Your FIRE number is the portfolio you would need to stop working, and the usual shortcut is to divide yearly spending by a withdrawal rate. At 4%, that is the same as multiplying spending by 25. If you spend 40,000 a year, the number is 40,000 ÷ 0.04 = 1,000,000. The 4% comes from studies of United States market history, each looking at retirements of around 30 years. It is a historical rule of thumb. It was never a guarantee, and several of its built-in assumptions will not hold for every reader.

The arithmetic

The formula is the number of years of spending you want to hold, expressed as a rate:

FIRE number = annual spending ÷ withdrawal rate

Withdrawal rate Multiple of spending FIRE number for spending of 40,000
3.0% 33.3× 1,333,333
3.25% 30.8× 1,230,769
3.5% 28.6× 1,142,857
4.0% 25× 1,000,000
4.5% 22.2× 888,889
5.0% 20× 800,000

Two consequences follow. Every 1,000 of yearly spending adds 25,000 to the target at 4%, so the number is driven more by spending than by anything about markets. And the number is very sensitive to the rate: moving from 4% to 3.5% adds 142,857, about 14%. The FIRE calculator does this sum, and the years of saving it takes, with your numbers.

Where 4% came from

Bengen (1994). William Bengen’s article “Determining Withdrawal Rates Using Historical Data” appeared in the Journal of Financial Planning in October 1994. It took the historical record of US stocks, intermediate-term US Treasuries and inflation from 1926, from the Ibbotson yearbook, and asked how much a retiree could have taken in the first year, then raised each year with inflation, without running out. His main case was a portfolio of 50% stocks and 50% intermediate-term Treasuries, rebalanced, held in tax-deferred accounts so that capital gains taxes did not apply. His reading of the results: assuming at least 30 years of portfolio life is required, a first-year withdrawal of 4% “should be safe”. In no past case did it exhaust the portfolio before 33 years. A 4.25% first-year withdrawal could, in the worst past case, have run out in as little as 28 years. Bengen also found that a stock allocation between 50% and 75% did best in his data.

The Trinity study (1998). Philip Cooley, Carl Hubbard and Daniel Walz, finance professors at Trinity University, published “Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable” in the AAII Journal in February 1998. They used annual US returns from 1926 to 1995 for the S&P 500 and long-term high-grade corporate bonds. They tested five stock and bond mixes from 100% stocks to 100% bonds, payout periods of 15, 20, 25 and 30 years, and withdrawal rates from 3% to 12% of the starting portfolio, with and without inflation adjustment. They reported that 3% and 4% rates were very unlikely to exhaust a stock and bond portfolio in their data. The paper states that it did not adjust for taxes or transaction costs.

The popular label “the 4% rule” blurs these two different studies, with different data and different assumptions, into one number.

What “4%” did and did not claim

Read the studies and a precise statement emerges. In US data from the 1920s onwards, a retiree who withdrew 4% of the starting portfolio in year one, raised the amount by inflation every year afterwards, and held a mix of stocks and bonds, would not have run out of money within 30 years in the historical start years tested. Each phrase in that sentence is a limit:

  • A fixed horizon. About 30 years. A 40-year retirement is a different problem.
  • US data. One country’s history, which includes the country that did unusually well in the twentieth century.
  • A portfolio mix. Mostly 50% or more in stocks. A lower stock share changes the result.
  • Constant real spending. The retiree never changes the spending plan, however the portfolio does. Real retirees usually adjust.
  • No investment costs, and no tax in the main cases. Bengen’s main case was tax-deferred; the Trinity authors did not adjust for taxes or transaction costs. A fund fee or advisory fee comes straight off the return.
  • History as the test. A rate that survives every start year in a sample has survived a limited number of overlapping periods. See the article on historical and Monte Carlo methods.

Longer horizons

If the portfolio earned a constant real return every year, the sustainable withdrawal rate for n years would be r ÷ (1 − (1 + r)^−n), with withdrawals at the end of each year. This table gives the result. It is a calculation with no volatility in it, so it is an upper bound on what a variable return could support.

Constant real return 30 years 40 years 50 years 60 years
2% 4.46% 3.66% 3.18% 2.88%
3% 5.10% 4.33% 3.89% 3.61%
4% 5.78% 5.05% 4.66% 4.42%

A related check: how long does a 4% withdrawal last, if the return is constant? The answer is −ln(1 − r ÷ 0.04) ÷ ln(1 + r) years. At a 1% real return it is 28.9 years. At 2% it is 35.0. At 3% it is 46.9. At 4%, where the return equals the withdrawal rate, the balance never falls. Someone retiring at 40 or 45 and planning to 95 or 100 faces a horizon of 50 years or more. At a 3% real return, the constant-return table above gives 3.89% for 50 years, before any allowance for volatility.

The same formula shows what costs do. If a fund charge takes a 3% real return down to 2% net, the number of years that a 4% withdrawal lasts falls from 46.9 to 35.0. These are illustrative figures with a constant return, not forecasts.

Investors outside the United States

The studies above used US data. Wade Pfau’s “An International Perspective on Safe Withdrawal Rates: The Demise of the 4 Percent Rule?” (Journal of Financial Planning, December 2010) repeated the exercise with 109 years of data for 17 developed countries and a 30-year retirement. Its summary says a 4% real withdrawal rate would have provided safety in only 4 of the 17 countries, and calls the rate “surprisingly risky”. The count depends on how “safe” is defined, and a looser definition, such as surviving 90% of start years, would give a different count. The paper’s point stands: the US was one outcome among several.

If your country has less history in a tool, the tool may default to US data. Glideworth’s UK households use a UK series that ends in 2020; households elsewhere default to US history unless you choose otherwise, as the limitations page notes.

Setting your own number with a spending-based projection

A multiple of one flat spending figure is a shortcut. A projection starts from what the plan actually contains:

  1. Spending by year, not a single figure. Spending commonly changes as people age, and housing, children and health costs rise and fall.
  2. Other income. Subtract pensions, state pensions and part-time earnings in the years they exist. A portfolio only has to cover the gap.
  3. Tax on the withdrawals taken. The tax depends on which account the money comes from.
  4. A horizon from retirement age to a life expectancy you choose, with a margin.
  5. A chance of success you accept, found by running the plan against many market sequences.

The number is then the smallest starting portfolio that meets the chosen chance of success. Glideworth’s solvers answer this directly: the required portfolio at a target chance of success, the most you can spend at that target, and the safe withdrawal rate for your plan, as described on the solvers page and in the guide to solvers and guardrails. Two caveats apply. The result is an estimate from your assumptions and from historical or random returns, and tax figures are estimates. And a different chance-of-success target gives a different number: the choice of 85% or 95% is part of the answer. The article on Monte Carlo results explains how to read that percentage.

A spending-based number also shows what a fixed rate hides. It can show that the plan is fine with a lower portfolio because a pension starts at 67, or that it needs more because spending rises in the first decade. It can also compare withdrawal rules, which is the subject of the article on withdrawal strategies.

Summary

  • FIRE number = spending ÷ withdrawal rate; 4% means 25×.
  • The 4% figure comes from US historical studies of roughly 30-year retirements (Bengen 1994; Cooley, Hubbard and Walz 1998), with fixed real spending and, in the main cases, no costs or tax.
  • Longer retirements, non-US markets, costs and taxes all push the rate below 4% in some calculations, and flexible spending pushes it up.
  • A projection that uses your own spending, other income and horizon replaces the multiple with a number that comes from your plan.

The free FIRE calculator works out the simple version of this question with your own spending, savings and withdrawal rate, and tests the number against US market history.

Filed under FIRE, Retirement, Withdrawal rates, Methodology

See where your plan lands.

Free to self-host today, with every feature. The hosted service opens later, at $9 a month or $90 a year.