How the test works
We run your savings through every stretch of real market history the data allows. The returns and inflation figures come from Aswath Damodaran of NYU Stern, whose yearly series for US stocks, bonds and bills runs from 1928 to 2025. For a 30-year plan that gives 69 start years, 1928 to 1996. For each one we go a year at a time:
- At the start of the year we take out your spending, raised by that period's actual inflation, minus any other income you entered (also raised with inflation) once it starts.
- What is left earns that year's stock return on your stock share and the 10-year Treasury bond return on the rest, minus your fees. The mix is reset every year.
- If a year's withdrawal is more than what is left, the money ran out that year.
The result counts the start years in which the money lasted, names the worst one, and shows what was left at the end in today's dollars.
A worked example
Take $1,000,000 in savings, $40,000 a year (a 4% starting withdrawal rate), 30 years, 60% in stocks and 0.5% a year in fees. The money lasted in 62 of 69 historical periods. It ran out for start years 1962 and 1964 to 1969. The worst was 1966: the money ran out in 1989, year 24 of 30.
The median ending balance, counting the periods that ran out as $0, was $1,132,163 in today's dollars. The thinnest finish among the periods that lasted was $31,621, for the 1973 start. With no fees, the same plan lasted in 65 of 69 periods.
What is the 4% rule, and what is it not?
The 4% rule is a rule of thumb from historical tests like this one: take 4% of your savings in the first year, raise that dollar amount with inflation every year, and the money lasted 30 years in most past periods. It is a summary of history, not a promise. It assumes 30 years and a fixed mix of stocks and bonds, and it is usually quoted before fees and taxes.
On this data with 0.5% fees, 4% ran out in 7 of 69 periods. A longer retirement asks more of the same rate: $40,000 a year from $1,000,000 over 40 years lasted in 44 of 59 periods.
What is sequence of returns risk?
Once you are withdrawing, the order of returns matters as much as their average. A fall in the early years means selling more to fund the same spending, and what was sold is not there for the recovery.
In the example above, the 1969 start earned an average of 5.1% a year after inflation and fees over its 30 years, and the money still ran out in 1993. The 1955 start averaged only 2.5% a year and lasted, ending with $529,195. The difference is the first ten years: 1969's lost 2.8% a year after inflation, 1955's gained 6.9% a year.
How long will $1 million last in retirement?
It depends on how much you take out, how it is invested, and which markets you retire into. Here is $1,000,000 over 30 years, 60% in stocks, 0.5% fees:
| Yearly withdrawal | Lasted 30 years | Worst start year |
|---|---|---|
| $30,000 (3%) | 69 of 69 | 1965, lasted with $406,891 left |
| $40,000 (4%) | 62 of 69 | 1966, ran out in 1989 |
| $45,000 (4.5%) | 52 of 69 | 1966, ran out in 1985 |
| $50,000 (5%) | 45 of 69 | 1969, ran out in 1985 |
| $60,000 (6%) | 35 of 69 | 1973, ran out in 1985 |
A record of past periods is not a forecast of the next one.
Why test against history instead of an average return?
An average return hides the order of returns, and the order is what breaks a plan. History also carries real inflation: prices rose 12.3% in 1974 and 13.3% in 1979 in this data. The limit is that the periods overlap. The 69 thirty-year periods share most of their years, so they are not 69 independent tests.
What does a 90% success rate mean?
That the plan lasted in 90% of the historical periods tested, nothing more. It is not a 90% chance about your future: the next 30 years will not repeat any past period, and the record covers one country's markets. Lasting in every period is no guarantee either. The worst start years and the near misses say more than the headline count.
What this test leaves out
- Yearly data: one withdrawal at the start of each year and one return for the year.
- US stocks (the S&P 500 with dividends, filled in with other large-company indices before it existed) and 10-year Treasury bonds only.
- Taxes. If your withdrawals are taxable, they have to cover the tax too.
- Changes in spending. Cutting back after a bad year would last longer than this shows; large one-off costs do not show up.
- Pensions without inflation increases. Other income here rises with inflation every year.
- Exact inflation timing. Inflation is the yearly change in the consumer price index for all urban consumers. The dataset does not say whether each year compares annual averages or December to December, so treat it as close, not exact.
Past results do not predict the future. Data: Aswath Damodaran, NYU Stern, Historical Returns on Stocks, Bonds and Bills, 1928 to 2025.
Sources
- Historical Returns on Stocks, Bonds and Bills: 1928-2024 (Aswath Damodaran), checked 2026-10-27
- histretSP.xlsx (Aswath Damodaran): Excel workbook in the same dataset folder as the linked histretSP.xls, checked 2026-10-27
- Data history / Usage Rules (Aswath Damodaran), checked 2026-10-27
