Is the 4 percent rule safe? We graded it twenty times.

The 4 percent rule is the most quoted number in retirement planning, and its keepers no longer agree on it. Bill Bengen, who derived the rule in 1994, revised his figure to 4.7 percent in a book published last year; Morningstar's research team puts the safe starting rate at 3.9. The rule was calibrated to a 30-year retirement, but it gets quoted as a constant that holds at any age, so rather than argue for a better number we tested it the way people actually use it: three retirees, at 50, 65, and 75, twenty runs of 2,000 simulated lifetimes each, at withdrawal rates from 3 to 6 percent. The success rates below run from 8 percent to 96.

What the rule actually says

Withdraw 4 percent of your portfolio in the first year of retirement, then give yourself a raise each year matching inflation, and a portfolio at least half in stocks would have survived every 30-year retirement in US market history. That is Bengen's original finding, later confirmed by the Trinity study. Three of its lab conditions matter here. The horizon is exactly 30 years, no more, no less. The withdrawal is all-in: taxes, healthcare, everything you spend comes out of the 4 percent. And nothing else exists in the model: no Social Security timing, no account types, no Medicare at 65, no ACA before it.

Bengen was clearer about the horizon than the shorthand suggests. He tied the 4 percent figure to "a minimum requirement of 30 years of portfolio longevity," identified 3 percent as the level that kept a portfolio alive for 50 years, and counselled his own clients to withdraw no more than 4 percent "especially if they retire early (age 60 or younger)." The rule was a ceiling tied to a horizon rather than a number attached to an age, so the questions below are about what happens when the horizon moves.

Real households have all of those things, so we built one and started asking questions.

What does the 4 percent have to cover?

Whatever you decide it covers. The two common readings succeed 81 and 68 percent of the time. Our first retiree is a single woman of 65 in Florida with $1 million: $600,000 in a 401(k), $300,000 in a brokerage account carrying $150,000 of gains, and $100,000 in a Roth, all invested 60/40. She claims Social Security now, $24,960 a year. The engine prices her Medicare coverage at about $7,200 a year on top of whatever budget she picks, and taxes on top of that.

Read the rule as written, everything inside the 4, and her budget is $57,800: with healthcare added and near-zero taxes, the portfolio supplies $40,040 in year one, 4.0 percent on the nose. That plan succeeds in 81 percent of her simulated lifetimes. But almost nobody reads it that way. The common reading treats 4 percent of the portfolio as spending money and lets healthcare and taxes ride on top, which for her means a $65,000 budget and a true first-year draw of 4.7 percent. That plan succeeds 68 percent of the time. Stretch the spending-money reading to 5 percent and it drops to 49, a coin flip against her own longevity.

Budgets in today's dollars. The true draw adds healthcare and taxes and nets out Social Security.

The remaining runs use the common reading, since it is the one people actually live by.

Does it matter which account the million sits in?

More than most people expect: the rule prices a portfolio, but the tax code prices the accounts inside it. We gave the same retiree the same million in a single account and reran her $65,000 plan. Held entirely in a brokerage account with no unrealized gains, the plan succeeds 76 percent of the time and she pays about $2,000 of total income tax through age 87, her simulated life expectancy. Held entirely in a traditional 401(k), every withdrawal is ordinary income that also pulls her Social Security into taxability: $49,000 of federal taxable income in year one instead of zero, about $191,000 of tax through 87, and a plan that succeeds 62 percent of the time. Her actual mixed household, part brokerage and part 401(k), lands between those two results at 68. Same million, same budget, same markets; the only variable is which kind of account holds the money.

What happens to the rule at 50?

It becomes a coin flip. Our second retiree leaves work at 50 in Arizona with $1.5 million: $900,000 in a 401(k), $500,000 in a brokerage account with $250,000 of gains, $100,000 in a Roth. Four percent is a $60,000 budget, and she is in some ways the rule's best customer: her early tax bills are tiny, and because little of her spending counts as taxable income, the engine's ACA model covers her $15,000 marketplace premium almost entirely with subsidies, so her true draw starts barely above 4 percent. She fails anyway, in half of her lifetimes: 50 percent, with a third of them broke before 80.

The reason is the horizon. The rule was tested over 30 years, and a 50-year-old with average health needs the money for closer to 40. Social Security is 17 years away and, with a career cut short, comes in at $24,000. The sequence matters too: every bad market in her first decade is financed entirely by the portfolio, since there is no other income to lean on. Cutting the rate to 3 percent, a $45,000 budget, restores the plan to 86 percent. That is worth pausing on: 3 percent is the same figure Bengen named for horizons this long, and he reached it from the historical record in 1994 while this engine reached it from simulated lifetimes and a tax model in 2026. The cost of that safety shows up at the other end: her median estate at 95 is $1.7 million in today's dollars. At 50 the useful range is narrow, 3 percent against 4, and the $15,000 a year between them is why our early-retirement post spent its pages on structure rather than on a rate.

And at 75?

At 75 the rule is timid. Our third retiree is 75 in Florida with $800,000, a $36,000 Social Security benefit already in payment, and the same mixed account structure in miniature. Four percent of her portfolio, a $68,000 budget all told, succeeds in 89 percent of her lifetimes; even measured all-in, taxes and Medicare included, her draw is above 5 percent and the result holds. Raise the rate to 5 percent and the plan still succeeds 76 percent of the time, more often than the 65-year-old's plan at 4. At 6 percent it succeeds 61 percent of the time. The arithmetic is mortality's: the median 75-year-old is funding 14 years, not 30, and a rule calibrated to the longer horizon leaves spending on the table for the shorter one.

How much does the same plan change with the starting age?

To isolate age, we took the 65-year-old's exact retirement, the same mixed million and the same $65,000 budget, and started it at six different ages:

The same dollars and the same spending succeed 8 percent of the time started at 50 and 91 percent started at 75, the widest swing anything in this article produces. The mechanics are visible in the left half of the line: before 65 there is no Social Security check and no Medicare, so the portfolio carries the entire budget plus a marketplace premium, through a retirement that also has to last longer. (This is a different question from the 50-year-old above, who sized her spending to her own larger portfolio; here the budget stays fixed while the start date moves.)

Is there a safe rate for each age, then?

There is a curve, not a number. Here are all three retirees at withdrawal rates from 3 to 6 percent, still on the common reading:

At 3 percent the three curves start nearly together, at 86, 87, and 96, and then separate as the rate rises. The older the retiree, the gentler the slope: moving from 4 to 5 percent takes the 75-year-old from 89 to 76 percent, the 65-year-old from 68 to 49, and the 50-year-old from 50 to 24. And notice that the 50- and 65-year-old curves lie almost on top of each other at every rate they share, even though the younger retiree has half again as much money: the extra $500,000 offsets roughly what the extra fifteen years take away.

Why these numbers sit below the 95 percent you have read

The original studies asked what survived the historical record: every actual 30-year window of US returns, gross of taxes and fees, ending on schedule. The engine asks a different question. It simulates each lifespan instead of fixing 30 years, which flatters the 75-year-old and punishes the 50-year-old; it draws returns and inflation from forward-looking assumptions rather than the US postwar record, which was one of the best runs any stock market has had; and it charges taxes and healthcare, with medical costs inflating faster than everything else. If the next half-century repeats the last one, the numbers above are too harsh. An engine can be internally consistent and still wrong about the world, and the return assumptions are the hinge these numbers all swing on.

So is 4 percent safe?

It depends on the horizon, which is what the rule was calibrated to in the first place. Across these three it came out conservative for one, roughly right for another, and too high for the third, and the horizon explains each case.

Bengen's 4.7 and Morningstar's 3.9 are careful answers to the question both were asked, which assumed a 30-year retirement and none of the rest of it. Whatever number you start from, the useful step is checking it against the household that will actually live on it.