Safe withdrawal rates: the Trinity study, the 4 percent assumptions and international data

What you will learn

• What the Trinity study measured, with which data, horizon and portfolio mix.
• The hidden assumptions behind "4 percent": fixed real withdrawals, 30 years, US returns, no costs.
• Why a 95 percent success rate is not the same as "safe".
• What international and German return history (Pfau, 2010) imply for a euro investor.

Explanation

A safe withdrawal rate (SWR) is the percentage of the starting portfolio withdrawn in year one and raised with inflation every year, such that the portfolio historically did not run out within a chosen horizon. It is a historical observation, not a forecast.

Bengen (1994) tested US stock and bond returns from 1926 for 30-year retirements and found that about 4 percent survived every starting year with 50 to 75 percent equities. The Trinity study (Cooley, Hubbard and Walz, 1998, "Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable") used US data from 1926 to 1995, five portfolios from 100 percent stocks to 100 percent bonds, horizons of 15 to 30 years and rates from 3 to 12 percent. For inflation-adjusted withdrawals of 4 percent over 30 years, the reported success rates were 95 percent for 100 percent stocks, 98 percent for 75/25 and 95 percent for 50/50. Rates of 5 percent and more failed much more often, and bond-heavy portfolios failed more often at every rate.

The success rate is the share of historical starting years in which the portfolio still had money at the end of the horizon; the failure rate is the rest. "Success" includes ending with €1. The starting years overlap heavily, so 70 years of data contain only a handful of independent 30-year periods. And a retiree gets one draw, so 2 to 5 percent failure is not small.

Assumptions inherited with 4 percent:

• A 30-year horizon. An early retiree at 45 may need 45 to 50 years; longer horizons mean lower surviving rates.
• Withdrawals fixed in real terms, never adjusted to markets (lesson 4).
• US returns, the most successful stock market of the twentieth century.
• Index returns with no fund costs, trading costs or taxes; each percentage point of yearly cost reduces the sustainable rate by a similar amount over long horizons.
• Annual rebalancing.

International evidence. Wade Pfau (2010, "An International Perspective on Safe Withdrawal Rates") applied the same method to 19 developed countries using Dimson, Marsh and Staunton data from 1900. He reports the SAFEMAX, the highest rate that survived every 30-year period, for a 50 percent stocks / 50 percent bills portfolio. For the US it was 3.96 percent (worst start 1937). For Germany it was 1.01 percent (worst start 1914), because two world wars and the 1923 hyperinflation destroyed nominal assets and interrupted markets; at 4 percent, over half of German 30-year periods in that data set failed. With a 50/50 mix, 4 percent survived every period in no country; it held in a few (including the US and Canada) only if a 10 percent failure rate was accepted.

For a German investor in 2026 the lesson is structural: home bias and nominal bonds in a currency that can fail are the vulnerabilities; a global equity portfolio, inflation-aware fixed income and a withdrawal rule that can bend (lessons 4 and 9) address them. Many practitioners start early retirements at 3 to 3.5 percent rather than 4 percent; treat that as a convention, not a guarantee.

Worked example

Hypothetical example: €800,000, withdrawals fixed in real terms, no fees or taxes. The euro amounts:

• 4.0 percent: €32,000 per year (€2,667 per month).
• 3.5 percent: €28,000 (€2,333).
• 3.25 percent: €26,000 (€2,167).

Sensitivity check: if the real return were constant at r, a 4 percent real withdrawal lasts n = −ln(1 − r / 0.04) / ln(1 + r) years:

For 2 percent: 1 − 0.02 / 0.04 = 0.5; −ln(0.5) = 0.693; ln(1.02) = 0.0198; 0.693 / 0.0198 = 35.0 years. One point of real return moves the horizon by a decade. Excluded: fund costs (subtract them from r), taxes on gains (lesson 7) and sequence effects (lesson 3).

Common mistake

Practice activity

On paper: Mira retires at 50 with €1,200,000 in a global 70/30 portfolio costing 0.3 percent per year. (a) Yearly amounts at 4.0, 3.5 and 3.0 percent? (b) With the constant-return formula and a net real return of 2.5 percent (2.8 percent minus costs), how long does a 4 percent real withdrawal last? (c) Name two reasons why the Trinity results overstate her safety.

Key takeaways

• Trinity and Bengen are US, 30-year, zero-cost historical back-tests; 4 percent is their result, not a law.
• Success rates count any positive ending balance and rest on few independent periods.
• Pfau's international data show German history far below 4 percent; global diversification and flexible rules are the response.
• Longer horizons, costs and taxes push a prudent starting rate below 4 percent.

Next step

Lesson 3 shows why two retirees with the same average return can end very differently: sequence-of-returns risk.