How to Test Whether International Diversification Is Still Worth It
Global equity markets have grown steadily more interconnected over the past several decades, which raises a fair and increasingly common question: does spreading a portfolio across countries still meaningfully reduce risk, or has integration quietly eroded the benefit? This piece gives a practical, repeatable test rather than relying on an outdated rule of thumb.
The core idea: a threshold test for adding an asset
Whether adding a new asset class improves a portfolio's efficient frontier, its best available combination of risk and expected return, depends on a specific relationship between the new asset's own return-to-risk ratio and its correlation with the existing portfolio. The test is: expected return of new asset / volatility of new asset > correlation × (expected return of portfolio / volatility of portfolio). If the new asset's own reward-to-risk ratio clears this correlation-adjusted threshold, adding even a small allocation to it improves the portfolio's risk-adjusted return; if it falls short, the new asset does not earn a place on the frontier no matter how attractive it looks in isolation.
The intuition behind the threshold is that a lower correlation lowers the bar an asset needs to clear. An asset with a mediocre standalone return-to-risk ratio can still improve a portfolio if its correlation with existing holdings is low enough, because the diversification benefit compensates for the weaker standalone profile. Conversely, an asset with an excellent standalone ratio but high correlation with the existing portfolio adds comparatively little, since it mostly duplicates risk the portfolio already has rather than offsetting it.
The math: applying the test to international stocks
Suppose a U.S.-only portfolio has an expected return of 8% and volatility of 15%, giving a return-to-risk ratio of 8 / 15 ≈ 0.533. International stocks are expected to return 7.5% with volatility of 18% and a correlation of 0.65 with the existing U.S. portfolio. The test threshold is correlation × portfolio ratio = 0.65 × 0.533 ≈ 0.347. International stocks' own ratio is 7.5 / 18 ≈ 0.417, which clears the 0.347 threshold, meaning adding international stocks should still improve the portfolio's risk-adjusted return, even though their standalone expected return of 7.5% is lower than the U.S. portfolio's 8%.
Push the correlation higher to see how sensitive this result is: if correlation between U.S. and international stocks were instead 0.85, closer to figures seen in some recent high-integration periods, the threshold becomes 0.85 × 0.533 ≈ 0.453, now higher than international stocks' own 0.417 ratio. At this higher correlation, the test fails, and adding international stocks under these specific return and volatility assumptions would no longer clearly improve the portfolio's risk-adjusted return, though it would not necessarily make it worse either; the calculation shows the benefit has genuinely shrunk to a point where it is no longer unambiguous, which is exactly the concern that rising global market integration raises.
A second example: when the test says no
Now test a different candidate asset to see the same framework flag a genuinely weak addition. Suppose an investor is considering a specific single-country emerging-market fund with an expected return of 9%, volatility of 28% (reflecting the concentration and political risk of a single country), and a correlation of 0.55 with the existing U.S. portfolio, using the same 8% return, 15% volatility U.S. portfolio as before, ratio 0.533.
The threshold is 0.55 × 0.533 ≈ 0.293. The candidate asset's own ratio is 9 / 28 ≈ 0.321, which does clear the 0.293 threshold, though by a much narrower margin than the diversified international example above. This narrow pass is itself informative: it suggests a small allocation could still help, but the margin is thin enough that a modest downward revision to the expected 9% return estimate, entirely plausible given how uncertain single-country return forecasts are, would flip the conclusion. Compare this to a broad, diversified emerging-market fund with a more conservative expected return of 7%, lower volatility of 20% from diversification across many countries, and a similar 0.55 correlation: its ratio is 7 / 20 = 0.35, comfortably above the 0.293 threshold with a much wider margin of safety than the concentrated single-country alternative, despite the lower headline expected return.
It is worth being explicit about what happens at the boundary case, where the candidate asset's ratio sits almost exactly at the threshold, since this is the situation investors face most often in practice rather than the clean, wide-margin passes and fails used in the examples above. When a candidate asset's ratio and the correlation-adjusted threshold are within a few hundredths of one another, the honest conclusion is that the data cannot confidently distinguish "worth adding" from "not worth adding" given normal estimation uncertainty in the underlying return, volatility, and correlation inputs, each of which typically carries a meaningful standard error of its own when estimated from historical data. In that situation, a reasonable response is not to force a binary yes-or-no decision from the arithmetic, but to add a modest, exploratory allocation, small enough that being wrong about the marginal benefit costs little, while continuing to monitor whether the input estimates converge more clearly toward a pass or a fail as more data accumulates.
What the evidence shows about the test's stability
Historical correlation studies between U.S. and international equities show the input to this test has not been stable over time: multi-decade averages have risen from figures often below 0.5 in earlier windows toward figures more commonly in the 0.7 to 0.85 range measured over recent decades, meaning a diversification case that was strong using data from thirty years ago is measurably weaker, though generally not eliminated, using more recent correlation figures, exactly the pattern the sensitivity check above illustrates numerically.
At the same time, studies decomposing correlation by market segment consistently find that correlation is not uniform across all forms of international exposure: broad emerging-market indexes, small-cap international stocks, and certain sector-specific international exposures have generally retained somewhat lower correlation with U.S. large-cap equities than developed-market large-cap international indexes have, meaning the diversification case, run through this same threshold test, tends to remain stronger for some segments of international exposure than for others even as headline "U.S. versus international" correlation has climbed overall.
Running this test on your own portfolio
Applying this framework in practice does not require precise forecasting, since the test is fairly forgiving to moderate estimation error, as the narrow-margin example above demonstrated; it requires reasonable, conservative estimates of expected return, volatility, and correlation, ideally drawn from long-run historical averages rather than recent, potentially unrepresentative short-term data, and then simple arithmetic to check whether the correlation-adjusted threshold clears. Running the test with a range of plausible inputs, rather than a single point estimate, and checking whether the conclusion holds across that range is a more robust approach than relying on one specific set of numbers that happened to produce a favorable answer.
For most investors, rerunning this test every few years, rather than once at the start of an investing career and never again, is a reasonable discipline given how much correlation figures have moved historically. It is also worth running the test separately for developed and emerging market allocations, rather than treating "international" as a single decision, since the evidence above shows their correlation profiles, and therefore their thresholds, can differ meaningfully.
One more refinement worth adding to the framework is testing not just whether an asset clears the threshold today, but how far it clears it, since the size of the margin is itself useful information for sizing the allocation, not just deciding whether to make one at all. The diversified emerging-market example above cleared its threshold by a wide margin, 0.35 against 0.293, while the single-country candidate cleared its own threshold, 0.321 against 0.293, by a much narrower one; a natural, defensible response to that difference is a meaningfully larger allocation to the diversified fund than to the concentrated single-country position, treating the margin above the threshold as a rough proxy for the confidence an investor should have in the conclusion, not just a pass or fail signal to act on identically regardless of how close the numbers actually came.
Actionable breakdown
- Estimate correlation between your holdings and the candidate asset.
- Use long-run historical averages, not a single recent window.
- Check the result across a plausible range, not one point estimate.
- Do not skip diversification just because expected returns look similar.
- A lower-return asset can still clear the correlation-adjusted threshold.
- Correlation, not raw return, is what the test is actually sensitive to.
- Run the test separately for developed, emerging, and small-cap segments.
- These segments carry different correlation profiles historically.
- Treating international as one number can hide a stronger sub-case.
- Recheck the test every few years as correlation figures evolve.
- A calculation from decades ago likely understates today's correlation.
- A narrow pass today can flip with a modest data update tomorrow.
It is also useful to note what this threshold test does not tell you: it evaluates whether adding a small, marginal allocation to a new asset improves the portfolio, holding everything else fixed, but it does not by itself tell you the optimal size of that allocation once added, which is a separate, more involved optimization problem depending on the full covariance structure across every asset in the portfolio, not just the pairwise relationship between the existing portfolio and the one candidate being tested. The threshold test is best understood as a first-pass screening tool, a quick way to rule out candidates that clearly do not belong before investing the greater effort a full portfolio optimization requires, rather than a complete substitute for that fuller analysis when precise allocation sizing genuinely matters to the outcome.
Common pitfalls
The first pitfall is using a single crisis-period correlation figure, when correlations spike, to conclude that diversification "does not work," instead of looking across the full range of historical market conditions to get a representative estimate. A second pitfall is ignoring that correlation is not a stable, permanent constant; it shifts with global capital flows, trade integration, and monetary policy regimes, so a test run a decade ago may no longer reflect current conditions.
A third pitfall is over-relying on a single point estimate for expected return, when the examples above show how a narrow pass or fail can flip with a modest, entirely plausible revision to that one input. A fourth pitfall is overcomplicating the decision process itself: a simple, broad, low-cost international allocation captures most of the available diversification benefit for most investors without needing a precise optimization exercise repeated constantly.
The bottom line
An asset with a lower expected return and higher standalone volatility can still meaningfully improve a portfolio if its correlation to your core holdings stays low enough, which is a test worth running with real numbers rather than assuming.
Related reading: international investing, asset allocation, what you are missing owning only U.S. stocks, extra risks in international funds, the diversification payoff of international stocks.