Key takeaways
- From January 1991 to December 2025, moving from no emerging markets to a 20% weight raised annualised volatility from 14.76% to 15.15% and deepened the worst monthly drawdown from 53.62% to 55.13%.
- Annualised return moved from 8.91% at no emerging markets to 9.00% at a 20% weight. The Sharpe ratio read 0.483 with no emerging markets, with 5% and with 10%.
- The monthly correlation between emerging and developed equities was 0.775 over those 420 months. The break-even was 0.732, which is why the sleeve added volatility rather than removing it.
- Taiwan Semiconductor Manufacturing was 15.14% of the MSCI Emerging Markets index at the end of August 2026, and the ten largest holdings were 37.53% of it.
- Emerging markets still won on return over the long run: the MSCI Emerging Markets index returned 9.16% a year net since the end of December 2000, against 7.53% for MSCI World.
The short answer: 5%, 10% and 20% moved the risk numbers by less than half a point
You want to know what a slice of emerging markets does to a portfolio you already own. Here's the measurement, over the 420 months from January 1991 to December 2025, using the Fama/French developed and emerging market research series in US dollars.
Developed-market equities on their own ran at 14.76% annualised volatility. Put 5% into emerging markets and the mix ran at 14.82%. At 10% it ran at 14.90%. At 20% it ran at 15.15%. Volatility rose at every weight tested. The worst drawdown measured on month-end values went the same way, from 53.62% to 55.13%. Annualised return went from 8.91% to 9.00%.
So on this sample, the answer is: almost nothing changed, and the little that changed went in the direction most people don't expect. That needs explaining, because it isn't what a diversifier is supposed to do.
Thirty-five years, seven weights, one table
All seven mixes are rebalanced monthly back to their target weight. Returns include dividends and are in dollars, before tax and before any dealing cost.
| Emerging weight | Return a year | Volatility | Worst drawdown | Sharpe |
|---|---|---|---|---|
| None | 8.91% | 14.76% | 53.62% | 0.483 |
| 5% | 8.94% | 14.82% | 54.00% | 0.483 |
| 10% | 8.97% | 14.90% | 54.37% | 0.483 |
| 20% | 9.00% | 15.15% | 55.13% | 0.480 |
| 30% | 9.02% | 15.50% | 55.88% | 0.474 |
| 50% | 9.00% | 16.48% | 57.40% | 0.454 |
| 100% | 8.64% | 20.18% | 61.20% | 0.389 |
The chart plots the volatility column. Three things stand out. The Sharpe ratio is flat to three decimals at zero, at 5% and at 10%, so over 35 years the first tenth of the portfolio changed the risk-adjusted return by nothing measurable. Return peaks around a 30% weight, at 9.02%, and then falls away. And the whole return column spans 0.38 percentage points between the best mix and the worst, which is inside the noise of any 35-year sample.
The drawdown column matters more than the volatility column for most people, because it's the number you have to live through. Every rebalanced mix bottomed in the same month, February 2009, from a peak in October 2007. Adding emerging markets made that hole deeper each time.
Why an emerging markets allocation can add volatility instead of removing it
The arithmetic here is not mysterious, and it's the part that gets skipped. Whether a new sleeve lowers the volatility of what you already hold depends on two things: how volatile the sleeve is relative to the core, and how closely the two move together.
The first slice of a new asset reduces portfolio volatility only when its correlation with the core sits below the ratio of the core's volatility to the sleeve's. Developed equities ran at 14.76% and emerging at 20.18%, so that break-even correlation was 0.732. The measured correlation across the 420 months was 0.775. Above the break-even, the sleeve adds volatility from the very first pound.
Run the same arithmetic forwards and the minimum-variance weight in emerging markets over this period was minus 7.94%. That is not a position a long-only portfolio can hold. That's the honest statement of the result: on the full sample there was no positive weight in emerging markets that lowered volatility, so the case for an emerging markets allocation has to rest on return, not on risk reduction.
Notice how little of that depends on emerging markets being risky. A sleeve twice as volatile as the core still cuts portfolio volatility if its correlation is below 0.5. The problem wasn't the volatility on its own, and it wasn't the correlation on its own. It was the pair.
The correlation matrix, and the decade the arithmetic turned against it
Correlation is not one number. It's a number with a date attached, and the date does most of the work here. The matrix below shows emerging markets against five developed groupings, all built from the same monthly dollar series, alongside the two volatilities and the break-even they imply.
| Period | vs Developed | vs North America | vs Europe | vs Japan | vs Asia Pacific ex Japan | EM volatility | Developed volatility | Break-even |
|---|---|---|---|---|---|---|---|---|
| 1991 to 1999 | 0.59 | 0.60 | 0.50 | 0.27 | 0.79 | 21.26% | 12.35% | 0.58 |
| 2000 to 2009 | 0.89 | 0.83 | 0.86 | 0.62 | 0.92 | 24.19% | 16.75% | 0.69 |
| 2010 to 2019 | 0.84 | 0.76 | 0.81 | 0.60 | 0.93 | 16.32% | 13.10% | 0.80 |
| 2020 to 2025 | 0.79 | 0.75 | 0.80 | 0.67 | 0.89 | 16.50% | 16.67% | 1.01 |
| 1991 to 2025 | 0.78 | 0.73 | 0.75 | 0.48 | 0.87 | 20.18% | 14.76% | 0.73 |
In the 1990s the correlation with developed markets was 0.59 against a break-even of 0.58. That's a hair on the wrong side, and it means emerging markets were close to genuinely diversifying then. By the 2000s the correlation was 0.89 and the break-even 0.69, which is not close at all. The gap between the two columns is the whole story of this article.
One row breaks the pattern. Between 2020 and 2025 emerging markets ran at 16.50% volatility while developed markets ran at 16.67%. The sleeve was, for the first time, the calmer of the two, so the break-even rose above 1 and any correlation at all lowered portfolio volatility. That's a real result, and it's the strongest thing anyone can say for the case. It also rests on a 72-month stretch.
The Japan column is worth a glance for a different reason. Japan's correlation with emerging markets over the full period was 0.48, well below emerging markets' 0.78 with developed equities as a whole. The most effective diversifier inside the developed index was a developed market. Our piece on international diversification works the same grid from the other side.
Drawdowns got deeper in three of the four largest falls
Volatility is a summary. Drawdown is what people actually experience, so here are the four worst equity episodes in the sample, with the same portfolios.
| Episode | No emerging | 10% emerging | 20% emerging |
|---|---|---|---|
| July 1997 to August 1998 | -0.57% | -7.05% | -13.18% |
| November 2007 to February 2009 | -53.62% | -54.37% | -55.13% |
| January to March 2020 | -21.77% | -22.09% | -22.42% |
| January to September 2022 | -25.46% | -25.37% | -25.29% |
Calendar 2008 alone went from a loss of 40.41% with no emerging markets to a loss of 43.23% with 20%. The 1997 to 1998 stretch is the extreme case and the one worth staring at. Developed markets came through it roughly flat, losing 0.57%, while a 20% emerging weight turned that into a 13.18% loss. The crisis was in emerging markets, which is exactly when a diversifier is meant to help and exactly when this one didn't.
2022 is the exception, and a small one. The emerging sleeve trimmed the loss by 0.17 percentage points at a 20% weight, because emerging markets fell slightly less than developed markets over that stretch.
The counter-case: the 2000s paid for the whole thing
The full-period table hides a decade in which every one of these numbers pointed the other way, and any fair reading has to put it on the page.
| Period | No emerging, return | 20% emerging, return | No emerging, volatility | 20% emerging, volatility |
|---|---|---|---|---|
| 1991 to 1999 | 14.54% | 14.28% | 12.35% | 12.85% |
| 2000 to 2009 | 1.38% | 3.25% | 16.75% | 17.84% |
| 2010 to 2019 | 9.73% | 8.79% | 13.10% | 13.34% |
| 2020 to 2025 | 12.40% | 11.50% | 16.67% | 16.08% |
From 2000 to 2009 a 20% emerging weight turned 1.38% a year into 3.25% a year. That is 1.87 percentage points a year for a decade, which dwarfs the 0.09 points the same weight added across the full 35 years. MSCI's own index tells the same story from a different construction: the MSCI Emerging Markets index returned 9.16% a year net since the end of December 2000, against 7.53% for MSCI World and 7.48% for MSCI ACWI.
The objection to the piece you're reading is therefore a serious one. A start date of 1991 buries a decade of emerging-market outperformance inside a longer sample that developed markets dominated. Anyone who bought at the end of 1999 has been rewarded for it since.
Here's what the rolling record does to that objection. Across all 301 overlapping ten-year windows in the sample, a 20% emerging weight beat a pure developed portfolio in 127 of them, or 42.2%. The median gap was minus 0.36 percentage points a year. The best window paid 2.59 points a year and the worst cost 1.43. So the sleeve won less than half the time, by amounts that were small either way, and its good decade was genuinely good. Both of those things are true.
The measured cost is also more recent than the measured benefit. The MSCI Emerging Markets index returned 9.29% a year net over the ten years to August 2026, against 13.01% for MSCI World, with a ten-year annualised standard deviation of 17.45% against 14.86%. Worse return, higher volatility, over the most recent full decade.
What the MSCI Emerging Markets index actually holds
The word "emerging" invites a picture of 24 economies. The index is narrower than that. At the end of August 2026 the MSCI Emerging Markets index held 1,178 constituents across 24 countries, covering roughly 85% of free-float market value in each. Taiwan Semiconductor Manufacturing alone was 15.14% of it. The ten largest holdings were 37.53%.
That concentration is the mechanism behind the correlation number. A sixth of the sleeve is one semiconductor company, and eight of the ten largest holdings in MSCI World sit in information technology or communication services, so the two indices are not describing separate economies. MSCI World, for comparison, held 1,280 constituents across 23 developed countries with its top ten at 26.61%, which is concentrated too, just less so.
The size gap is the other thing worth knowing. MSCI put the emerging index's float-adjusted value at 12,338,300.43 million dollars at the end of August 2026 and MSCI World's at 91,704,810.20 million. Emerging markets were therefore 11.86% of that combined large- and mid-cap universe.
Set that against output. The IMF's World Economic Outlook data put emerging market and developing economies at 60.565% of world GDP measured at purchasing power parity in 2025, up from 34.936% in 1990. The two figures don't measure the same thing, and the IMF's group is far broader than MSCI's 24 countries, so the gap between 11.86% and 60.565% is not a mispricing you can arbitrage. It's a statement that listed equity and national output are different quantities. Our piece on GDP growth and stock returns tests whether the growth has reached shareholders, and largely finds it hasn't.
What counts as emerging is also a decision, not a measurement. MSCI's classification framework, published in June 2023, rests on three criteria: economic development, size and liquidity, and market accessibility, with economic development used only to decide what qualifies as developed. Countries move between the buckets when those assessments change.
The fee is not the obstacle here
A common assumption is that the emerging sleeve is expensive enough to explain the shortfall. On the two iShares core trackers a UK investor is most likely to meet, it isn't. The iShares Core MSCI EM IMI UCITS ETF (EIMI) carried a total expense ratio of 0.18% and held 3,061 positions in its August 2026 factsheet. The iShares Core MSCI World UCITS ETF (SWDA) carried 0.20% and held 1,279. The emerging fund was the cheaper of the two.
The emerging fund's own factsheet names greater liquidity risk, restrictions on investment or transfer of assets, and failed or delayed delivery of securities among its key risks. None of that shows up in a total expense ratio. But the headline fee is not what moved these numbers, and a difference of 0.02 percentage points cannot account for a return gap of 3.72 points a year over the last decade.
What this test cannot tell you
Four limits are worth stating plainly, because each one could move the answer.
It's one sample. 420 months sounds like a lot, and it is 35 years, which is three and a half non-overlapping decades. That's why the rolling-window result matters more than the full-period one. It is also a backtest, and a backtest describes a path that happened rather than the range of paths that could have.
It's in dollars. A sterling investor's numbers differ, because the currency leg sits between the index and the outcome, and it doesn't behave the same way in both sleeves. The sterling version isn't tested here, so every level in these tables belongs to a dollar investor.
It's gross. No withholding tax, no dealing spread, no platform fee, no tracking difference. Index-style returns are the ceiling of what anyone actually received, in both sleeves.
And the boundary itself moves. Greece appears on both of Ken French's country lists, developed and emerging, which is a fair warning that the line is drawn by committee. MSCI says it considers a market for upgrade only when the change in status looks irreversible, so a country that succeeds eventually stops being part of the emerging series, which is not a neutral thing to do to a long-run average. Correlations estimated from that series inherit the problem, and they carry ordinary estimation error on top of it, as our piece on correlation instability sets out.
One thing the test is robust to is the rebalancing rule. At a 20% weight, monthly rebalancing returned 9.00% a year, annual rebalancing 9.16%, and simply letting the mix drift 8.86%. Drift ended the period at an 18.6% emerging weight, so there was no great migration to correct.
What would change the conclusion
The break-even correlation is the number to watch, and it isn't fixed. It's the ratio of developed volatility to emerging volatility, and it moved from 0.58 in the 1990s to 0.73 across the full sample to 1.01 in the 2020 to 2025 window. Once that ratio passes 1, emerging markets are the calmer asset and the sign of this whole argument flips. It has already passed 1 once, on that 72-month stretch.
The second is the correlation itself, which peaked at 0.89 in the 2000s and read 0.79 from 2020 to 2025. A sustained move back toward the 0.59 of the 1990s would restore the diversification case without any change in returns at all.
The third is composition, and it's the one that gets least attention. At 15.14% in a single semiconductor manufacturer and 37.53% in ten names, the index's behaviour is increasingly a question about a handful of supply chains rather than about 24 economies. If that concentration keeps rising, the case for treating the sleeve as a separate market gets harder to make, and "emerging markets" stops being a useful label for the risk being taken on. How much of the rest of a portfolio already sits in the same place is the subject of home bias.