The 4% Rule in 2026: What 150 Years of Data Actually Shows

The 4% rule is the most quoted number in early retirement, and almost nobody checks it. We ran a fixed 4% real withdrawal through every 30-year window in 155 years of US market history. It survived 97.9% of the time - and the failures tell a sharper story than the headline.

The 4% rule says you can retire on roughly 25 times your annual spending, withdraw 4% of that starting balance in the first year, adjust it for inflation each year after, and expect the money to last 30 years. It is the single most repeated number in early retirement, and it is usually quoted with far more confidence than anyone has actually earned. So we checked it against the record. Running a fixed 4% real withdrawal through every 30-year window in 155 years of US market history, the money lasted the full three decades in 1,476 of 1,507 windows: 97.9% of the time.

That number is high enough to explain why the rule became gospel and imperfect enough to explain why the fine print matters. This article is the story of where the rule came from, what it actually claims, and exactly where it has broken. If you want the raw window-by-window survival tables at several withdrawal rates, our companion piece on safe withdrawal rate history is the reference table. This one is the argument around the numbers.

Where the rule came from

The 4% rule is not an economic law. It is one financial planner’s backtest that happened to hold up. In 1994, William Bengen published “Determining Withdrawal Rates Using Historical Data” and did something the industry had mostly hand-waved before: he took actual US market history, walked a hypothetical retiree through every 30-year starting point he had data for, and asked which fixed withdrawal rate would have survived even the worst of them. His answer, using a stock-and-bond mix, was about 4%. The retiree who happened to start at the worst possible moment still made it 30 years pulling roughly that rate.

A few years later the “Trinity study” (three Trinity University professors, 1998) reframed the same idea as success probabilities across different rates and horizons, and the “4% rule” as a catchphrase was born. Two things are worth holding onto here. First, the number was reverse-engineered from a single worst-case historical start, not derived from any theory about markets. Second, it was calibrated to US data specifically, over a century that the US spent becoming the winning market in a field of many. Both of those facts age into caveats later in this piece.

What the rule actually says (and what it doesn’t)

The most common misreading is that “4%” means you withdraw 4% of your current balance every year. That is a different, safer strategy that can never technically run out, because you are always spending a fraction of whatever is left. The real 4% rule is stricter and more brittle: you spend 4% of your starting balance in year one, then hold that dollar amount constant in real terms for the rest of retirement, regardless of what the market does. A 2008 in year three does not lower your withdrawal. You keep pulling the same inflation-adjusted paycheck out of a portfolio that just fell by half. That rigidity is exactly why the rule can fail, and why a fixed-withdrawal backtest is the honest way to test it.

The other thing the rule quietly assumes is a 30-year horizon. Bengen was modeling a conventional retirement at 65. Someone coasting to an early retirement at 45 is asking a 45- or 50-year question, and the 4% rule was never validated over spans that long. We come back to that below.

The rule, recomputed on 155 years

Here is our own pass at it, computed fresh from the bundled Shiller US total-return series (real, inflation-adjusted, dividends reinvested) rather than borrowed from Bengen or anyone else. Because this data is monthly, we can test far more starting points than an annual backtest: every month from February 1871 through August 1996 opens a distinct 30-year window, for 1,507 windows in total. Each window starts a retiree with a fixed 4% real withdrawal and asks the one binary question that matters: did the money survive all 30 years, or did it hit zero first?

Every monthly-start 30-year window in an all-US-equity portfolio, 1,507 windows from Feb 1871 to Aug 1996, using real total return and a fixed real withdrawal of the starting balance. A window "survives" if the inflation-adjusted balance never hit zero across the full 30 years. Historical backtest of starting points that actually happened, not a forecast. See the methodology note below.

A note before anyone cross-checks: these figures deliberately differ from the ones in our 30-year-window companion article, which reports 96.8% at 4%. That is not a contradiction. That article uses annual-start windows and annual withdrawal steps; this one uses monthly-start windows and monthly steps, which samples more starting points and smooths the within-year timing. Two honest methods, two slightly different counts, same conclusion: 4% is strong but not spotless. When a number gets quoted as often as this one, it is worth seeing that it moves a little depending on exactly how you slice history, and settles in the high-90s either way.

The shape of the table is the real lesson. Below 4% the rule is close to bulletproof: 3.5% missed only four windows and 3.0% missed none, across 155 years that include the Great Depression, two world wars, and 1970s stagflation. Above 4% the ground gives way quickly. Half a point to 4.5% roughly triples the failure count, and 5% left one in six retirees short. The 4% figure sits right at the knee of that curve, which is why it has survived as the default: it is close to the most you can pull while still clearing almost all of history.

Where the rule breaks

The honest part of the 4% rule is its 2.1% failure rate, and those 31 failed windows are not random noise. They cluster in exactly two places.

All 31 of the 4% failures fall in these two start-date cohorts. Counts derived from the same 1,507-window backtest above; every other starting month in 155 years survived a 4% withdrawal for the full 30 years.

The larger cluster is the surprise. It is not the 1929 retirees, the ones who watched their portfolio fall roughly 80% in real terms; most of them, brutally, still made it, because the recovery that followed was violent enough to refill the account. The bigger group of failures belongs to people who retired in the mid-to-late 1960s into a market that looked fine and then delivered a decade where inflation quietly ate real returns from both stocks and bonds at once. There was no dramatic crash to warn them. The portfolio just failed to keep pace with a rising cost of living, year after year, while the fixed real withdrawal kept pulling the same amount out.

The single worst starting month in the entire dataset was retiring near the 1929 peak: that portfolio ran out of money about 16 to 17 years in, less than the promised 30. That is the concrete failure the 4% rule carries. It is uncommon, it happened to people who retired at genuinely terrible moments, and it happened.

Why the order of returns is the whole game

What links both failure cohorts is not the average return over their 30 years. Some of them earned a perfectly respectable long-run average. What ruined them was that the bad years came first. A portfolio that drops early, while you are withdrawing a fixed amount, sells more of itself at low prices and has a permanently smaller base to recover from. The same bad decade arriving late, after the portfolio has grown, does far less damage. This is sequence-of-returns risk, and it is the reason a single average-return number can never tell you whether a retirement is safe. We dig into the mechanics in is Coast FIRE risky, and it is also why any honest simulation has to keep real crashes attached to their real aftermaths instead of reshuffling years independently.

The spread this produces is enormous, and it is the other half of the 4% story. Among the windows that did survive at 4%, the median retiree finished 30 years with about 3.3 times their starting balance still invested, in real terms. The luckiest finished with more than 17 times what they started with. Same rule, same 4%, wildly different endings, decided almost entirely by which sequence of history you happened to draw. The rule “working” 97.9% of the time hides the fact that most of the time it works by leaving a fortune untouched, and a small fraction of the time it fails outright. It rarely lands neatly on zero.

If watching that play out beat by beat is more your speed than reading survival tables, our market-timing game deals you a lifetime of these draws one at a time, on the same real history, so you feel how much the order matters rather than just taking our word for it.

The 30-year fine print

Every number above is a 30-year test, because that is the question Bengen asked and the horizon the rule was built for. Early retirement quietly changes the question. Retire at 45 and you may be funding 45 or 50 years, not 30, and the failure rate at a fixed 4% rises as the horizon lengthens, because there are simply more years for a bad sequence to strike and less runway to recover before the money is needed. The 4% rule is not wrong for early retirees; it is answering a shorter question than they are asking. That gap is exactly where a flexible spending plan, a lower starting rate, or a bit of ongoing income does the most work.

Methodology, stated plainly

Every figure here comes from one script (scripts/four-percent-article-stats.mjs) run against one bundled dataset: the Shiller US real total-return series, monthly, CPI-deflated, dividends reinvested, February 1871 through mid-2026. A window is 360 consecutive months. We start each window with a balance of 1.0 and, every month, apply that month’s real market return and then withdraw one-twelfth of the annual spend, where the annual spend is a fixed percentage of the original balance held constant in real terms for all 30 years. A window fails if the balance ever reaches zero before the 360 months are up. We test every month that has a full 30 years of data after it, which is 1,507 windows. There are two honest limits. This is all-US-equity, so it slightly overstates the volatility a real stock-and-bond retiree would feel and understates nothing about the growth. And it is US-only history, the market that won; a century and a half of American returns is a success-biased sample, and the future is under no obligation to match it. Read these as a well-documented record of the past, not a promise.

The honest takeaway

The 4% rule earns its fame. Across 155 years and 1,507 starting points it worked 97.9% of the time, and below 4% it barely ever failed at all. But the rule is a rough summary of a distribution, not a guarantee, and the 2.1% of the time it broke, it broke for a reason worth respecting: retirees who drew a bad sequence early, whose fixed withdrawal kept draining a portfolio that could not recover in time. If you take one thing from the data, let it be that “4%” is a reasonable place to start a conversation about your own plan, not a number to bank a 45-year retirement on without checking it against your own horizon, your own allocation, and the full range of ways history could have gone.

Frequently asked