Why a US-Only Portfolio Underperforms a Diversified One (A Fair, Same-Years Test)

Tested against the exact same historical years, a diversified portfolio beat an all-US one for retirement survival: 100% success versus 94.9% over 1991-2025. Here's the fair, apples-to-apples comparison, and the honest caveats behind it.

Tested against the exact same historical years, a diversified portfolio beat an all-US one for retirement survival. Over 1991-2025, a 100% US-equity portfolio survived a 30-year retirement in 94.9% of resampled paths. A mix of 50% US stocks, 20% international developed markets, and 30% bonds survived in 100% of them. Same years, same withdrawals, same starting balance: the diversified portfolio simply retired more of the paths safely.

That is the whole finding, and it is a narrow one on purpose. This article is careful about what it does and does not claim, because the honest comparison is more interesting than the loud one.

Why the comparison has to be same-years

You cannot fairly say one portfolio beats another unless you test them over the same stretch of history. A portfolio that only ever saw the calm markets of one decade will always look safer than one that lived through a crash, no matter what it holds. The comparison only means something when both sides face the identical sequence of good and bad years.

That constraint is what makes diversification hard to test far back. The international developed-markets index data we bundle starts in 1991. Before that, there is no international series to diversify into, so any earlier test would be US assets on both sides, which is not a diversification test at all. So the fair window is 1991 through 2025: 35 years where both a US-only portfolio and an internationally diversified one can be replayed against the exact same market history.

Within that window, here is what the engine found.

The fair test: same years, two portfolios

Both portfolios below were run through the same historical block bootstrap: 10,000 resampled paths drawn from 1991-2025 in 7-year blocks (so real runs of good and bad years stay glued together, which is the whole point of sequence-of-returns risk). Each path is a 30-year retirement starting with $1,000,000 and withdrawing $40,000 a year in real terms, a 4% withdrawal rate. “Success” means the portfolio never ran dry across the full 30 years.

Illustrative only, not advice. Both rows are the site's own historical block-bootstrap engine: 10,000 paths, 7-year blocks drawn from the 1991-2025 window, 30-year horizon, $1M starting balance, $40k real annual withdrawal (4%). The two rows share the exact same years, so the difference is the portfolio, not the era. Run your real number →

A 5-point gap between 94.9% and 100% can sound small. In retirement terms it is not: it is the difference between a plan that failed in roughly 1 out of 20 histories and one that failed in none of them over this window. The mechanism is old and boring, which is why it is trustworthy. US and international markets do not crash and recover on the same schedule, and adding bonds dampens the deepest drawdowns. When your withdrawals hit a portfolio that did not fall as far in its worst years, more paths survive the dangerous early stretch. That is all diversification is doing here.

The honest caveat: the full US history is a different test

Here is the part it would be easy to get wrong, so read it slowly. For context, US equities alone have actually held up reasonably well across their full available track record. Tested against the entire 1872-2025 span, 154 years, a 100% US-equity portfolio survived a 30-year retirement in 92.5% of paths.

But that 92.5% number is not a fair comparison to the 100% above, and you should not line them up as if it were.

Illustrative only, not advice. A separate, non-comparable run: the same engine and assumptions, but drawn from the full 154-year US-only history rather than the 1991-2025 window above. It cannot be compared like-for-like to the diversified portfolio, because international data does not exist before 1991. Shown only as honest context on how US equities have done on their own longest record.

Why keep it separate? Because we cannot test diversification before 1991, when the international data begins. The 92.5% covers a century and a half that the diversified portfolio never got to face, including deep pre-war and mid-century episodes that the 1991-2025 window simply does not contain. It is a different era, a different set of shocks, and a longer sample. Comparing it directly to the diversified number would be comparing two different questions.

What it is good for is one honest observation: even measured against its own longest and most complete track record, a US-only portfolio comes in at 92.5%, still below the 100% the diversified portfolio reached over the shorter window where we can actually test both. That does not prove diversification would have won across all 154 years, because we have no international data to check that. It just means US-only, on every measurement we can run, did not clear the bar the diversified mix did. Over the years we can test both fairly, diversification measurably helped.

What this does not say

To keep the claim clean, a few things this article is deliberately not arguing:

• It is not arguing that the long US history is misleadingly rosy or overstates safety. It does not. On its full record the US-only portfolio scored 92.5%, a hair below its 1991-2025 score, not above it. The long history is not flattering US-only results.
• It is not arguing that international always wins, or that these exact weights (50/20/30) are optimal. They are one illustrative mix over one window.
• It is not arguing that past survival rates are future guarantees. They are the historical record replayed, nothing more.

The single defensible takeaway is the narrow one at the top: over the fair, same-years window, the diversified portfolio survived more retirements than the US-only one.

How to use this

If you are building a retirement plan, the practical lesson is not “chase the highest historical number.” It is that a plan tested only against US stocks is being graded on a single market’s luck, and that spreading the same money across more than one market lowered the failure rate over every fair test we could run. Diversification did not raise the ceiling on returns here; it raised the floor on survival, which is the number that actually matters when you are the one withdrawing.

If you want to see how these survival rates are built path by path, read how the Monte Carlo simulation works. If you want the deeper US-only history of 30-year withdrawal windows on its own, that lives in the safe-withdrawal-rate history guide. And if you are still sizing the target itself, the 25x rule is the place to start. Then run your own numbers, move the sliders, and watch the floor move with them.

Frequently asked