Country Comparison
Why does a US-only simulation support a higher withdrawal rate than the 16-country pool? This table is the raw material behind that gap: real returns, inflation and worst-case windows for every country in the dataset, 1900–2025. None of it depends on your plan — these are properties of each country's own market history.
| 🇦🇺 Australia | 4.43%4.36%–4.50% | 6.45% | 1.42% | 3.76% | 0.83%1947 | 2.57% | 2.94×1974 |
| 🇩🇰 Denmark◆ | 4.40%4.35%–4.45% | 6.13% | 2.02% | 3.58% | 1.86%1951 | 2.41% | 2.97×1911 |
| 🇺🇸 United States | 4.31%4.29%–4.32% | 6.81% | 1.41% | 2.88% | 2.82%1955 | 3.22% | 2.20×1973 |
| 🇸🇪 Sweden | 3.93%3.91%–3.94% | 6.01% | 1.94% | 3.43% | 0.63%1903 | 2.36% | 2.89×1911 |
| 🇳🇱 Netherlands | 3.81%3.78%–3.83% | 5.21% | 1.09% | 3.00% | 0.65%1945 | 2.06% | 2.14×1940 |
| 🇬🇧 United Kingdom | 3.71%3.68%–3.72% | 4.93% | 0.90% | 3.88% | -0.15%1945 | 2.23% | 3.44×1973 |
| 🇨🇭 Switzerland◆ | 3.70%3.68%–3.74% | 4.93% | 1.72% | 2.11% | 1.13%1962 | 1.86% | 2.45×1910 |
| 🇳🇴 Norway | 3.40%3.37%–3.44% | 4.07% | 1.07% | 3.65% | -1.69%1949 | 0.52% | 3.37×1911 |
| 🇪🇸 Spain◆ | 3.01%2.97%–3.05% | 3.97% | 0.24% | 5.47% | -1.66%1954 | 0.19% | 4.66×1974 |
| 🇫🇮 Finland | 2.74%2.71%–2.77% | 5.08% | 0.29% | 6.63% | -1.41%1900 | 1.61% | 11.98×1912 |
| 🇩🇪 Germany◆ | 2.62%2.57%–2.68% | 3.65% | -17.33% | 27.35% | -7.30%1902 | -2.80% | 1.28e+12×1915 |
| 🇮🇹 Italy | 2.54%2.53%–2.55% | 1.96% | -1.30% | 7.63% | -4.73%1916 | -1.86% | 49.26×1938 |
| 🇧🇪 Belgium◆ | 2.38%2.36%–2.40% | 2.68% | 0.20% | 4.86% | -3.90%1914 | -2.02% | 6.69×1909 |
| 🇵🇹 Portugal | 2.17%2.10%–2.24% | 0.07% | 0.17% | 6.96% | -7.05%1955 | -2.27% | 23.99×1915 |
| 🇫🇷 France | 2.17%2.13%–2.20% | 0.17% | -1.57% | 6.55% | -8.54%1943 | -1.99% | 15.37×1940 |
| 🇯🇵 Japan◆ | 1.72%1.67%–1.78% | 3.10% | -1.67% | 7.01% | -10.61%1919 | -3.30% | 186×1940 |
How to read this
Sort by the worst 30-year column and by the real-equity column, and you get almost the same ordering as the withdrawal rate. That is the point: a withdrawal rate is set by the bad tail, not by the average. A country can have a respectable long-run average and still be near the bottom here if it had one 30-year stretch that destroyed capital.
Denmark and Portugal make the mechanism concrete. Both are small, open, coastal European economies — on any test of economic similarity they belong in the same bucket. But between 1973 and 1985, Danish equities returned +230% in real terms while Portuguese equities returned −95%: revolution, nationalisations, and a decade of 20%+ inflation. Notice too that Portugal had no safe asset either — real bonds returned 0.17%/yr over 125 years, because the inflation that wrecked equities wrecked bonds at the same time. That correlation is what ends a retirement plan.
This is also why 'just use the countries most similar to the US' does not work as a filter. The large developed markets that most resemble the US — Germany, France, Japan, Italy — are the four biggest capital-market catastrophes in this dataset, because resembling the US mostly means being a large country that was in the middle of the wars. Screening on the first half of the century tells you nothing about the second: the rank correlation between 1900–1949 and 1950–2025 real returns is −0.02.
Two worst-30-year columns, and why they differ
The headline column holds 60% home-country equities and 40% home-country bonds, so it is a property of that country alone. The second column uses the same 30% home / 30% global / 40% bonds reference blend as the withdrawal-rate column, so the two can be read together. Neither is the simulator's own default allocation, which is 40/40/20 — change the mix there and you will get different numbers than either column here.
The gap between the two columns is the value of global diversification, and it is largest exactly where it matters: Portugal's worst 30-year window is −7.05%/yr for a home 60/40 portfolio but −2.27%/yr once a global equity sleeve is added. (On home equities alone it is −8.62%/yr.) Most of what makes a national catastrophe catastrophic is that it is concentrated at home. Note that the second column is therefore not a property of that country in isolation — it depends on what the rest of the world was doing.
About the withdrawal-rate column
Unlike every other column, this one is not a plain statistic — it is a simulation result, and it is pinned to one fixed reference scenario so the countries stay comparable. It is not a recommendation, and it is not your plan:
$1,000,000 · 30 years · 30/30/40 home equity / global equity / home bonds · 0.05% fees · fixed real withdrawals · block bootstrap, 5,000 paths × 5 seeds · reported at 90% success
The smaller range next to each rate is the spread across the random seeds — sampling noise in the bootstrap, not a confidence interval for the country's true safe rate. Run your own numbers in the simulator; these exist to rank the countries against each other, not to size a retirement.
Data notes
All figures come from the Jordà-Schularick-Taylor Macrohistory Database over 1900–2025, in local currency and adjusted for local inflation. Years after 2020 are an unofficial extension — see the data sources page for how it is built and what it does not cover.
◆ 21 values inside this window are modelled rather than observed, and they sit in exactly the war and crisis years that drive the worst-case columns, so they are listed rather than averaged in silently. Two are declared market closures (Tokyo, 1946–47: exchange shut, assets frozen, modelled as zero nominal return). The rest are inferred from the published dataset, which does not preserve the observed/imputed distinction: a bond return of exactly zero is what the upstream pipeline writes when the underlying series is missing, and in two German years the global-equity series falls back to the domestic one because the exchange rate is unavailable.
- Belgium 1914, 1915, 1916, 1917, 1918, 1919
- Switzerland 1915
- Germany 1944, 1945, 1946, 1947, 1948
- Denmark 1915
- Spain 1937, 1938, 1939, 1940
- Japan 1946, 1947
The countries and years affected are listed above so you can judge for yourself how much weight to put on the worst-case columns for those markets. Where they matter most is Belgium, Germany and Spain — whose worst windows overlap the years being disclosed.