Skip to main content

International Diversification

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,282 words

International diversification is the practice of holding equities (and other assets) from multiple countries rather than concentrating in one's home market, on the premise that imperfectly correlated national markets lower portfolio volatility for a given level of expected return. Its core tension is timing-dependent: the diversification benefit is real and persistent on long horizons and in normal regimes, yet it tends to shrink exactly when it is most wanted — correlations across countries rise sharply during crashes — and it has been steadily eroded by decades of financial integration. The companion paradox is behavioral: despite the theoretical case, investors worldwide systematically under-hold foreign assets (the "home bias" puzzle).

How it works (mechanics)

The math is standard Markowitz portfolio theory. Portfolio variance depends not only on each market's variance but on the covariance between them. When two markets have correlation below 1.0, combining them produces a portfolio whose volatility is lower than the weighted average of the parts — the lower the correlation, the larger the "free" risk reduction. Returns are assumed roughly additive while risk is sub-additive, so the efficient frontier shifts up-and-left.

Three drivers determine how much benefit you get:

  • Cross-country correlation — the key input. Lower is better.
  • Relative volatilities — high-volatility, high-return markets (e.g. emerging markets) can still improve a portfolio if their correlation to the core holding is low enough.
  • Currency exposure — foreign returns to a domestic investor are local equity return plus the currency move. Currency adds its own (sometimes diversifying, sometimes amplifying) volatility, which is why hedged vs. unhedged is a separate decision (see sibling node).

The theoretical benchmark is the global market-cap-weighted portfolio: in a fully integrated, frictionless world, every investor should hold each market in proportion to its share of global market capitalization. By that logic a U.S. investor would hold the non-U.S. share of global equity — which varies materially with the index and date. In the MSCI ACWI (large/mid-cap free-float) the U.S. has recently run around 60–64%, leaving roughly a third (~36–40%) abroad after a long stretch of U.S. outperformance; on broader or earlier measures the non-U.S. share has been closer to half. Investors in small markets, by contrast, would hold the overwhelming majority of their equity abroad.

How it's used in practice

Most practitioners do not run the formal optimizer for country weights (estimation error makes mean-variance optimization notoriously unstable). Instead the common implementations are:

  • Global market-cap indexing — owning a total-world fund (e.g. an MSCI ACWI or FTSE Global All Cap tracker) so weights self-adjust. Vanguard and other large managers advocate this as the default neutral starting point.
  • Home-tilted allocations — many investors deliberately overweight home equities versus the cap-weight neutral, citing currency match to liabilities, tax, and lower costs. Vanguard's published framework explicitly endorses a moderate home bias rather than insisting on the pure global weight.
  • Carve-outs — a fixed strategic sleeve (e.g. "X% developed international, Y% emerging markets") rebalanced periodically.

The currency decision is handled separately: Vanguard hedges foreign bonds back to the home currency (to isolate the fixed-income characteristics) but generally leaves equity currency exposure unhedged, since currency is a smaller share of equity volatility and unhedged foreign stocks act as partial insurance against a domestic-currency decline.

Adoption, debate & evidence

The principle is near-universally taught and endorsed by index providers, large asset managers, and academia — but the magnitude of the benefit is genuinely contested.

Correlations have risen with integration. The NBER and multiple journal studies document that correlations between international equity markets, in both developed and emerging countries, increased significantly over recent decades, shrinking the historical diversification gain. The benefit has not vanished — several papers conclude economic gains remain "substantial" despite higher correlations, and emerging/high-country-risk markets still offer the largest gains because they are less integrated — but the easy diversification of the 1970s–80s is gone.

Diversification fails in the left tail. This is the most important honest caveat. Cross-country correlations are regime-dependent and spike during crises. Page & Panariello's "When Diversification Fails" (Financial Analysts Journal, 2018) finds left-tail correlations rise far above their normal-market and right-tail levels; separately, commonly cited summary figures (e.g. Two Sigma's analysis) put broad pairwise equity correlations near ~0.40 before the 2008 crisis and spiking toward ~0.70 during it, where they then plateaued for several years. By late 2008 a handful of principal components explained the great majority of variation across asset classes worldwide. So the diversification you measure on full-sample data overstates the protection you actually receive in a crash.

Home bias persists and is unexplained. Despite the theory, investors massively overweight domestic equities. The classic illustration is that an Australian investor historically held only ~17% foreign equity where naive theory implied something closer to ~98% (figures from the home-bias literature; the exact "optimal" number is model-dependent and overstated by frictionless assumptions). Home bias has been documented broadly across developed and emerging markets in the international-finance literature. Proposed explanations — barriers to foreign investment, information asymmetry, hedging of domestic non-tradable (labor) income, transaction and information costs, and behavioral familiarity — each capture part of it, but the puzzle is considered essentially unresolved. Notably, some work (e.g. Coeurdacier–Rey) argues the puzzle is "not as bad as you think" once labor-income hedging and real exchange-rate risk are modeled, which legitimately reduces the optimal foreign share below the naive benchmark.

Strengths & limitations

When it works: long holding periods, normal regimes, and especially allocations into less-integrated emerging/frontier markets, where correlation to developed cores is lowest. It also diversifies concentration risk — single-country sector and political risk — which volatility statistics understate.

When it fails: systemic crises, when correlations converge toward 1.0 and the protection evaporates precisely when needed. It is a poor tail-risk hedge; that job belongs to genuinely uncorrelated or convex assets, not foreign equities.

The #1 misuse: treating a single full-sample correlation number as if it were stable, and assuming the historical benefit will repeat. Correlations are non-stationary and rising; sizing an allocation on backward-looking averages overstates expected protection. A close second is conflating diversification with return enhancement — international holdings can underperform the home market for very long stretches (a decade-plus), and that tracking-error pain is the main reason investors abandon the strategy at the worst time.

Sources

Dispute flags: (1) The magnitude of remaining diversification benefit is contested — "declining/largely gone" (NBER, integration studies) vs. "still substantial" (cross-country local-perspective studies). (2) The "optimal" foreign share (~98% naive benchmark) is model-dependent and overstated by frictionless assumptions; Coeurdacier–Rey argue the true optimum is materially lower. Specific crisis correlation figures (~0.40 → ~0.70 in 2008) are commonly cited summary numbers, not exact universal constants.