The 4% rule isn't a rule. Here's the math behind it.

The 4% rule gets quoted like scripture. It isn't. It's the floor of a 1994 study by one financial planner, William Bengen, who ran historical numbers and named the worst case he could find. Bengen himself has been telling people for nearly two decades that 4% is too low. Nobody listens. So here's what the math actually says.

Where the 4% number comes from

October 1994. Journal of Financial Planning. A paper called "Determining Withdrawal Rates Using Historical Data." Author: William P. Bengen, a California financial planner who was tired of his profession quoting average market returns as if a retiree could withdraw them.

Bengen used U.S. market data going back to 1926. He tested rolling 30-year retirement windows. A retiree starting in 1926. Then 1927. Then 1928. All the way through. For each starting year, he asked one question: what's the highest inflation-adjusted withdrawal rate that would have survived all 30 years without running the portfolio to zero?

The asset mix was 50/50 stocks and intermediate-term Treasuries. The withdrawal was taken in year one, then bumped each subsequent year for CPI inflation.

The worst starting year in the entire historical record was 1966. A retiree who walked out the door that year ate a brutal one-two punch: late-60s and 70s inflation that mauled bond returns, then the 1973-74 bear market that gutted equities. That retiree's maximum sustainable rate was 4.15%.

Bengen rounded down. Called it SAFEMAX. 4%. The number stuck.

Read what Bengen actually wrote and one thing is obvious: he was naming a worst case, not a recommendation. The average safe withdrawal rate across all his historical starting years was closer to 7%. In some retirement windows, 10% would have worked. The 1966 retiree was the unlucky one. The 4% was the floor.

What the Trinity Study actually found

Four years later, three Trinity University finance professors (Philip Cooley, Carl Hubbard, Daniel Walz) ran a similar exercise. Different framing. Instead of asking "what rate would have worked every time," they asked "what percentage of historical 30-year windows would each rate have survived?"

They used 1926-1995 Ibbotson data. Tested multiple stock/bond mixes. Tested 15, 20, 25, and 30-year horizons. Their conclusion, in their own words: "Withdrawal rates of 3% and 4% are extremely unlikely to exhaust any portfolio of stocks and bonds during any of the payout periods."

The 30-year table is the one everyone cites. Here are the inflation-adjusted success rates for a 50/50 portfolio, the same allocation Bengen tested.

Withdrawal rate	30-year success (50/50)	30-year success (75/25)
3.0%	100%	100%
3.5%	100%	100%
4.0%	95%	98%
4.5%	79%	88%
5.0%	58%	76%

Two things jump out. First, the 4% rate is not bulletproof at 50/50. It failed roughly 5% of historical windows. Second, going stock-heavier raised the success rate at every withdrawal level. The 5% rate, treated as reckless in most blog posts, succeeded three-quarters of the time at 75/25.

That is not a rule. It is a probability distribution.

Bengen revised his own number

Here is the part the personal-finance internet largely refuses to print. Bengen has spent the last two decades raising his own figure.

In his 2006 book Conserving Client Portfolios During Retirement, he added small-cap stocks to the mix. Result: 4.5% for tax-free accounts, 4.1% for taxable. By 2012 he was talking about 4.2% as a starting point with annual inflation adjustments. In later interviews he settled on 4.5% to 4.7% as a more honest SAFEMAX.

In 2020 and again in 2022, he went further. Given low inflation and a reasonable CAPE ratio at retirement, Bengen argued the safe rate could be 5% or higher. Quote from Bengen, 2021: "I think a withdrawal rate of 5% is safe today for new retirees."

The man who wrote the 4% rule does not believe in the 4% rule. The number that survives in the popular imagination is the conservative floor of his earliest, narrowest study. The number he uses in his actual practice has moved.

What sequence-of-returns risk does

Here is the part of Bengen's research that matters more than the headline number. He named a real, mathematical phenomenon: the order of returns matters far more than the average.

Two retirees, same starting balance, same average 30-year return. One gets the bad years up front. One gets them at the end. The first retiree runs out of money. The second dies wealthy. Same math on paper. Different outcomes in life.

Why? Withdrawals in a down market sell more shares to produce the same dollar income. Those sold shares are gone. They can't participate in the eventual recovery. A 30% drawdown in year three of retirement is catastrophic in a way that a 30% drawdown in year twenty-five is not.

This is the actual contribution of Bengen's work. Not the number 4%. The recognition that the average return your portfolio earns over 30 years is not the rate you can safely withdraw, because the path matters. Michael Kitces, who has built much of his career on this research, puts it bluntly: a 7% average historical return does not justify a 7% withdrawal rate, because the worst-case path delivers that 7% with a sequence that destroys you.

Sequence risk is highest in the first 5 to 10 years of retirement. After that, the math forgives almost anything. This is why the 1966 retiree got crushed and the 1982 retiree (who experienced the entire bull run of the 80s and 90s) could have safely withdrawn over 10%.

What this means for your FIRE math

If you are doing 25x annual spending as your FIRE number, you are using 4% as a planning constant. Fine for a back-of-envelope. Wrong as a finished number.

What the research actually supports:

• 4% is the worst-case historical floor for a 30-year retirement. If you retire at 65, this is roughly the right number to plan around. If you retire at 40 and need money for 50 years, the floor drops. Wade Pfau and others have estimated closer to 3% to 3.5% for very long horizons.
• The average successful withdrawal rate is closer to 7%. Most retirees who used 4% died with multiples of their starting portfolio. Kitces' data shows over 90% of 4% retirees ended 30 years with more than they started, and the median ended with 2.8x.
• Stock allocation matters more than retirees admit. The Trinity table is unambiguous. A 75/25 portfolio survives higher withdrawal rates than 50/50. The instinct to derisk into bonds at retirement actively reduces sustainable income.
• Starting valuations matter. Pfau has consistently argued that current-day CAPE ratios should pull the safe rate down. Bengen, more recently, has argued the opposite given today's inflation profile. Both agree the static 4% number ignores the conditions at the moment you stop earning.
• Flexibility crushes rigidity. A retiree who cuts spending 10% in a bad year and ratchets up after a good run can support a meaningfully higher initial rate than one who blindly inflation-adjusts. Kitces' ratcheting rule is one version. Guyton-Klinger guardrails are another.

The 4% number is a starting point for one specific scenario: a 30-year retirement, U.S. assets, a 50/50 mix, taken from 1994 historical data, rounded down from the worst case ever recorded. Treat it that way. Don't treat it as a law of physics.

Bengen wrote a paper. The paper was good. The shorthand the internet built on top of it is worse than the paper. If you are basing a 40-year early-retirement decision on a number Bengen himself walked away from, you are using stale math. Read the original. Read the revisions. Adjust accordingly.