The 4% rule in ten minutes
It is the most famous rule in personal finance: withdraw 4% of your capital in the first year of retirement, raise that amount with inflation every year after, and your portfolio will last thirty years. It is so famous that people call it "the rule" and leave it at that.
This page lays out what the rule actually says (it is subtler than the barstool version), where its numbers came from, what it quietly assumes, and why the state of the art now treats it as an excellent place to start and a bad place to stop. By the end you will know how to use it for what it is worth: a ballpark, not a plan. And if you want to see why that number holds mathematically, and what would move it, the arithmetic is taken apart layer by layer in Why 4%? The mathematical anatomy of the rule.
The mechanics, step by step§
Meet Camille, who retires with $1,000,000, invested 60% in global stocks and 40% in bonds.
- Year 1: she withdraws $40,000 (4% of 1,000,000).
- Year 2: inflation ran at 2.5%. She withdraws 40,000 × 1.025 = $41,000. It makes no difference whether her portfolio gained 15% or lost 20%: the withdrawal is the same.
- Year 3 and after: same logic, last year's withdrawal adjusted for inflation.
Three properties follow straight from that machinery, and they explain everything else in the subject.
Purchasing power is constant. That is the rule's great virtue: your standard of living never depends on the mood of the markets. It is an annuity contract you sign with your own portfolio.
The effective withdrawal rate, though, floats. If Camille's portfolio drops to $700,000 after a crash in year 2, her $41,000 is suddenly 5.9% of capital. The rule does not care, and that is exactly where the risk lives.
The "4%" is counted once and never again. The rate applies to the starting capital, on the day you leave, and to nothing else. Hence a well-known paradox: two identical neighbors, one who left in 2021 with $1M and draws $40k, the other who left in 2022 after a crash with $800k and draws $32k, take out different amounts although their portfolios are now worth roughly the same. That is not a footnote. It shows the rule is a simplification of something deeper: the withdrawal rate depends on the valuations you start from (Valuations, the CAPE, and what they say about your withdrawal rate).
Where the numbers came from§
The rule has two founding acts, covered in Bengen, the Trinity study, and the birth of the safe withdrawal rate.
Bengen, 1994. William Bengen replays every possible US retirement since 1926: leaving in 1926, in 1927, in 1928, and so on. For each vintage he computes how many years a 50/50 portfolio of stocks and government bonds would have lasted under a given indexed withdrawal. The result: the worst vintage in the record, 1966, right before fifteen years of flat markets and high inflation, carried an initial withdrawal of 4.15% for thirty years. Bengen would later name that floor SAFEMAX. The round "4%" comes from there. It is the rate of the worst case in American history, not an average: the median vintage carries close to 6%.
Trinity, 1998. Three Trinity University professors (Cooley, Hubbard, Walz) turn the approach into a grid of probabilities: for each withdrawal rate, allocation and horizon, what share of the historical windows survived? The famous cell: 4%, a 50/50 portfolio, 30 years, 95 to 96% success depending on the update. That study is where the idea of a "success probability" comes from, and it still shapes every simulator (Failure probability: reading it, choosing it, and not letting it run you).
What the rule assumes without saying so§
The marketing power of "4%" buried the long list of its assumptions. Each one deserves a look against your own situation.
| Implicit assumption | What your plan actually looks like |
|---|---|
| A 30-year horizon | Leaving at 40 means 45 to 55 years (Horizon, life expectancy, and 50-year retirements) |
| US stocks and bonds, 1926 to 1995 | The best market of the century, over its best century (Beyond the United States: Anarkulova, Cederburg and the world sample) |
| No fees | An index fund costs 0.1 to 0.4% a year, an insurance wrapper sometimes another 1% (Building it with US-listed ETFs: the retiree's shopping list) |
| No tax on withdrawals | Tax on a withdrawal is a spending item: count it (US taxes in the withdrawal phase) |
| Perfectly rigid spending, indexed to inflation | Real spending moves, and that is a margin you can use (Real spending in retirement (the retirement smile, Die With Zero)) |
| No other income, ever | Pensions, side work and inheritances exist (Pensions and other income in the plan) |
| The retiree runs the rule mechanically for 30 years | Nobody watches a portfolio melt without reacting (The psychology of spending: why spending it is the hard part) |
| Success = one dollar left on the final day | Ending at 82 with $3,000 counts as "success" for the simulator alone |
None of these assumptions kills the rule as a ballpark. Together, though, they explain why the number that comes out of a serious look at your situation can land well away from 4%, in either direction.
What the state of the art says§
Thirty years of research have sharpened the picture, and the results converge on four points.
1. Over 30 years in the United States, the rule did hold. Including through 2000-2009, the worst modern decade for stocks: the retiree of January 2000, who took two 50% crashes in their first ten years, was still solvent in 2024 under the 4% rule. The historical frame was not absurd.
2. Over a long horizon, 4% is too high for a rigid rule. That is the central result of the Early Retirement Now series (ERN's Safe Withdrawal Rate series: a reader's guide): over 50 to 60 years, the rate that would have survived every US vintage falls toward 3.25 to 3.5%. The failure probability of a rigid 4% over 50 years, measured on American history, runs around 10 to 20% depending on the allocation. Too much for a life plan.
3. Outside the United States it is worse. On the global sample, the broad sample of 16 developed countries since 1870 (Beyond the United States: Anarkulova, Cederburg and the world sample), the "safe" rate of a domestic 60/40 sits well below 4%. Twentieth-century France, Japan, Germany and Italy handed their retirees sequences that American history simply does not contain. A globally diversified investor today sits somewhere between those two worlds.
4. Starting valuations move the rate. Every one of the worst vintages began in an expensive market, with a high CAPE. Modern rules therefore condition the initial rate on valuations (Valuations, the CAPE, and what they say about your withdrawal rate, CAPE-based rules: tying the withdrawal to valuations (ERN)). It is one of the best established improvements in the field.
So, keep it or toss it?§
Keep it, in its place. The 4% rule is excellent in three roles and fails in a fourth.
Excellent as a unit of measure. "This annual expense needs 25 times its own size in capital" is the most useful mental reflex in the field. A recurring $100 a month "costs" $30,000 to $40,000 of capital, and that one calculation changes how you weigh a subscription, a car, a move.
Excellent as a starting point for sizing. Aim at 25 times your spending, then refine with a real model (Using the FIRE simulator) and your own situation. That is the right order. The mistake is not starting at 4%, it is stopping there.
Excellent as a sanity check on a sales pitch. When someone promises "8% a year, risk free", the rule tells you instantly that this is twice what the best market in history sustained under a rigid rule over thirty years. A reflex worth having.
Bad as an actual withdrawal strategy. Nobody should run a blind indexed withdrawal mechanically for forty years. Not because the rule is wrong, but because it ignores the information that keeps arriving: markets, health, real spending, valuations. The whole withdrawal-strategies part of this book (Withdrawal strategies: the map before the territory) is about what goes in its place, rules that listen to the portfolio (Guyton-Klinger: the original guardrails, their power and their limits, Modern guardrails (Morningstar): the state of the art, Amortization-based withdrawal (ABW/TPAW): the actuarial approach). At equal capital they fund more spending, or protect the same standard of living far better.
The essentials§
- The rule: 4% of the starting capital, then the same amount indexed to inflation; never once broken over 30 years in Bengen's US data.
- The "4" comes from the worst American vintage, 1966, and not from an average; it was already a cautious number... inside its own frame.
- Outside that frame (a long horizon, the whole world, fees, taxes, high valuations), the equivalent rigid rate is closer to 3 to 3.5%.
- Flipped into the "25x multiple", it is the best mental reflex in the field.
- As an actual strategy it is beaten by the adaptive rules: read Withdrawal strategies: the map before the territory before you carve 4% into your plan.
Going further§
- William Bengen, "Determining Withdrawal Rates Using Historical Data", Journal of Financial Planning, 1994.
- Cooley, Hubbard & Walz, "Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable", AAII Journal, 1998 (the Trinity study, Bengen, the Trinity study, and the birth of the safe withdrawal rate).
- Early Retirement Now, SWR Series parts 1 and 26 ("Ten Things the Makers of the 4% Rule Don't Want You to Know"): the most complete modern critique (ERN's Safe Withdrawal Rate series: a reader's guide).
- Morningstar, The State of Retirement Income (annual): the recommended rate, recomputed every year with forward-looking returns (Modern guardrails (Morningstar): the state of the art).
- Bengen himself, in A Richer Retirement (2025) and in recent interviews. He now finds 4% too conservative and puts forward 4.7% for a portfolio of five equity styles, while reminding everyone that his rule only holds inside the US frame over 30 years. ERN's Part 65 disputes that lift (Factors (Fama-French, value, momentum) in the withdrawal phase). The two moves, his upward and the "whole world, long horizon" research downward, show how much the number depends on the frame.