrase “4% rule,” and it does not contain the figure 4.15%. In later work Bengen named the highest withdrawal rate that survived the worst historical case the SAFEMAX, and for his original stocks-and-Treasuries portfolio that worst-case figure was 4.15% (Wade Pfau, Retirement Researcher). The round 4% is the planner's version of that number.

The Trinity study, 1998

Four years later Philip Cooley, Carl Hubbard and Daniel Walz, finance professors at Trinity University, published “Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable” in the AAII Journal of February 1998 (AAII). It asked the question differently. Where Bengen looked for the single worst case, the Trinity authors reported a portfolio success rate: the percentage of all past payout periods that a portfolio supported, where success meant ending the period with a value above zero.

They used annual returns from 1926 to 1995 for the S&P 500 and long-term high-grade corporate bonds, and tested five mixes (100% stocks, 75/25, 50/50, 25/75 and 100% bonds), payout periods of 15, 20, 25 and 30 years, and withdrawal rates from 3% to 12% of the starting value, both as fixed dollar amounts and adjusted for inflation. Taxes and transaction costs were left out. For inflation-adjusted withdrawals they found that rates of 3% to 4% continued to produce high success rates for stock-dominated portfolios, and concluded that most retirees would benefit from holding at least 50% in stocks. The authors updated the tables with data through 2009 in the Journal of Financial Planning in 2011 (Portfolio Success Rates: Where to Draw the Line).

The two studies are often cited as one. They agree on the neighbourhood of 4% and differ in what the number means. Bengen's is a floor: nothing in the record did worse. Trinity's is a frequency: this share of historical periods survived. Note also what Trinity counted as success. A portfolio that ended thirty years with a single dollar passed, and one that ended with several times its starting value passed equally.

Every one of these is built into the number. Change one and the number changes.

• Thirty years. Bengen set a minimum of 30 years of portfolio life and had clients aged 60 to 65 in mind. Trinity tested nothing longer than 30. A 40-year retirement, or a 20-year one, has a different answer.
• US market history. Both studies used US returns from 1926 onward, one country over one period. The worst case in that record is treated as the worst case there is.
• Roughly 50% to 75% in stocks, held throughout. The rule does not hold for a portfolio of bonds and cash, and it assumes the retiree keeps the allocation and rebalances through every decline.
• No fees. Index returns have no fund expenses or advisory charges. Whatever a portfolio costs each year comes out of the withdrawal it can support.
• No taxes. Bengen assumed a tax-deferred account and did not deduct the income tax due on withdrawals. The 4% is a gross figure.
• Rigid spending. The retiree spends the same real amount in the year after a crash as in the year after a boom, and never spends more after good decades. Few households behave that way in either direction.
• Success means not running out. Nothing in the rule aims to leave an estate or to spend the money down. In most historical periods a 4% withdrawal left a large balance at the end.

The arithmetic is the easy part. The table shows the first-year withdrawal at 3%, 4% and 5% of the starting balance. All figures are before tax, and each would rise with inflation in later years.

Run backwards, the rule gives the familiar multiple: a portfolio of 25 times the annual amount it has to provide. At 3% the multiple is about 33, and at 5% it is 20. The difference between those is the whole debate about the rule, expressed in years of saving. Our guide to how much you need to retire works through the target, and the coast FIRE calculator uses the same multiple for people still saving.

We reran the test with a public dataset and our own drawdown engine. The returns are Aswath Damodaran's annual US figures from NYU Stern, after inflation (Historical Returns on Stocks, Bonds and Bills, 1928 to 2025). The portfolio is 60 / 35 / 5 in the S&P 500, 10-year Treasury bonds and Treasury bills, rebalanced yearly, with no fees or taxes. For each retirement beginning from 1928 to 1996, the 69 start years with a full 30 years of data, we found the highest starting withdrawal rate that would have lasted the full 30 years, with the withdrawal taken at the start of each year.

Three things stand out. The typical retirement could have sustained far more than 4%: the median start year supported 6.47%, and the best, 1982, supported 10.05%. The rule is set by the worst few years, not the usual ones. Those worst years are bunched together in the mid to late 1960s, when retirees met a long inflation before the 1973 to 1974 fall. And in this version of the test 4% is not quite a floor: 4 of the 69 start years (1965, 1966, 1968 and 1969) fell short of it, the lowest being 1966 at 3.76%. A 50/50 portfolio of stocks and 10-year Treasuries does slightly worse, with a low of 3.66% and 5 start years below 4%.

The honest answer is that the research has not moved far from it, in either direction, and that the people who study it most treat the number as a starting estimate, not a rule.

Morningstar: 3.9% for 2026

Morningstar publishes a yearly study that estimates the starting rate for a new retiree from forward-looking return and inflation forecasts, not from history. The current edition, published on December 3, 2025, puts the highest safe starting withdrawal rate at 3.9% for a retiree who wants a consistent level of inflation-adjusted spending, with a 90% probability of having money left at the end of a 30-year retirement. That figure applies to portfolios holding between 30% and 50% in stocks, and it excludes Social Security and other income (Morningstar). As of November 14, 2026, it is the latest edition.

The same figure in earlier editions shows how much it moves with market conditions: 3.3% in 2021, 3.8% in 2022, 4.0% in 2023 and 3.7% in 2024. Morningstar is explicit that these are estimates for people retiring now and are not a signal for existing retirees to change their spending each year. The study also found that retirees willing to tolerate some fluctuation in spending could start at nearly 6%.

Bengen: 4.5%, then 4.7%

Bengen has revised his own figure upward as he added asset classes to the test portfolio. Adding small-company stocks raised the historical worst case from about 4% to 4.5%. His 2025 book, A Richer Retirement: Supercharging the 4% Rule to Spend More and Enjoy More, adds micro-cap, mid-cap and international stocks and Treasury bills, and arrives at 4.7%. That is still a worst case: it is the rate that survived for a retiree of October 1968. Across all the retirement dates in the book the figure ranged from 4.7% up to 16.2%, for a retiree of July 1932, and averaged 7.1% (Morningstar interview with Bengen, May 2026).

In the same interview Bengen attaches conditions that are easy to miss in a headline. The figures are for tax-deferred accounts, and he puts the rate for a taxable portfolio perhaps 10% to 20% lower. He regards a 30-year horizon as too short for many people and suggests adding a margin of safety to life expectancy. And he names inflation as the greatest threat to a retiree's withdrawals, which is what the late 1960s in the chart above show.

Why the two differ

Morningstar's 3.9% and Bengen's 4.7% are answers to different questions. One is a forecast-based estimate with a 90% success target for a portfolio with at most half in stocks. The other is the historical worst case for a portfolio spread across seven asset classes. A reader choosing between them is choosing between assumptions about future returns, not between a cautious expert and a generous one. Both sit within a point of the original figure, three decades on.

History offers only a handful of independent 30-year periods, so the other way to test a withdrawal rate is simulation. Our Monte Carlo retirement calculator runs 10,000 paths of yearly returns. Its opening assumptions are a 60 / 35 / 5 portfolio of stocks, bonds and cash with an expected real return of 5.8% a year and volatility of 12.3%, both taken from the Damodaran averages for 1928 to 2025, less a fee of 0.25%. Spending is fixed in today's dollars, as the rule requires. The table shows the share of paths in which the money lasted.

At 30 years, 4% lasts in 90.5% of paths, which is the result the calculator opens on. The row is the rule; the columns either side show how much it depends on the horizon. Over 20 years the same 4% lasts in 99.1% of paths and even 5% lasts in 93.5%. Over 40 years 4% lasts in 80.9%, and a 3% withdrawal lasts in 95.1%.

Why longer retirements need less and shorter ones allow more

A withdrawal plan is partly living on returns and partly spending down capital. Over a short horizon most of the money comes from capital: with no return at all, a portfolio lasts 20 years at 5%. Over a long horizon the plan depends on returns compounding for decades, so there is more time for a poor stretch to do damage and less capital to spare for each year. The table below turns the question around and asks what starting rate 90% of simulated paths survive for each length of retirement.

On these assumptions the 90% rate for 30 years is 4.00%, close to Morningstar's 3.9% although the inputs differ: ours are historical averages with 60% in stocks, theirs are forecasts with less. Someone retiring at 55 and planning to 95 is looking at 3.40%. Someone of 75 planning to 95 is looking at 5.35%, which is one reason a rate set at 65 need not hold for life. Lower the expected return by a point and every figure falls; the calculator lets you rerun the table with your own inputs.

The 4% is before tax

The rule produces a gross withdrawal. What can be spent depends on the account it comes from. A withdrawal from a traditional IRA or 401(k) is ordinary income in full. A qualified Roth withdrawal is tax-free. A sale in a taxable account is taxed only on the gain, usually at capital gains rates, though the account has been paying tax on dividends along the way.

To show the size of it, take the simplest case: a married couple, both 65 or older, whose only income for 2026 is a 4% withdrawal from traditional IRAs. Federal income tax is computed with this site's 2026 tables: the standard deduction, the additional deductions for people 65 and older, and the 2026 brackets (IRS).

The tax is modest on smaller balances and grows faster than the withdrawal, because more of it reaches higher brackets: the after-tax rate is 4.00% of a $1,000,000 IRA and 3.51% of a $5,000,000 one. A budget drawn up in after-tax dollars has to be grossed up before it is compared with 4%. Which account to draw from, and in what order, is its own subject: see our guides to retirement withdrawal order and taxes in retirement.

Fees come out of the same 4%

The original studies used index returns with no costs. Every cost a portfolio carries, whether fund expenses, platform charges or advice, reduces the return that funds the withdrawal. In the simulation, the 30-year success rate of a 4% withdrawal moves with the annual cost as follows.

The practical point is only that the cost belongs in the calculation. A withdrawal rate quoted from research is net of nothing.

From age 73, the IRS requires a minimum withdrawal each year from traditional IRAs and most workplace plans. The amount is the account balance at the end of the previous year divided by a distribution period from the Uniform Lifetime Table. Roth IRAs have no required distributions while the owner is alive (IRS). Dividing 1 by the table's factor turns each age into a percentage of the balance.

The RMD starts just under 4%, at 3.77%, passes it at 75 and keeps climbing: 4.95% at 80, 8.20% at 90. Two points follow. An RMD larger than the 4% rule's amount does not have to be spent. It has to be withdrawn and taxed, and whatever is not needed can be reinvested in a taxable account. And the comparison is between unlike things. The rule fixes a dollar amount at retirement. The RMD is recalculated every year from whatever the account is then worth, so it falls after a bad market and rises after a good one. Our RMD calculator projects the amounts for your own balance and age.

The 4% rule keeps spending steady and lets the portfolio absorb everything. Each alternative moves some of the adjustment onto spending. What each one trades is set out briefly below.

• Guardrails. Spend a fixed real amount, but cut it when the portfolio falls behind and raise it when the portfolio runs ahead. The best-known version, from Jonathan Guyton and William Klinger in 2006, reduces withdrawals by 10% when the current withdrawal rate has risen 20% above its starting level and increases them by 10% when it has fallen 20% below (Kitces.com). The trade: a higher starting rate, paid for with cuts that can be long and, under the original rules, deep. Our calculator models a simple cut-only version, a 10% reduction in any year that starts below 80% of the starting balance. At 4% it raises the 30-year success rate from 90.5% to 94.2%; the rule is triggered in 52% of paths, and where it is, the median time at reduced spending is 13 years.
• Fixed percentage. Withdraw the same percentage of the current balance every year. The portfolio can never reach zero, since each withdrawal is a fraction of what is left. The trade: income moves in step with the market. A year in which the portfolio falls 20% is a year in which spending falls 20%, with no smoothing, and in a long decline the income can shrink to well below where it started.
• The RMD method. Withdraw the percentage in the table above, whether or not the account is subject to RMDs. It is a fixed-percentage rule whose percentage rises with age, so it spends more of the portfolio as the remaining horizon shortens and cannot run out. The trade: the same year-to-year variation as any percentage rule, and a low starting figure, 3.77% at 73, that leaves early retirement comparatively lean.
• The spending “smile.” The rule assumes spending rises with inflation every year for life. David Blanchett's 2014 study of household survey data found that retirees' spending tends to fall in real terms through retirement, by about 1% a year on average, with larger changes early and late in retirement, a pattern he named the retirement spending smile (Journal of Financial Planning). A plan built on that pattern can start higher. The trade: it relies on the household following the average, and late-life care costs are the part of the smile that turns up.

None of these removes the underlying risk. They decide who bears it: the portfolio, in the form of a chance of running out, or the household, in the form of a variable income. Our guide to sequence of returns risk looks at guardrails, cash reserves and the other responses in more detail.

The rule applies to what the portfolio must fund, and for most households that is well short of the whole budget. Subtract Social Security, any pension and any annuity income from planned spending, including the tax bill, and test only the remainder. For a household that plans to spend $100,000 a year:

Two consequences are worth drawing out. First, guaranteed income changes how much of the plan is exposed to markets at all. The household in the last row is relying on its portfolio for a fifth of its spending, and a poor decade threatens the extras, not the essentials. That household can reasonably think about its withdrawal rate differently from one with no other income. Morningstar's research makes the same observation: flexible spending methods work better when a substantial, predictable income floor sits underneath them.

Second, the timing matters. Someone who retires at 62 and claims Social Security at 70 has eight years in which the portfolio funds everything, followed by decades in which it funds much less. A single withdrawal rate describes neither phase. The Monte Carlo calculator handles this by letting other income begin in a chosen year; our guide to when to claim Social Security and the break-even calculator cover the claiming decision itself. For those considering buying guaranteed income, see the types of annuities and the annuity payout calculator.