compoundsimThe 4% Rule: Failure Rates by Withdrawal Rate and Years
The 4% rule says savings last 30 years if you withdraw 4% in year one and raise it with inflation. We test 3% to 6% over 20 to 40 years in 1,000 futures.
Published · Updated
The 4% rule says savings last 30 years if you withdraw 4% in year one and raise it with inflation. We test 3% to 6% over 20 to 40 years in 1,000 futures.
Published · Updated
The 4% rule is a rule of thumb for retirement: withdraw 4% of your savings in the first year, raise that amount with inflation every year after, and the money should last 30 years. Applied to a mix of 60% stocks and 40% bonds across 1,000 futures drawn from a model fitted to past data (the S&P 500 since 1993, US aggregate bonds since 2003), it ran out of money within 30 years in 5.9% of them (the failure rate). At 3% that fell to 1.5%; at 5% it rose to 16.8%. Stretch the withdrawals to 40 years and even 4% failed in 12.5%. The 4% rule is a 30-year rule; change the span and the answer changes.
The rule grew out of US research in the 1990s. In 1994 the financial planner William Bengen used US stock and bond data from 1926 onward to find the withdrawal rate that lasted at least 30 years whatever year a retirement began, a safe withdrawal rate. His answer was about 4% of the savings in the first year. In 1998 three researchers at Trinity University in the US tabulated how often savings lasted for each withdrawal rate and span, in what is known as the Trinity study.
The rule rests on three assumptions that are easy to miss:
Sources: W. P. Bengen, "Determining Withdrawal Rates Using Historical Data," Journal of Financial Planning, October 1994; P. L. Cooley, C. M. Hubbard and D. T. Walz, "Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable," AAII Journal, February 1998.
We ran 1,000 paths (possible futures) the way the withdrawal simulator does, with these settings:
The monthly returns come from a model fitted to the full history. The model moves between a calm regime, when prices move gently, and a wild regime, when stocks swing more and tend to fall. Every withdrawal rate uses the same 1,000 paths with the same months in the same order, so only the amount withdrawn differs.
Here the failure rate is the share of paths whose money ran out before the end: 100% minus the success rate the simulator shows (the share of paths with money left).
| Withdrawal rate | Failure rate over 30 years | Median balance after 30 years (start = 100%) |
|---|---|---|
| 3% | 1.5% | 312% |
| 3.5% | 2.9% | 270% |
| 4% | 5.9% | 227% |
| 4.5% | 11.5% | 186% |
| 5% | 16.8% | 150% |
| 6% | 32.5% | 68% |
Withdrawing 4% for 30 years gave a failure rate of 5.9%: small, but not zero. Raising the rate a little raises the failure rate a lot. The extra withdrawn each year is small, but it adds up over 30 years, and more has to be sold when markets are down.
Lowering the rate lowers the failure rate, though even 3% does not bring it to zero.
At the same withdrawal rate, the number of years makes a large difference.
| Withdrawal rate | 20 years | 30 years | 40 years |
|---|---|---|---|
| 3% | 0.2% | 1.5% | 3.6% |
| 4% | 1.0% | 5.9% | 12.5% |
| 5% | 4.2% | 16.8% | 26.1% |
Over 20 years, even 5% failed in only 4.2% of paths. Over 40 years, 4% failed in 12.5%. The longer the span, the more chances there are to hit a bad stretch, and the more is withdrawn in total.
The earlier someone retires and the longer they draw down, the further 4% moves from the assumption the research made. To keep a 40-year plan below the failure rate of 4% over 30 years (5.9%), the table takes the rate down to 3% (3% over 40 years: 3.6%).
At 4%, the share of paths that had run out was 0.0% by year 10, 1.0% by year 20, 3.0% by year 25 and 5.9% by year 30. Most of the paths that ran out did so after year 20. Most of them had weak markets in roughly the first ten years of retirement. The withdrawals in those years left less to recover with, and the balance wore down over many years after. Why the timing of a crash matters so much is explained in "Sequence of returns risk."
On the other hand, most of the paths that lasted ended with more than they started with. After 30 years at 4%, the balances (today's money, start = 100%) were:
| 10th percentile | 25th percentile | Median | 75th percentile | 90th percentile |
|---|---|---|---|---|
| 19% | 100% | 227% | 480% | 829% |
At the median, 227% of the starting balance was left after 30 years of withdrawals. The 4% rule is not the rate that spends the money down to zero in an average future. It was chosen to survive futures where returns came in an unlucky order, so in many futures money is left over. The other side of that: the 10th percentile (about the 100th of the 1,000 paths from the bottom) was 19% after 30 years, so one path in ten ended at or below it (including the paths that ran out).
The withdrawal simulator starts with the settings used here: 60% stocks and 40% bonds, a 4% withdrawal rate and 30 years. Its "Success rate by withdrawal rate" chart compares 2% to 8% over the same paths.
How the simulator builds its 1,000 paths is explained in "How Monte Carlo Simulation Works for Investing, Without Math."