The 4% rule, revisited — what 30 years of withdrawal data actually shows

Ask a stranger at a dinner party how much you can spend in retirement and, if they know anything at all, they will say four percent. Multiply your annual spending by 25, the shorthand goes, and that is the pot you need. Some call it the Rule of 300: take your monthly costs, multiply by 300, and there is your number.

What almost nobody has done is read the thing it came from. The 4% rule is not a law of finance handed down by a committee. It is the conclusion of a single fourteen-page article, written in 1994 by a financial planner in Southern California named William Bengen, working with a spreadsheet and a book of historical returns. The paper is famous. The arithmetic inside it is not — and almost everything people believe about the rule is a little bit wrong. It was never meant to be a rule. The four was never the typical answer; it was the worst answer in the worst case in the worst decade of modern American financial history. So we rebuilt the spreadsheet from scratch. This is what falls out.

A starting rate, not a yearly one: the 4% applies to the *first* withdrawal; after that you follow the dollars, not the percentage.

What Bengen actually did

Bengen's clients kept asking the same thing: how much can I take out each year without running dry? So he went to the data. He took the long run of American market history that begins in 1926, built a portfolio split between large US stocks and intermediate Treasury bonds rebalanced once a year, assumed it sat in a tax-advantaged account, and imagined a retiree who withdrew a fixed percentage in year one and raised that dollar figure by inflation every year after — exactly the way Social Security adjusts its payments.

The clever part was what he did next. He did not test one retiree; he tested a whole parade — one starting in 1926, one in 1927, one in 1928 — each living through the actual returns and inflation of the thirty years that followed. For each starting rate he asked a brutal question: did anyone in this parade run out of money before thirty years were up? At a 4% starting rate, the answer was no. The highest rate that survived *every* start year came out to roughly 4.15%. Bengen rounded down and a rule was born; he later named the concept SAFEMAX, the maximum historically safe withdrawal rate.

Here is the first thing the rebuild makes vivid. Four percent was not the average outcome — it was the floor, defined entirely by the most unlucky retiree in the dataset: someone who retired into the stretch around 1966–69, just before a savage bear market and the long grind of 1970s inflation arrived together. That one cohort sets the number for everyone. For almost everyone else, four percent was wildly conservative — the typical safe rate sits closer to seven percent, and many cohorts could have spent eight, ten, even more and still died with a fortune.

Rebuilding the spreadsheet from scratch

The reconstruction is simpler than its reputation. You need three things: a year-by-year series of stock returns and a matching one of bond returns (the standard US data from 1926), a year-by-year series of inflation, and a starting allocation and withdrawal rate to test.

The engine is one line of logic, applied year after year. Start with a portfolio value. Subtract the chosen withdrawal. Let what remains grow or shrink by that year's blended return. Next year, raise the withdrawal by last year's inflation, subtract, apply the return, and carry on — thirty times, rebalancing to target each year. If the balance ever hits zero, that cohort failed. The famous four percent is just the highest rate at which no path in the parade ever touches zero.

To make the worst case concrete, here is the kind of ledger the spreadsheet produces for the stagflation cohort — a portfolio hammered while inflation forces the withdrawals up, clawing to the finish with almost nothing to spare. Figures are an illustrative reconstruction modelled from historical series, not a definitive record.

Note how the withdrawal column keeps climbing — from $40,000 to nearly $137,000 — even as the balance erodes in the final third. Bad markets and high inflation, arriving together, are what nearly broke this portfolio. It finished with roughly one year's spending left: the very definition of SAFEMAX.

The Trinity Study, and the switch from "floor" to "odds"

Four years later, three finance professors at Trinity University — Cooley, Hubbard and Walz — published the paper that spread the idea: the Trinity Study (1998). The method rhymed with Bengen's, but they made one consequential change of emphasis, and it is the source of most of the confusion around the rule to this day.

Bengen reported a *floor*: the single rate that never failed. Trinity reported *odds*: the percentage of historical periods in which a portfolio survived at a given rate and allocation — the portfolio success rate. At 4% over thirty years, a 50/50 portfolio succeeded in about 95% of windows; tilt to 75% stocks and it climbed toward 98%.

Read carefully: a 95% success rate does not mean you have a 95% chance of a comfortable retirement. It means that, across the overlapping thirty-year periods in *one country's* record, the portfolio lasted in 95 of 100. Those periods overlap heavily, so the "probability" is doing less work than it appears. It is a description of the past, dressed in the language of chance. The authors revisited the work through 2009 — capturing the dot-com bust and 2008 — and the headline conclusions barely moved.

The canonical "4% rule" cells are 50/50 and 75/25 at 4% (≈95% and ≈98%). Read this as "how often did this work in the past," not "what are my odds" — those are not the same sentence. Notice how success holds at 3–4% for stock-heavy and balanced portfolios but collapses above 6–7%, and for very bond-heavy mixes.

The American exception nobody mentions at dinner parties

Here is the bias baked into all of it. Bengen's data, the Trinity data, the figure your dinner-party stranger quotes — all of it rests on one country's twentieth century. And the American twentieth century was, for an investor, about as good as it gets: two world wars left US industry intact and dominant while rivals rebuilt from rubble.

In 2010 the retirement researcher Wade Pfau took 109 years of data across seventeen developed countries and asked whether the 4% rule would have worked outside America. His subtitle was deliberately provocative: "The Demise of the 4% Rule?" Even granting generous assumptions, a 4% inflation-adjusted withdrawal would have been safe in only a handful of the seventeen. In several European markets it failed in the majority of windows; in Italy a retiree using it would have run dry more often than not. American retirees did better than almost anyone, second only to Canada.

Three studies, three different "right" answers

Go looking for the safe withdrawal rate today and you will find three reputable sources giving three different numbers. The truth is gentle: they are answering three different questions.

Bengen's own answer has gone up. In a 2025 book, *A Richer Retirement*, he reran his life's work with a more diversified portfolio — adding mid-, small- and micro-cap US stocks, international equities and Treasury bills. Broader diversification lifted the worst-case floor to 4.7%, which he calls the Universal SAFEMAX. He is blunt that 4.7% is the *floor*, not the recommendation: the average safe rate across his cohorts is around 7%, and under today's conditions he suggests something closer to 5–5.5% for retirees with normal flexibility. Retirees clinging to 4%, he says, are probably cheating themselves. He also adds a two-factor adjustment: a rich stock market (high Shiller CAPE) and high expected inflation both pull the safe rate toward the floor.

Morningstar answers the question Bengen could not: what about the future? Its annual research runs forward-looking simulations rather than historical replays, feeding in current bond yields, valuations and expected inflation, then asking for a rate with a 90% modelled probability of lasting thirty years. That produces a more cautious, more volatile number — just 3.3% when yields were on the floor in 2021, drifting back up as yields recovered. Notably it finds the highest safe rates come from fairly *modest* equity weightings (30–50% stocks), because too much equity invites exactly the early-crash risk that breaks portfolios.

These are not three answers to one question. They are one answer each to three different questions — a worst-case floor, a probability of success, and a forward-looking forecast. Knowing which one you are looking at is most of the battle.

What the data fingerprints as the real risk

Run enough cohorts and a clear culprit emerges — and it is not general volatility. Portfolios failed because of two specific things happening *early*.

The first is sequence-of-returns risk. The order of returns matters enormously when you are drawing money out, even if the average is identical. A steep loss in the first few years, while you are also withdrawing, sells off shares at depressed prices that are never there to recover; the same loss late in retirement barely registers. Researchers call the danger zone the five years either side of your retirement date — the most fragile moment in a portfolio's life.

The second is sustained inflation, and on the historical evidence it is the deadlier of the two. A crash usually recovers within a year or two; prolonged inflation is corrosive because it forces the withdrawal up every single year, draining the portfolio fastest precisely when it can least afford it. The stagflation cohort that defines the original rule was killed less by the bear market than by the decade of rising prices that followed. Bengen, after thirty years with this data, names inflation outright as the retiree's greatest enemy.

So what should the number actually be?

The honest answer is that the question contains a hidden, wrong assumption. There is no single number, because a fixed withdrawal rate is an anchor, not an operating manual. It is a sensible place to *start* a plan and a terrible way to *run* one.

What thirty years of research converge on is that flexibility beats precision. A retiree willing to trim spending after a bad year, or skip an inflation raise occasionally, can start meaningfully higher than one who insists on the same cheque no matter what. Morningstar finds simple dynamic rules can lift a safe starting rate from around 3.9% to nearly 5.7%. The best-known approach, the Guyton-Klinger guardrails, sets upper and lower bounds on the withdrawal percentage and nudges spending only when you stray outside them. Your right number also bends to a handful of personal facts.

Two practical notes the original studies set aside but you cannot. The famous figures assume a tax-advantaged account, so a taxable account supports less real spending after tax. And they exclude outside income entirely — for most real retirees a state pension does much of the heavy lifting, so the portfolio only has to cover the gap, and a higher rate on a smaller required pot is far less frightening than it sounds.

The bottom line

The 4% rule is one of the most useful pieces of financial folk wisdom ever produced, and almost everything memorable about it is a simplification. It was never a rule — only the worst case in one favourable country's history, generalised into a slogan. As a *floor* it has proved remarkably durable; thirty years and several crashes later, the worst-case number has if anything edged up, to about 4.7%. As a *forecast* of what a typical retiree can spend it is conservative, leaving most people with a large unspent estate. As an *instruction* to be followed rigidly through a collapse and a decade of inflation, it is simply wrong, and the data says so plainly.

The most honest version of the rule is not a number at all. It is a starting point of somewhere around four to five percent for a thirty-year horizon — adjusted up or down for your real circumstances, and run with enough flexibility to ease off when the early years go badly. Use the 4% rule calculator and the safe-withdrawal-rate calculator as a starting point, keep watch on inflation and on the order of your returns, and remember which question you are asking. The worst-case floor, the odds of success, and the forward-looking forecast are three different things wearing the same four-percent mask.