How the backtest works
Instead of assuming one steady return, the backtest replays real history. For every year from 1928 onward, it imagines someone retiring that January with your savings and withdrawal rate, takes the withdrawal, raises it each year by that year's actual inflation, and grows what is left by the actual returns of US stocks and 10-year Treasury bonds in your chosen mix. A start year counts as a success if money remains at the end of the retirement period.
With $1 million, a 4% first-year withdrawal, a 60/40 mix, 0.1% fees and a 30-year retirement, the money lasted in 64 of the 69 start years from 1928 to 1996, about 93%. The five failures are retirements that began between 1964 and 1969, when high inflation in the 1970s pushed withdrawals up while stocks and bonds struggled. Dropping the rate to 3.5% makes every start year succeed.
Reading the chart
The chart shows how much money was left at the end of each retirement, in the buying power of the year it began. Most start years ended with more than they started with, often far more, which is why many retirees following the 4% rule die with large balances. The dips show the dangerous periods: retiring just before a crash or a long stretch of high inflation.
History only offers about 70 overlapping 30-year periods, so treat the success rate as a guide rather than a precise probability. Spending flexibility, such as skipping inflation raises after a bad year, makes any rate safer than the backtest shows.
Common questions
Did the 4% rule ever fail?
For a 60/40 portfolio over 30 years, five start years between 1964 and 1969 ran out, the worst after about 25 years in this backtest; most periods ended with money to spare.
What withdrawal rate never failed?
Around 3.5% survived every 30-year period since 1928 for a 60/40 mix in this data.
Why include bonds?
Bonds cushion stock crashes. Very high stock shares often end richer but can fail harder after early crashes.
Sources
This calculator gives estimates for planning. Real returns, rates and prices change, so treat the result as a guide rather than a promise. Read how we build and check our calculators.