The short answer
There’s no single right number, but the trade-off can be measured. With $1 million, 60% stocks and spending that rises with inflation for 30 years, the most you could have spent and lasted in all 69 start years since 1928 was about $37,300 a year. Accepting a plan that lasted in 95% of starts raised that to $39,000; 90%, to $41,700; 80%, to $46,400. Going from 90% to 100% cost about $4,400 a year per $1 million.
“Success” here means every year’s withdrawal was paid in full for all 30 years; one short year counts as a failure. A lower target is easier to live with if you’re willing to adjust spending, which is what the rest of this page tests.
What “success” means, and what it doesn’t
At $40,000 a year (the 4% rule), the plan lasted in 65 of 69 start years: a 94% historical success rate. That doesn’t mean a 6% chance of going broke. The 4 starts that fell short were 1965, 1966, 1968 and 1969, and they ran out in years 26 to 28. Each also assumes spending that never changed while savings shrank for decades, which few retirees would actually do.
Consecutive start years share most of their market history, so these aren’t independent coin flips. They show how a plan held up through real events, not a probability of the future. See how historical testing works.
What each step toward 100% costs
| Target | Start years that lasted | Highest spending per $1M | Extra spending vs the step above | Start years that fell short (year money ran out) |
|---|---|---|---|---|
| 100% | 69 of 69 | $37,300 | — | None |
| 95% | 66 of 69 | $39,000 | +$1,700 | 1965 (year 30), 1966 (year 28), 1968 (year 30) |
| 90% | 63 of 69 | $41,700 | +$2,700 | 1964 (year 30), 1965 (year 26), 1966 (year 24), 1967 (year 30), 1968 (year 26), 1969 (year 25) |
| 80% | 56 of 69 | $46,400 | +$4,700 | 13 starts, from 1929 to 1973 |
The first steps buy the most spending. The last step, from 95% to 100%, costs only $1,700 a year, because only a handful of late-1960s starts need protecting. Each amount is the exact edge: $100 more and one more start year falls short.
- Supported the 90% target or more
- Below the 90% target
- Below the 95% target
When plans fail: late, not early
| Spending per $1M | Lasted | Earliest year money ran out | Median year among failures |
|---|---|---|---|
| $40,000 | 65 of 69 | Year 26 | Year 28 |
| $45,000 | 58 of 69 | Year 21 | Year 25 |
| $50,000 | 50 of 69 | Year 18 | Year 23 |
At $40,000 a year, every failure came after year 25. Spend more and failures come sooner: at $50,000 the earliest came in year 18. See how 5%, 6% and 8% held up. Most of them came from the same window, retirements that began just before the 1970s’ high inflation. See the worst year to retire and what happened to 1966 retirees.
Historical success vs simulated odds
The planner shows two numbers for the same plan: the share of real start years that lasted, and the share of 3,000 simulated retirements that lasted. They usually tell a similar story, but they aren’t the same. $37,300 a year lasted in all 69 real start years, yet in 96.1% of simulated ones. At $40,000, the figures were 94% historical and 92.9% simulated.
The simulations stitch together real five-year stretches of history in new orders, so they can string together bad years worse than any real sequence. That’s why a plan can survive every start year on record and still fall short of 100% in simulation. For when simulation is the stricter number and when it’s the more generous one, and why other calculators can differ far more, see Monte Carlo vs historical testing.
Flexibility changes what “failure” means
A failure in these tests assumes spending never changes. Real retirees adjust. With guardrails (trim spending about 10% after big drops, raise it after strong years, never below 80% of the starting amount), the same plans held up far better:
| Starting spending per $1M | Fixed spending lasted | With guardrails lasted | Lowest year with guardrails |
|---|---|---|---|
| $40,000 | 65 of 69 | 69 of 69 | $32,000 |
| $45,000 | 58 of 69 | 68 of 69 | $36,000 |
| $50,000 | 50 of 69 | 62 of 69 | $40,000 |
So the more useful question is often not “what success rate?” but “how low would I be willing to go in a bad stretch?” See the 4% rule vs guardrails. In the planner, your own guardrails and spending floor are part of Plus ($69 a year); the guardrails calculator is free.
How to choose your target
This is a judgment, not a formula. Some things point toward a higher target, closer to 100%:
- Little room to cut spending, because most of it is essentials.
- Most of your income comes from savings rather than Social Security or a pension.
- A long retirement, or a lower share in stocks.
And some make a lower target, around 80% to 90%, easier to live with:
- Social Security or a pension covers your essential spending.
- You’re willing to trim spending for a few years after a big drop.
- You plan to check in every year and adjust before small problems grow.
These figures are for 30 years and 60% stocks. The dollar amounts shift with your retirement length and mix: try a 90% or 95% standard for your own numbers in the safe withdrawal rate calculator.
Common questions
What is a good success rate for a retirement plan?
There is no single right number, but you can measure the trade-off. With $1 million, 60% stocks and fixed spending over 30 years, $37,300 a year lasted in every start year since 1928, while $41,700 lasted in about 90% of them. A lower target is easier to live with if you are willing to trim spending in bad years.
Does a 90% success rate mean a 10% chance of running out of money?
No. It means the plan paid every year in full in about 90% of past start years (or simulated ones), with spending that never changed. In history the failures came late, came from the same late-1960s window, and assumed the retiree never adjusted. It measures how much margin a plan has; it is not a forecast.
Why doesn’t my plan show 100% in simulated markets when it lasted in every real start?
Simulated markets stitch together real five-year stretches of history in new orders, so they can produce runs of bad years worse than any real sequence. In our test, $37,300 a year lasted in all 69 real start years but in 96.1% of 3,000 simulated retirements.
Is aiming for 100% a mistake?
Not necessarily, but it has a cost. In this test, a plan that lasted in every start year allowed about $4,400 a year less spending per $1 million than one that lasted in about 90% of them. Most start years ended with more money than they began with, so a 100% target means spending less in the many good outcomes to protect against the few worst ones.
Test your own number
Your plan’s historical success and simulated odds are free in the retirement planner: enter your savings, spending and Social Security and see how often your plan lasted. To compare targets for your retirement length and mix, use the safe withdrawal rate calculator.
How these figures are calculated
- $1,000,000 of savings, 60% S&P 500 stocks (dividends reinvested) and 40% 10-year Treasuries, rebalanced yearly. The first year’s spending rises with inflation every year and never changes otherwise. Withdrawals come out at the start of each year. No taxes, fees or other income.
- Historical results use all 69 complete 30-year retirements from 1928 to 1996. A start “lasted” if every year’s withdrawal was paid in full. Simulated results are 3,000 retirements built from random five-year blocks of the same history, the planner’s setting.
- Spending targets are the highest amounts, in $100 steps, that met each standard. Guardrails use the planner’s default rules, based on Guyton and Klinger (2006); fixed spending follows William Bengen’s 1994 approach.
- Data: stock, bond and inflation returns since 1928, from Aswath Damodaran (NYU Stern) and the U.S. Bureau of Labor Statistics. Results are hypothetical and based on U.S. history; they are not a prediction or financial advice. See the full methodology.