The 4% rule: where it came from, and what the research actually says now

The best-known number in retirement planning came out of two specific papers with specific assumptions, and almost every argument about it today is really an argument about those assumptions.

Updated September 2026

Two papers, four years apart

The figure everyone calls the "4% rule" was never published under that name, and it does not come from one study. It comes from two. The first is William Bengen's "Determining Withdrawal Rates Using Historical Data," in the October 1994 Journal of Financial Planning. The second is "Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable," by Philip Cooley, Carl Hubbard and Daniel Walz in the February 1998 AAII Journal. All three authors of the second paper were professors of finance at Trinity University in San Antonio, which is why it is remembered as the Trinity Study. The methods overlap and the conclusions rhyme, but they tested different things, and the differences are where most of the modern debate lives.

What the 1994 paper actually tested

Bengen used annual US return and inflation data from Ibbotson Associates' Stocks, Bonds, Bills and Inflation: 1992 Yearbook. The baseline portfolio was 50% common stocks and 50% intermediate-term Treasury notes, continually rebalanced, with every asset assumed to sit in a tax-deferred account. The withdrawal method was precise: take a set percentage of the portfolio's starting value in year one, then raise that dollar amount every year in line with actual inflation. The outcome measured was "portfolio longevity" — how many years before withdrawals exhausted the money — calculated separately for each of the 51 retirement start years beginning in 1926.

The results were narrower than the slogan suggests:

• At a 3% first-year rate, every start year lasted at least 50 years, the maximum the paper charted. The same held up to roughly 3.5%.
• At 4%, no start year was exhausted before 33 years, and most lasted 50 years or longer.
• At 4.25%, the worst start year could have run dry in as little as 28 years.
• At 6% with that same mix, 31 of the 51 start years would have outlived the assets and only 20 were adequate for 30 years — under 40%.

So 4% was not chosen because it was round. It was the highest tested rate that cleared a self-imposed 30-year minimum in every historical start year, with the next step up failing. One wrinkle the retellings miss: the paper states its own worst case two ways, as about 35 years in the commentary on one chart and as 33 years in the section on strategy. The lower figure is the one to plan against. The paper also compared asset mixes: raising stocks to 75% lifted the count of start years reaching the 50-year ceiling from 40 to 47, while shortening two cases — a 1966 start from 33 years to 32, and a 1969 start from 36 to 34.

What the Trinity Study added

The 1998 paper ran a much wider grid: withdrawal rates from 3% to 12%, payout periods of 15, 20, 25 and 30 years, and five fixed allocations from all-stock to all-bond. Stocks were the S&P 500. The fixed-income side was long-term, high-grade corporate bonds. Data ran from 1926 to 1995, which gave 41 overlapping 30-year windows. The authors stated plainly that they did not adjust for taxes or transaction costs. A window counted as a success if the portfolio finished with any value above zero.

Trinity Study, Table 3: share of the 41 overlapping 30-year windows the portfolio survived, with each year's withdrawal indexed to the Consumer Price Index.

Indexing to inflation is the expensive part

The most striking number in the 1998 paper is not in the table above. The 75/25 mix at a 6% withdrawal rate supported 95% of 30-year windows when the dollar withdrawal was held flat, and only 68% once that withdrawal was raised each year with the CPI. Same portfolio, same rate, same history — the indexing alone accounts for the gap. Anyone quoting a success rate without saying whether withdrawals were indexed has dropped the most important assumption in the study. The authors also flagged that many economists at the time believed CPI overstated the real rise in living costs by 1.0 to 1.5 percentage points a year, which would bias their inflation-adjusted table downward.

The critiques that have held up

Sequence of returns

Average returns do not decide the outcome; their order does. Bengen's charts showed the 1973–74 decline damaging portfolios whose withdrawals had begun as much as 20 or more years earlier, while the 1937–41 decline reached back only about 9 or 10 years and the early-1930s slump only four or five. The Depression ranked third largely because falling prices shrank the inflation-indexed withdrawals and bond returns were modestly positive, together offsetting part of the market loss. The binding constraint is a weak market combined with high inflation early in retirement — which is why the paper warned against raising withdrawals after a few strong early years.

The data are American

Both papers rest on one country's twentieth century. Wade Pfau's December 2010 Journal of Financial Planning article, "An International Perspective on Safe Withdrawal Rates: The Demise of the 4 Percent Rule?", applied the same historical-simulation approach to 109 years of data across 17 developed markets, with 30-year horizons and retirements beginning between 1900 and 1979. Even on assumptions he called overly optimistic, a 4% inflation-adjusted rate would have been "safe" in only 4 of the 17 countries, and a fixed 50/50 allocation failed at some point in all 17. The US figure is not a law of finance; it is one country's record.

Thirty years was a choice

The horizon was set to match a conventional retirement, and the arithmetic does not stretch. Bengen's worst case at 4% lasted 33 years against a 30-year test — three years of slack. A 40- or 50-year retirement simply was not covered by that result, which is the single most important caveat for anyone planning to stop work early. Even a conventional plan runs further than instinct suggests: 30 years from age 65 reaches 95.

Costs were left out

Neither paper deducted investment costs, so every success rate in the table above is an upper bound. The SEC's investor bulletin on fees shows the size of the effect with $100,000 growing at 4% a year for 20 years: at an ongoing fee of 0.25% the balance reaches about $208,000, at 0.50% about $198,000, and at 1% about $179,000 — roughly $29,000 between the cheapest and most expensive case. How far costs push the Trinity figures down cannot be read off that grid, and the reason is worth being explicit about: a percentage fee scales with the balance, while an inflation-indexed withdrawal is a fixed real dollar amount that grows as a share of a shrinking portfolio. The two drains work differently and do not simply add.

Why the original author's figure went up

In an August 2025 AAII Journal interview, Bengen put his "Universal SAFEMAX" at 4.7%, up from 4.5%, which had itself replaced 4.0% in a January 2018 interview. The increase is not a bullish forecast. He attributes it to testing a far wider set of asset classes than the 1994 paper's two — seven in total, spanning US stocks by company size, international stocks, and shorter- and longer-dated government holdings — in a mix he reports as 55% stocks, 40% bonds and 5% cash. The point is methodological: a different set of inputs produces a different worst case, not that any particular mix is preferable. The binding case remains a retirement starting in October 1968, which met early bear markets followed by years of high inflation. Of roughly 400 historical retirements modeled, 399 could have started above 4.7%, and the historical average was 7.0%. In the same interview the interviewer put it to him that the odds of depletion over 30 years climb sharply once the rate passes 5.25%; his answer was the general one, that higher rates drain a portfolio faster.

The misunderstanding that causes the most trouble

The research tested 4% of the initial balance, indexed to inflation thereafter — a fixed real income that ignores what markets do next. Most people describe something different: taking 4% of the current balance each year. That second method can never exhaust a portfolio, because the withdrawal shrinks with the account, but the income swings hard. They are not the same strategy and they do not have the same failure mode. A second trap is the definition of success: a portfolio ending the 30 years with one dollar left counted as a success in the 1998 tables.

The indexing matters more than people expect, and it runs in the opposite direction from the erosion inflation is usually discussed as causing: here the withdrawal grows while the balance it is drawn from may not. At the most recent published 12-month CPI-U change, a $40,000 first-year withdrawal becomes $41,360 in year two, then keeps climbing whether or not the portfolio has recovered — which is why the papers' inflation-indexed columns fail so much more often than their flat-withdrawal columns. The guide to what inflation does to money sitting still covers the index and its limits; the inflation calculator compounds a chosen rate over a full retirement.

What the numbers look like in context

The Federal Reserve's Survey of Consumer Finances for 2022, the most recent available and published October 18, 2023, found 54.3% of US families held retirement accounts, with a median value of $86,900 among those who had them and a mean of $334,000. A 4% first-year withdrawal on that median balance is $3,476 a year.

That is the context the rule is usually quoted without. The whole withdrawal-rate debate presupposes a portfolio large enough for the rate to be the binding question. On the median balance, moving from a 4% starting rate to 4.7% changes the first-year income by about $608, while adding $50,000 to the balance changes it by $2,000 — more than three times as much. For most households the size of the balance decides the answer long before the rate does.

Working through it

These papers answer one narrow question: given a fixed horizon and US history, what starting percentage survived the worst case? They say nothing about taxes, costs, variable spending, or income from outside the portfolio such as Social Security. The parts that are genuinely personal — how long a horizon to plan for, how much income variability is tolerable, how much slack to leave — are the parts no historical study settles.

• The FIRE calculator applies a withdrawal rate you choose to a target portfolio, which is where the 40-to-50-year horizon problem becomes visible.
• The retirement income calculator works the other direction: from a balance to an annual income.
• The retirement calculator and the 401(k) calculator cover the accumulation side.
• The net worth calculator helps separate the portfolio that funds withdrawals from assets that do not.

This guide is educational and reports published research. It is not a recommendation about any withdrawal rate, asset or product.

Common questions

Who created the 4% rule?

William Bengen, in "Determining Withdrawal Rates Using Historical Data," published in the October 1994 Journal of Financial Planning. The separate 1998 Trinity Study by Philip Cooley, Carl Hubbard and Daniel Walz of Trinity University tested a wider grid of rates and periods and is often confused with it. Neither paper used the phrase "4% rule."

Does the 4% rule mean I withdraw 4% of my balance every year?

No, and this is the most common misreading. Both papers tested 4% of the portfolio's value in the first year only, with that dollar amount then raised each year by inflation. Taking 4% of the current balance annually is a different strategy: it can never deplete the portfolio, but the income rises and falls with the market.

Why did the 1998 Trinity Study report 98% success at 4% rather than 100%?

Its inflation-adjusted 30-year table shows 98% for a 75/25 stock-bond mix, 95% for 100% stocks and 95% for 50/50, measured across 41 overlapping 30-year windows from 1926 to 1995. One or two historical windows depleted the portfolio. A window counted as a success if any value above zero remained at the end, so even the 98% figure sets a low bar for what surviving means.

Does the 4% figure work for a 45-year retirement?

The research does not support that conclusion. Bengen's 1994 paper set a 30-year minimum and found the worst historical start year at 4% lasted 33 years, leaving three years of margin. A 40- or 50-year horizon was outside what the result tested. At 3% every start year in his data lasted at least 50 years, the longest period his charts displayed.

Why has Bengen's own figure risen to 4.7%?

In an August 2025 AAII Journal interview he attributed the increase to testing a far wider set of asset classes than the 1994 paper's two - seven in total - in a mix he reports as 55% stocks, 40% bonds and 5% cash, rather than to a more optimistic market view. The binding historical case is still a retirement beginning in October 1968. He noted 399 of roughly 400 modeled retirements could have started above 4.7%, with a historical average of 7.0%.

What is the strongest objection to using US historical data?

Wade Pfau's December 2010 Journal of Financial Planning study, "An International Perspective on Safe Withdrawal Rates: The Demise of the 4 Percent Rule?", applied the same method to 109 years of data from 17 developed markets, with 30-year horizons and retirements from 1900 to 1979. Even under assumptions he described as overly optimistic, a 4% inflation-adjusted rate was "safe" in only 4 of the 17 countries, and a fixed 50/50 allocation failed at some point in all 17.

Did these studies account for taxes and investment fees?

No. The 1998 paper states it did not adjust for taxes or transaction costs. The 1994 paper assumed all assets sat in tax-deferred accounts. That makes every published success rate an upper bound. The SEC's investor bulletin on fees shows the scale: $100,000 growing at 4% for 20 years ends near $208,000 at a 0.25% ongoing fee versus about $179,000 at 1%. How much lower the real success rates sit cannot be read off the Trinity grid, because a percentage fee scales with the balance while an inflation-indexed withdrawal does not.

Sources

• Bengen, "Determining Withdrawal Rates Using Historical Data," Journal of Financial Planning (October 1994); FPA reprint in the March 2004 "Best of 25 Years" collection — The 1994 paper's data source (Ibbotson SBBI 1992 Yearbook), the 50/50 stock and intermediate-term Treasury baseline, continual rebalancing, the tax-deferred assumption, 51 scenario start years beginning in 1926, the 3%, 4%, 4.25% and 6% longevity findings, the 33-year worst case and the paper's alternative "about 35 years" statement, the 40-to-47 comparison at 75% stocks, the 1966 (33 to 32) and 1969 (36 to 34) penalties, the reach-back periods for the 1973-74, 1937-41 and early-1930s events, the deflation-plus-positive-bond-returns explanation for the Depression ranking third, and the warning against raising withdrawals after strong early years. The URL is the March 2004 reprint; its editor's note confirms the October 1994 original issue.
• Cooley, Hubbard and Walz, "Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable," AAII Journal (February 1998) — The Trinity Study's 1926-1995 data period, the S&P 500 and long-term high-grade corporate bond series, withdrawal rates of 3% to 12%, payout periods of 15, 20, 25 and 30 years, five allocations from all-stock to all-bond, 41 overlapping 30-year windows, the exclusion of taxes and transaction costs, the success definition of an ending value above zero, the full inflation-adjusted 30-year success-rate table (Table 3), the 95%-versus-68% unadjusted/indexed contrast at 6% for the 75/25 mix, the authors' affiliation at Trinity University, and their note that CPI was believed to overstate living-cost increases by 1.0 to 1.5 percentage points a year.
• Pfau, "An International Perspective on Safe Withdrawal Rates: The Demise of the 4 Percent Rule?", Journal of Financial Planning (December 2010) — The exact article title (re-verified on the page), 109 years of data across 17 developed markets, 30-year horizons, retirements between 1900 and 1979, the finding that a 4% inflation-adjusted rate was safe in only 4 of 17 countries, that a fixed 50/50 allocation failed at some point in all 17, and the author's own characterization of the assumptions as overly optimistic.
• "Is 4.7% the New Safe Retirement Withdrawal Rate?", AAII Journal (August 2025) — Bengen's Universal SAFEMAX of 4.7%, the prior 4.5% figure and the January 2018 move up from 4.0%, the October 1968 binding case, the seven asset classes and the 55% stock / 40% bond / 5% cash mix, 399 of roughly 400 retirements able to start above 4.7%, the 7.0% historical average, and the 5.25% depletion remark as the interviewer's premise rather than a finding of Bengen's.
• US Bureau of Labor Statistics, Consumer Price Index news release for August 2026 (dated archive copy, released September 11, 2026) — The 3.4% 12-month change in the all-items CPI-U to August 2026, the 2.4% 12-month change excluding food and energy, and the September 11, 2026 release date. The dated archive URL is used in place of the rolling bls.gov/news.release/cpi.nr0.htm, which always serves the latest release.
• FRED, Consumer Price Index for All Urban Consumers: All Items in U.S. City Average (CPIAUCSL) — Confirming that August 2026 was the latest published CPI observation as of October 9, 2026. This is a rolling series page: it will show later observations in future and will not evidence "latest" as of this date.
• Federal Reserve, "Changes in U.S. Family Finances from 2019 to 2022: Evidence from the Survey of Consumer Finances" (October 2023) — The 54.3% share of US families holding retirement accounts in 2022, the $86,900 conditional median account value and the $334,000 conditional mean, in 2022 dollars.
• Federal Reserve, Survey of Consumer Finances index page — Confirming the 2022 SCF is the most recent survey conducted and that its summary article was published October 18, 2023. This index page is updated when a new survey is released, so the "most recent available" claim will need re-checking once 2025 SCF results appear.
• SEC Office of Investor Education and Advocacy, Investor Bulletin: "How Fees and Expenses Affect Your Investment Portfolio" (current version, updated July 23, 2025) — The fee illustration and its published outcomes: a $100,000 portfolio at a 4% annual return over 20 years reaching roughly $208,000 at a 0.25% ongoing fee, $198,000 at 0.50% and $179,000 at 1%. The superseded sec.gov PDF (ib_fees_expenses.pdf) is no longer cited. The bulletin itself does not state the dollar gap; the roughly $29,000 figure is the difference between its own rounded endpoints.

Read the other guides, or browse every calculator.