If you have ever worked out a FIRE number, you have almost certainly used the 4% rule, even if you never called it that. Every FIRE calculator tells you to multiply your yearly spending by 25, and 25 times your spending is just the 4% rule turned around, because 4% of 25 is exactly 1. My own Lean FIRE number is built that way. And yet when I finally sat down and asked where the 4% actually came from, I realized I had been trusting a number that I had never once seen anyone calculate.
So I rebuilt the test that produced it, from scratch, on 98 years of real US market data, and then I put the whole thing into a free calculator so you can run it on your own numbers.
Here is the short answer first. The 4% rule says you can take 4% of your portfolio in your first year of retirement, raise that dollar amount with inflation every year after, and expect the money to last at least 30 years, because it did in every 30 year stretch of US history. When I reran the test with 60% in the S&P 500 and 40% in intermediate government bonds, the highest rate that survived every 30 year retirement starting between 1928 and 1996 was 4.01%, and the year that set it was 1966.
In this article I will walk you through where the rule came from, exactly how the math works year by year, what happened to the unlucky person who retired in 1966, why most retirees could have spent far more than 4%, what changes when a retirement runs 50 years instead of 30, which is my situation, and how you can find your own safe withdrawal rate with the calculator.
Where the 4% rule came from
The 4% rule comes from a single paper. In October 1994 a financial planner in Southern California named William Bengen published "Determining Withdrawal Rates Using Historical Data" in the Journal of Financial Planning. The question he was trying to answer is the one every retiree asks, which is how much you can spend each year without running out, and a common answer back then was to look at the market's average return and spend a little less than that. Bengen saw the problem with averages, because every retiree lives through one specific order of years, and a crash in your first few years does far more damage than the same crash 20 years later, since you are selling to pay your bills while everything is down.
So instead of averages he used history. He took the actual US record from 1926 onward, the return of large US stocks, the return of intermediate term government bonds, and the actual inflation of every year, and he played out a retirement starting in every single one of those years. He found that a first year withdrawal of 4%, raised with inflation every year after, never emptied a portfolio in less than 33 years, and he recommended holding somewhere between 50% and 75% in stocks. He also found that 3% to 3.5% lasted 50 years or more in every case he tested. The people who did worst were the ones who stopped working in the late 1960s, right before the crash of 1973 and 1974 and the decade of high inflation that came after it.
Bengen has kept refining this for 30 years. In his 2025 book, A Richer Retirement, he raised his number to 4.7% for a 30 year retirement by spreading the money across more kinds of assets, including small company stocks and international stocks. The original 4%, from a plain mix of large US stocks and government bonds, is still the number the FIRE community uses, so that is the one I rebuilt.
The math, one year at a time
The method is simple enough to do in a spreadsheet, which is a big part of why it spread so widely. You set your first year of spending as 4% of what you hold on the day you retire. On every 1 January you take out that year's spending and rebalance whatever is left back to your mix, which here is 60% stocks and 40% bonds. Over the year the stock part earns the S&P 500's real total return for that year, dividends included, and the bond part earns the real return of intermediate government bonds. Then the next year's spending goes up by that year's inflation, no matter what the market did, and you repeat this for 30 years. If at any point the money cannot cover a full year of spending, the plan has run out.
The one detail most people get wrong is that the 4% only applies once. After year 1 you stop looking at the balance entirely and simply raise last year's dollar amount by inflation. Taking 4% of whatever the portfolio is worth each year would be a completely different plan, and a much safer one, because a percentage of what is left can never actually run out.
Here is what that looks like for someone who retired on 1 January 1966 with $1,000,000, in the dollars of the time.
| Year | Money on 1 January | Spending that year | S&P 500 | Bonds | Inflation | Spending as a share of the money |
|---|---|---|---|---|---|---|
| 1966 | $1,000,000 | $40,000 | down 10.0% | up 4.7% | 3.5% | 4.0% |
| 1967 | $920,582 | $41,400 | up 23.8% | up 1.0% | 3.0% | 4.5% |
| 1968 | $1,008,282 | $42,642 | up 10.8% | up 4.5% | 4.7% | 4.2% |
| 1969 | $1,045,807 | $44,646 | down 8.2% | down 0.7% | 6.2% | 4.3% |
| 1970 | $948,700 | $47,414 | up 3.6% | up 16.9% | 5.6% | 5.0% |
| 1974 | $997,905 | $58,133 | down 25.9% | up 5.7% | 12.3% | 5.8% |
| 1975 | $815,121 | $65,283 | up 37.0% | up 7.8% | 6.9% | 8.0% |
| 1995 | $255,119 | $187,923 | up 37.2% | up 18.2% | 2.5% | 73.7% |
Look at the last column, because it is the whole story of a failing retirement. The spending started at 4% of the money and, through no fault of the retiree, it crept to 5.8% by 1974 and 8% by 1975, and by the final year it was eating 73.7% of what was left. That happens because the dollars keep rising with prices, which rose 4.8 times over those 30 years, while the money sitting underneath them shrinks.
Now do that for every start year from 1928 to 1996, which gives you 69 separate 30 year retirements with a complete record behind them. For each one you find the highest first year rate that still lasted the full 30 years. The lowest of those 69 numbers is the safe withdrawal rate, the highest rate that would have survived every retirement in the record. Bengen later gave this number a name, SAFEMAX.
The data I used, and how close it gets to Bengen
I tried to match Bengen's ingredients as closely as free, public data allows.
| Ingredient | What I used | Where it comes from |
|---|---|---|
| Stocks | S&P 500 total return, dividends reinvested | Aswath Damodaran's historical returns table, NYU Stern |
| Bonds | Intermediate government bonds, about 5 years to maturity | The Ibbotson SBBI index from Morningstar to 1975, then 5 year Treasury returns built from Federal Reserve yields |
| Inflation | US consumer prices, December to December | Bureau of Labor Statistics |
| Years | 1928 to 2025 | 98 years, 69 full 30 year retirements |
| Rebalancing | Once a year, every 1 January | Same as Bengen |
| Taxes and fees | None | Same as Bengen |
The method is identical to his, and the data is very close without being a perfect copy. His record started in 1926 and mine starts in 1928. He used Ibbotson's large company stock series, and I use Damodaran's S&P 500 numbers, which describe the same market and differ slightly in a few years. For the bonds, I use Ibbotson's own intermediate government index up to 1975, and after that I rebuild 5 year Treasury returns from the Federal Reserve's published yields, which I checked against the index for the years where both exist and found them about half a point a year apart on average. The answer lands right where his did, with about 4% and the late 1960s as the worst time to retire.
I tried one other version along the way that is worth mentioning. With 10 year Treasury bonds instead of intermediate ones, 4% fails for anyone who retired in 1965, 1966, 1968 or 1969, because longer bonds lost far more value when interest rates climbed through the 1970s. So the kind of bonds you hold matters more than most people realise, and Bengen's 4% depends on bonds that mature in around 5 years.
What happened to the person who retired in 1966
The 4% rule is really the story of one retiree, so let me tell it in today's dollars. Say you retired on 1 January 1966 with $2 million in today's money, put 60% in the S&P 500 and 40% in intermediate government bonds, and spent $80,000 a year in today's money, which is 4%.
The market fell 10% in your very first year. Then it recovered for a while, and at the start of 1973 you still had $1.68 million in today's money. Then 1973 and 1974 took the S&P 500 down 14.3% and then 25.9%, while prices rose 12.3% in 1974 alone, and by January 1975 your $2 million had become $999,000 in today's money. Half of it was gone in 9 years, while your spending was exactly as big as it had been on day one, in real terms. Inflation then rose again at the end of the decade, with prices up 13.3% in 1979, so even the great bull market of the 1980s and 1990s could only slow the decline. By 1982 you had $598,000 in today's money, by 1990 you had $445,000, and at the end of 1995, after your 30th and final year of spending, you were left with about $36,000, which covers 5 months.
You made it, and that is precisely what the 4% rule is. It is the rate that got the unluckiest retiree in the record to the end with almost nothing left over. Push the rate any higher and this same person runs out. At 4.5% they run out after 23 years, and at 5% after 19 years, which matches what Bengen found for retirees of the late 1960s.
Most retirees could have spent far more
Here is the part that almost nobody mentions. The 4% rule is set by the single worst start year, so for almost everyone else it was far too cautious.
Each bar is one retirement year, from 1928 to 1996, and it shows the highest spending rate that would have lasted 30 years from that start. The dashed line is 4%. Only the late 1960s get close to it.
The typical start year in this record could have spent 6.58% a year and still lasted 30 years. Someone who retired in 1982 could have spent 10.16% a year, more than double the rule. And if everyone had followed the 4% rule anyway, 46 of the 69 retirees would have ended their 30 years with more money in real terms than they started with, and the typical one would have finished with 1.67 times their starting wealth. Only 9 start years in the whole record had a safe rate below 4.5%, and they were 1937 and 8 of the years between 1962 and 1973.
That is the honest trade you make with the 4% rule. You are insuring yourself against being the 1966 retiree, and in most histories the price of that insurance is dying with a lot of unspent money. Whether that trade is worth it is a personal decision, and it is the reason flexible spending strategies exist at all.
How much should be in stocks
Bengen's finding that 50% to 75% stocks works best holds up well in this data too. Here is the safe withdrawal rate for different mixes and different lengths of retirement, each one the highest rate that survived every complete start year in the record.
| Years the money has to last | 25% stocks | 40% stocks | 50% stocks | 60% stocks | 75% stocks | 100% stocks |
|---|---|---|---|---|---|---|
| 30 years (69 starts) | 3.59% | 4.01% | 4.02% | 4.01% | 3.98% | 3.49% |
| 40 years (59 starts) | 2.95% | 3.46% | 3.71% | 3.73% | 3.73% | 3.34% |
| 50 years (49 starts) | 2.56% | 3.07% | 3.35% | 3.55% | 3.56% | 3.24% |
This table shows two things clearly. A portfolio that is mostly bonds does worse, because over decades inflation takes most of what bonds earn while the spending keeps climbing. And a portfolio that is all stocks also does worse, because the person who retired in 1929 with 100% in stocks lived through the Great Depression with nothing steady to sell, while stocks lost 65% between 1929 and 1932. The middle of the table is where the money lasts longest, and the longer your retirement, the more the stock share starts to matter.
Why 4% is a 30 year rule, and why I need 50
This is the part that matters most to me personally. Bengen designed his test for someone retiring at 65 who needed the money to last 30 years. I stopped working in my thirties, so if I make it to 90, my money has to last well over 50 years, and the table above says the answer changes quite a lot over that kind of stretch.
Over 50 years, with 60% in stocks, the highest rate that survived every complete start year in the record is 3.55%, and the worst year is once again 1966. With 100% in stocks it drops to 3.24%, set by 1929. That lines up almost exactly with what Bengen himself wrote back in 1994, that 3% to 3.5% lasted 50 years or more in every case he tested. So if you are retiring early, the honest version of the rule is closer to a 3.5% rule, and 25 times your spending is really a 30 year number. For a 50 year retirement, about 28 times your spending is what history actually supports.
My investments sit in the US market, which is exactly why the US record is the one I care about, and why I wanted to see these numbers for myself instead of trusting a rule built for people retiring at 65 in the 1990s.
What the 4% rule leaves out
Before you rely on any of these numbers, here are the things the test simply does not include, and some of them matter a lot.
The first is fees. Bengen's test charges no fees and my rebuild follows it, but real money always costs something to hold. A 0.1% index fund fee on the stock part is enough to make the 1966 retiree run out in their 30th year, and the safe rate slips to 3.99%. A 1% advisor fee on the whole portfolio is far worse, because it drops the safe rate to 3.54%, the 1966 retiree runs out after 23 years, and 8 of the 69 start years fail at 4%. If you pay an advisor 1% a year, your real safe rate is about 3.5%.
The second is taxes, which depend entirely on your mix of 401(k), Roth and taxable accounts, so the test leaves them out. The third is the income most people actually have, like Social Security, a pension or a bit of work, all of which make the real picture safer. The fourth is that real people change their spending. Nobody keeps spending the same amount through a 50% crash, and if you are willing to cut back in bad years with a set of rules, you can start higher than 4%. I wrote about exactly how that works in surviving a stock market crash on Lean FIRE, using the guardrails strategy.
The fifth is that this is the US, which had one of the best stock markets of any country in the last century. If your money is in another market, the US numbers do not transfer automatically. I ran the same kind of test on Indian data in this 26 year backtest of a ₹10 crore retirement, and India only has about 26 years of total return history, which is far too short to produce a rule like this. And the last one is that 98 years of history is fewer independent retirements than it sounds. The 69 start years overlap heavily, so the whole record holds only about 3 completely separate 30 year retirements, and the future can always produce something worse than the worst thing in the past.
How to find your own safe withdrawal rate
This is the part I built the calculator for. The retirement backtest calculator in its US mode runs the exact math in this article, on the same data, so you can test any numbers you like instead of trusting mine.
Open it, make sure the US is selected, and fill in the sentence at the top. Put in the money you would retire with, how much of it you would keep in the S&P 500, what share you want to spend each year, and how many years the money has to last, which is set to Bengen's 30 and goes anywhere from 5 to 60. Then pick the year you retire into, and the default is 1966 because that is the year that defines the rule. Press the button and it plays your plan through that exact history, year by year, the same way the table above does.
Right under the result you will find the line that answers this whole article. It tells you the most you could have spent from every start year with a full record behind it, for your stock share and your number of years, and which year would have run out if you had spent even a little more. With the defaults it reads 4.01% and 1966. Change the years to 50 and it reads 3.55%. That number is your historical safe withdrawal rate. Below it you will see a bar for every single start year since 1928, so you can click on any year and watch what your plan would have done through it, and a table of every year in your plan, with that year's returns, inflation, money and spending.
Remember that this number is the edge, the rate at which the worst retiree in the record ended with almost nothing, so most people choose to leave some room below it. And if you want to go beyond a single history and test your whole plan, with your actual spending, your savings years, your side income and 1,500 possible futures, that is exactly what the FIRE planner does.
So, is the 4% rule safe?
For a 30 year retirement with roughly 50% to 75% of your money in US stocks and the rest in intermediate government bonds, history says yes, just barely, with almost nothing left for the worst start year in the record. For an early retirement of 40 to 50 years it is too high, and something closer to 3.5% is what the past actually supported. Add fees and it drops further. And in most histories, 4% left people far richer than they needed to be, which is the real argument for spending flexibly instead of following any fixed rule without ever adjusting.
The 4% rule is the worst case of one country's past, worked out honestly, and once you understand the math behind it you can run it on your own life and decide for yourself.
Where the numbers come from
Every figure in this article comes from the same data the backtest calculator uses. Stock returns are the S&P 500 including dividends from Aswath Damodaran's historical returns table at NYU Stern, updated January 2026. Bond returns up to 1975 are the Ibbotson SBBI US Intermediate Term Government Bond Index, using Morningstar data as published by Hartford Funds, and from 1976 they are 5 year Treasury returns that I built from the Federal Reserve's H.15 Treasury yields. Inflation is the December to December change in the US consumer price index from the Bureau of Labor Statistics. Bengen's original paper appeared in the October 1994 Journal of Financial Planning, and his 4.7% update is from his 2025 book A Richer Retirement, covered here by Advisor Perspectives.
The disclaimer, please actually read it
This article explains how the 4% rule was built and rebuilds the same test on historical US data. It is general information to help you understand the math, and it is not financial, investment or tax advice. I am not a licensed financial advisor, and I know nothing about your income, your family or your goals. A backtest describes the past and promises nothing about the future, and this one ignores taxes, fees, health shocks and large one off costs. The rates shown are the edge of what would have survived, where the worst retiree finished with almost nothing, so please do your own math for your own life and talk to a qualified professional before you act.