The "4% rule" is the most famous rule of thumb in retirement planning. It says that if you withdraw 4% of your portfolio in your first year of retirement, and then increase that amount each year in line with inflation, your money should last at least 30 years. It comes from financial planner William Bengen's 1994 research, which tested historical US market returns.

We re-ran the test ourselves using Robert Shiller's S&P 500 data from 1871 to 2023, to see how often the rule held up — and, more usefully, when it didn't.

How the test works

For every possible retirement start year from 1871 to 1993 (123 start years, each with a full 30 years of data):

  • The retiree starts with a portfolio invested 100% in the S&P 500, with dividends reinvested.
  • At the start of year one they withdraw 4% of the starting value. Each later withdrawal is the same amount adjusted for that year's inflation, so their spending power stays constant.
  • The remaining balance earns that year's actual market return.
  • The retirement "succeeds" if money remains after 30 withdrawals.

This is a stricter test than Bengen's original, which assumed a mix of stocks and bonds. A 100%-stock portfolio has higher expected returns but much bigger swings.

The results

Initial withdrawal rate Lasted 30 years Failed Median ending balance (in real terms)
3.0% 100% of start years none 3.9× the starting value
3.5% 100% none 3.4×
4.0% 97.6% 3 3.0×
4.5% 91.9% 10 2.6×
5.0% 80.5% 24 2.1×
6.0% 69.9% 37 1.3×

At 4%, the portfolio survived 30 years in 120 of 123 historical start years. At 3.5% or below, it never failed. Beyond 4%, failure rates climb quickly: at 5%, roughly one retirement in five ran out of money.

The three failures at 4%

Retirement began Money lasted What happened
1929 24 years Retired straight into the Great Crash; the portfolio lost most of its value in the first three years while withdrawals continued.
1966 26 years A weak, inflationary decade followed. Stocks went nowhere in real terms while inflation pushed withdrawals up every year.
1969 27 years Similar to 1966: a poor first decade, made worse by the 1973–74 bear market.

The near misses tell the same story. Retirements beginning in 1965 and 1968 survived, but with only 2–4% of their starting real value left after 30 years. Starting in 1973, the money lasted, ending at about one-fifth of its starting value.

These are all examples of sequence-of-returns risk: poor returns in the first decade of withdrawals do far more damage than poor returns later.

What about retiring in 2000?

A retirement that began in January 2000 does not yet have 30 years of data, but it is on track to be one of the worst on record. After 23 years of 4% inflation-adjusted withdrawals, the portfolio held only about 21% of its starting value in real terms — two bear markets in its first nine years and modest real returns overall did the damage. With seven more years of withdrawals to go, it would need strong markets to last the full 30 years.

The other side: most retirees underspend

The failures get the attention, but the median result is just as important. In the typical historical case, a retiree following the 4% rule ended 30 years later with around three times their starting wealth, after inflation. That suggests most retirees following a fixed rule spent much less than they could have — the price of protecting against the rare bad sequence.

How to use this in your own plan

  1. Treat 4% as a starting point, not a guarantee. It held in about 98% of historical cases for a US stock portfolio over 30 years. Retirements longer than 30 years, portfolios with higher fees, or markets other than the US can need a lower rate.
  2. Be flexible. The rule assumes you never cut spending. In practice, trimming withdrawals for a year or two after a large market fall greatly reduces the risk of running out.
  3. Pay attention to your first decade. If the first years of retirement bring a deep bear market, that is the moment to reassess — not year 25.
  4. Consider valuations. All three failures began when the market was expensive by historical standards, with a CAPE ratio between 21 and 27 against a long-run median of about 16.5. (January 2000, the start of the on-track-to-fail retirement above, had a CAPE of almost 44.) See how CAPE relates to future returns.
  5. Model a range, not a single line. Our calculator shows a 10th-to-90th percentile range from 1,000 simulated paths. Use the 10th-percentile figure to judge whether your plan survives a poor run.

Notes and limitations

Data: Robert J. Shiller, U.S. Stock Markets 1871–Present, via datasets/s-and-p-500. Annual January-to-January total returns calculated from monthly average prices with dividends reinvested; inflation from Shiller's CPI series. The test uses a 100% S&P 500 portfolio and ignores taxes, fund fees and bonds. US stocks had one of the strongest long-run records of any market, so results for other countries would likely be worse. Calculations are our own.

This guide is educational and not investment advice. Past performance does not guarantee future results.