Not as a rule that held everywhere. Wade Pfau tested a 4% first-year withdrawal, raised with inflation, against 19 developed countries’ returns since 1900, with half in stocks and half in bills. In every country there were start years when the money ran out within 30 years. The US and Canada came closest, at 3.96%. Japan was lowest, at 0.26% (Pfau, Retirement Researcher). Two later analyses point the same way. These are findings about the past under the studies’ own assumptions, not forecasts.
In short
- The 4% rule is a finding from US returns since 1926 (stocks and intermediate Treasuries; Bengen, 1994).
- Across 19 countries, the highest withdrawal rate that survived every 30-year start ranged from 0.26% to 3.96%, with half in stocks and half in bills (Pfau).
- For the UK back to 1870, Monevator puts the best case at 3.1% for 30 years, and 2.4% for 50 (Monevator, 2025).
- A 2025 study of 38 developed countries put the rate for a 5% chance of running out at 2.31% for a 65-year-old couple (Anarkulova, Cederburg, O’Doherty and Sias).
- Pfau’s table has no fees, and Monevator ignores costs and taxes. Tax and fees both lower what a given withdrawal can pay for. We have not confirmed how Anarkulova et al. treat them.
Where 4% came from
William Bengen’s 1994 paper used US stock and intermediate-term Treasury returns from 1926, split half and half in his main example. The first year’s withdrawal was a share of the starting pot, and each later withdrawal was raised with inflation. He found that 4% had never exhausted a portfolio in under 33 years. At 4.25%, a portfolio could run dry in as little as 28 years (Bengen, Journal of Financial Planning, October 1994; reprint).
That is a careful result about one country’s 20th century, a century that went well for US investors. The 2025 UBS Global Investment Returns Yearbook puts US equities at 9.7% a year from 1900 to 2024, with inflation at 2.9% (summary edition, p. 6). That is about 6.6% a year after inflation (1.097 ÷ 1.029 − 1). For world equities over the same years, the figure after inflation is 5.2% (Cambridge Judge Business School).
What the international data found
Wade Pfau first ran Bengen’s test on 17 countries in the Journal of Financial Planning in December 2010 (abstract). He later updated it to 19 countries with Dimson, Marsh and Staunton’s data to 2010, for retirements starting from 1900 to 1981. His US figure (3.96%) sits just under Bengen’s 4% because the setups differ: Pfau used 50/50 stocks and bills with start years from 1900, Bengen stocks and intermediate Treasuries from 1926. The selection below is from that updated table. The last column is our arithmetic.
| Market | Highest safe rate | Starts where 4% failed | Pot (× spending) |
|---|---|---|---|
| United States | 3.96% | 1.2% | 25.3 |
| Canada | 3.96% | 1.2% | 25.3 |
| World (GDP-weighted) | 3.58% | 3.7% | 27.9 |
| Australia | 3.50% | 13.4% | 28.6 |
| United Kingdom | 3.36% | 24.4% | 29.8 |
| World excluding the US | 3.21% | 22.0% | 31.2 |
| Spain | 2.16% | 53.7% | 46.3 |
| Germany | 1.01% | 56.1% | 99.0 |
| Italy | 0.80% | 80.5% | 125.0 |
| Japan | 0.26% | 37.8% | 384.6 |
Pfau, 50% stocks / 50% bills, 30 years. Highest safe rate = the highest first-year withdrawal rate that survived every start year. Starts where 4% failed = share of start years in which a 4% withdrawal ran out within 30 years. Pot = pot as a multiple of yearly spending at that rate (100 ÷ rate). Source: Pfau, “The Shocking International Experience of the 4% Rule”, Table 3.2, 10 of its 21 rows. Withdrawals adjusted for inflation, rebalanced yearly, no fees. The multiple column is 100 ÷ the rate, worked out by us.
Three things stand out:
- Nobody reached 4% with this mix. The US and Canada missed by 0.04 percentage points. Every other market missed by more.
- The worst start years cluster around the world wars. Japan’s lowest rate belongs to people who retired in 1937, Italy’s to 1940 and Germany’s to 1914, per the same table. Those years are in the record, and the record is what the rule was built from.
- The multiple moves faster than the rate. Going from 4% to 3.36% looks small. As a pot, it is 25 versus 29.8 times a year’s spending.
The pot each rate implies
The same comparison in money, for an example household spending 30,000 a year in any currency. Each pot is the spending divided by the rate:
| Rate | Working | Pot needed |
|---|---|---|
| 4% (the US rule of thumb) | 30,000 ÷ 0.04 | 750,000 |
| 3.58% (Pfau, world portfolio) | 30,000 ÷ 0.0358 | 837,989 |
| 3.36% (Pfau, United Kingdom) | 30,000 ÷ 0.0336 | 892,857 |
| 3.1% (Monevator, UK best case, 30 years) | 30,000 ÷ 0.031 | 967,742 |
| 2.31% (Anarkulova et al., 5% chance of ruin) | 30,000 ÷ 0.0231 | 1,298,701 |
The studies differ in method, data and horizon, so these rows are not a ranking. They show how much the size of the target depends on which history you lean on.
Two later analyses
The UK back to 1870. Monevator’s 2025 series uses UK data back to 1870. It lets the best stock-and-bond mix be chosen in hindsight (“perfect foresight”, in its own words), and it ignores costs and taxes. Even so, the best 30-year rate is 3.1%, with 100% equities. For longer retirements, it gives 2.7% for 40 years and 2.4% for 50 (Monevator).
38 developed countries. Aizhan Anarkulova, Scott Cederburg, Michael O’Doherty and Richard Sias built a new dataset of asset-class returns for 38 developed countries. In their words, “a 65-year-old couple willing to bear a 5 percent chance of financial ruin can withdraw just 2.31 percent per year” (Journal of Pension Economics and Finance, 2025, 24(3), 464–500). Their data, method and measure of failure differ from Bengen’s and Pfau’s, so the 2.31% is not directly comparable with the table above.
The strongest objections, and what survives them
1. “Those countries had wars and hyperinflation. That isn’t relevant now.” The rule was built from history, and history includes those years. Pfau’s point cuts the other way: “As those countries never made the dataset, even the results I describe here include survivorship bias.” Markets that disappeared altogether aren’t in any of these tables. Conceded: the past is not a forecast, in either direction.
2. “With a better allocation, more countries pass 4%.” True. When Pfau picked the best allocation in hindsight, the highest safe rate was above 4% in Canada (4.4%), New Zealand, Sweden, Denmark and the world portfolio (4.1% each), and 4.0% in the US. His 2010 abstract sums up that kind of setup: even with “overly optimistic assumptions”, 4% would have been safe in only 4 of the 17 countries. Hindsight is not available on the day someone retires.
3. “Nobody holds only their home market.” Partly answered by Pfau’s own rows. A GDP-weighted world portfolio reached 3.58%, and the world excluding the US 3.21%. Spreading across countries got closer to 4% than most single countries did, but not to 4%.
4. “Real retirees cut spending in bad years.” These tests don’t model that. Bengen, Pfau and Monevator all assume the withdrawal rises with inflation no matter what markets do. A flexible spender is outside their results. This article doesn’t compute that case.
5. “Early retirees need more than 30 years.” That makes it harder, not easier. Monevator’s UK best case falls from 3.1% at 30 years to 2.4% at 50.
What the studies leave out: tax and fees
Every rate above is before tax and, except where noted, before costs.
- Tax. If a share of each withdrawal goes to tax, less of it can be spent. With an example effective rate of 20% on the whole withdrawal, a 4% withdrawal pays for 4% × (1 − 0.20) = 3.2% of the pot in spending. To spend 4% of the pot, the withdrawal would have to be 4% ÷ 0.8 = 5%. The real rate depends on the country, the account and how much of the withdrawal is gain, so this blog gives none.
- Fees. Monevator’s rule of thumb is to cut the withdrawal rate by half the yearly cost. With an example cost of 0.5% a year, the UK’s 3.1% becomes about 3.1 − 0.25 = 2.85%.
What this article does not do
It doesn’t say which rate anyone should use, and it recommends no allocation, fund or account. The allocations above are the ones the studies tested. Past returns don’t promise future ones. What the figures on these pages are, and what they leave out, is on the limitations page.
How we worked out the numbers on this page
- From the sources: every rate and failure share is copied from the linked study, with its assumptions: Bengen (US, 1926 onward, 50/50 stocks and intermediate Treasuries), Pfau (19 countries, 1900–2010, 50/50 stocks and bills, 30 years, no fees), Monevator (UK from 1870, best allocation in hindsight, no costs or taxes), Anarkulova et al. (38 countries, 5% chance of ruin, 65-year-old couple).
- Our arithmetic: the multiple (100 ÷ rate), the pots (30,000 ÷ rate, rounded to the nearest unit), the US real return (1.097 ÷ 1.029 − 1 ≈ 6.6%), the tax example (4% × 0.8 and 4% ÷ 0.8) and the fee example (3.1 − 0.5 ÷ 2).
- Example inputs: 30,000 of spending, a 20% tax rate and a 0.5% cost are round illustrations, not anyone’s numbers.
This article is part of FIRE outside the US. The other side of the equation, how long it takes to build the pot, is in Savings rate and years to financial independence.