International diversification (and home bias)

Of all the protections in a withdrawal plan, spreading your stocks across the world is the cheapest. It costs nothing in expectation, because it swaps one risk for another instead of buying insurance. And it covers the gravest risk there is: a country falling behind for a generation, and that country being yours. The developed world's century settles the point (Beyond the United States: Anarkulova, Cederburg and the world sample). Retirees have almost never been ruined en masse by "the markets". They were ruined by their own market: the Japanese retiree of 1990, the German of 1914, the French retiree of 1929 or 1960, whose home portfolio lost to inflation for thirty years.

Home bias is still the most common flaw in real portfolios, in the United States as everywhere else. This article works through it in full: how big the bias is and where it comes from, then the evidence, national fates and what diversification did to withdrawal rates. Currency comes next, the one real subtlety, and the place where an American has to correct the story he usually reads. Then the modern argument about how much of a world index is American. Emerging markets, the total dose and how to put it in place close the file. The good news is that one fund does the whole job (Building it with US-listed ETFs: the retiree's shopping list).

Home bias: how big, and where it comes from§

The fact is settled. Savers in every country overweight their home market, and Americans are no exception. The classic studies (French and Poterba, and the updates since) find domestic allocations of 60 to 90% in markets worth 2 to 5% of the world. The American version looks milder only because the denominator is enormous: US stocks are around 60% of world market capitalization, and American investors hold something like 80% of their equity at home. The reflex is the same one; the arithmetic hides it better.

And the American version is the sharper bet, not the safer one. The United States is the country that won the last century, which is exactly why its record cannot be the plan (Beyond the United States: Anarkulova, Cederburg and the world sample). Holding 80 or 90% of your stocks in one market is a statement that the winner keeps winning, made by someone whose pension, house and job market ride on the same answer. The world sample warns about one thing above all others, a country falling behind, and an American plan concentrated at home is the purest form of that exposure.

The reasons behind the bias are easy to understand, and none of them survives a look. Familiarity first: you know Apple, you know Coca-Cola. But knowing the name is not knowing the risk, and the risk that matters, the fate of a national economy, is the one you already share. Currency next, "foreign money adds risk". It adds a little, and the section below explains why that is not an argument for staying home. Cost and convenience third: US funds are cheap, they are the default in every workplace menu, and a menu with no world fund is a real constraint. It is a bounded one, since a total-world fund costs a few basis points more than an S&P 500 fund and any IRA can hold one. Short history last. The backtests on the consumer sites start in 1990, or worse in 2010, a window in which the home market not only failed to kill anyone but beat everything in sight. That is the short window in all its glory (Simulator traps: independence, US bias, survivorship).

The evidence: how national fates differ§

A century of gaps. The Dimson, Marsh and Staunton Yearbook gives real stock returns from 1900 to 2023. The United States about 6.5% a year, the world about 5%, France about 3.5%. Germany and Japan run 3 to 4% with catastrophic interruptions, Italy about 2%, Austria about 1%. The averages hide what matters, which is the crossings, the years spent inside a bear market. French stocks did about −80% in real terms from 1929 to 1950, postwar inflation included. Japan spent thirty years below its real peak, 1990 to 2020. Germany was reset to zero twice. A 45-year plan crosses episodes of that kind, or it does not survive them. And no country, the United States included, holds a guarantee of being spared in this century.

What it does to withdrawal rates. Pfau (2010), then the whole broad-sample line, all find the same thing. Domestic SAFEMAX rates in most developed countries sit well below the American number. France drops to about 1.25 to 2% for a domestic 50/50, depending on the window. A world portfolio, run from any national base, climbs back toward 3 to 3.5%. So international diversification does more than smooth returns. It closes the dominant failure mode of the historical record. That is exactly what a broad-sample model replays, with coherent national paths where the simulated retiree lives one country's fate (Beyond the United States: Anarkulova, Cederburg and the world sample, Historical windows, bootstrap, parametric: the three families of models). And it is why the most serious criticism of that model is precisely that a globally diversified investor today comes out ahead of the domestic retirees it simulates. The full ranking, one bar per country, is a figure in Beyond the United States: Anarkulova, Cederburg and the world sample: on the Jorda-Schularick-Taylor panel, eight of sixteen countries carry less than 2.5%, while the equal-weighted world basket carries 3.3%. The United States on that same panel carries 3.75%, under the 4.15% the Bengen tradition prints.

The correlation objection. "Markets correlate at 0.8 now, so diversifying is pointless." Two answers. Short-run correlations really have risen, because shocks are global and a crash is global. Diversification no longer buys much protection against a bad year. But long-run divergences are untouched. US stocks returned about +200% real from 2010 to 2020, Europe +90%, emerging markets +30%, and the ranking was reversed from 2000 to 2010. What ruins a retiree is not the bad year. It is a bad decade in the market he happens to own (Sequence of returns risk: the retiree's real enemy, Market regimes (growth × inflation, sticky bears) and why they matter). International diversification protects against regimes, not against crashes, and a regime is what a withdrawal plan has to survive.

Currency: the subtlety, and why it runs backwards for an American§

Intuition says currencies add risk, so hedge them away. The truth depends on the asset class, and for a dollar investor it also runs against the version of the story written for Europeans. Worth getting straight once.

On the equity side, leave it unhedged. Three reasons, and the popular one is not among them. The added volatility is small: currency puts roughly 1 to 2 points on a sleeve already running at 15 to 16%, and correlations do part of that work. Over decades the effect shrinks further, because real exchange rates drift back toward each other instead of trending forever, so currency moves a given year's return around and adds little to a thirty-year one. Hedging, meanwhile, costs. A forward contract charges you the interest rate gap between the two currencies, which has run 1 to 2 points a year against some of them depending on the era. That gap is not a dead loss, since its counterpart is the expected decline of the high-rate currency, and what remains in practice is 0.2 to 0.3% a year of friction and fees. Paying a certain cost every year for thirty years, to remove a wiggle that averages out, is a poor trade.

The crisis cushion works the other way for you. European writers add a fourth reason to leave equities unhedged, and it is a real one on their side of the water: in a global panic the dollar, the Swiss franc and the yen rise against the euro, so a euro investor's American stocks are cushioned by the implicit dollar position that comes with them. In 2008 the MSCI World lost about 3 points less in euros than in dollars. Read that as evidence about the euro investor, not about you. You are the safe haven. When the panic comes, money runs to the dollar, foreign currencies fall against it, and your unhedged foreign stocks lose a little more in dollars than they lose at home. The safe-haven currency line in the table of defenses is one you already own, for free, by being paid in dollars (Defensive assets: the map and the roles). The mechanism is symmetric, and so is its consolation: when the dollar sags, you gain. The euro investor lost 7 points more than the dollar investor through the 2000 to 2003 bear market, which is another way of saying the American was cushioned by a falling dollar over exactly that stretch. So currency on your foreign stocks is not a hedge. It is a second bet, roughly fair and largely uncorrelated, stapled to them. Keep it because removing it costs real money every year, not because it will save you in a crash.

On the bond side, hedge it or stay in dollars. The exact opposite, and it is already settled (Bonds in retirement: types, duration, the exact job). What the bond sleeve is there for is stability. Layering 10% currency volatility on a sleeve that yields 4% destroys the thing you bought. Hence the rule worth remembering. Currency risk belongs on the equity side, where it is small and roughly fair, and never on the bond side, where it wrecks the job.

Emerging markets, and the total dose§

Emerging markets, 10 to 12% of world market capitalization, add real regime diversification: demographics, commodities, cycles of their own. They also add real friction, governance and political risk. Russia in 2022 showed it, with assets marked to zero overnight, political tail risk in the flesh (Hyperinflation and extreme scenarios). The reasonable position is their market weight. A total-world fund holds them at about 10%, which is the perfect default. Owning none of them is defensible too, and a developed-world fund stays acceptable. Overweighting them is a style bet, to be treated like a factor tilt (Factors (Fama-French, value, momentum) in the withdrawal phase): optional, written down, and sized to be held.

The practical target for a retiree's equity sleeve is simple. Roughly 100% world, developed plus or minus emerging at market weight, unhedged. Tolerate at most two deliberate and bounded departures, the ex-US valuation tilt above and a factor tilt (Factors (Fama-French, value, momentum) in the withdrawal phase). One total-world index fund does all of it in a single holding, VT being the best-known American example, and the choice among the handful of equivalents comes down to cost and account plumbing (Building it with US-listed ETFs: the retiree's shopping list). Where a workplace menu offers no world fund, the extra home weight it forces on you is the residual bias to watch: keep it within 10 to 15 points of market weight rather than double it. Past that, you are doubling down on home again.

The test in simulation. The whole argument lives inside the world sample, the broad-sample model that replays the developed century by drawing entire national fates (Beyond the United States: Anarkulova, Cederburg and the world sample, Historical windows, bootstrap, parametric: the three families of models). Compare your plan's failure probability under the central case, calibrated on your world funds, with its failure probability under that world sample. On a first reading, the gap is what international diversification spares you relative to the one-country retiree of the last century. One caveat comes right after. These long panels are built on each country's own domestic stocks and bonds. They set the bound of a national fate, and they do not react to your allocation. So the A/B has to be read on the models that do start from your holdings. Swap the world sleeve for an all-US sleeve of the same size, same withdrawal rule, same horizon, then compare the outcomes under the historical windows and under a bootstrap of your own history. As soon as the window contains a decade of divergence, 2000 to 2010 for instance, the gap jumps off the screen. Ten minutes that cure home bias for good.

The essentials§

Going further§