From January 2006 to April 2026, in euros, five of the nine asset classes we tested had monthly correlations with world equities below 0.2: trend following (-0.14), euro money market (-0.12), gold (-0.09), EUR-hedged global government bonds (0.01) and euro government bonds (0.18). A low correlation is not the same as protection.
Only gold and trend following rose in all three equity falls of 2008, 2020 and 2022; euro government bonds lost 12.3% in the 2022 fall, and money market funds held their value without earning much. This study measures each class against MSCI World in euros, in all months, in the months equities fell, and in the three stress periods that tested diversification most.
Key takeaways
- Average correlation can hide a failed hedge. Euro government bonds had a 0.18 correlation with world equities over twenty years, but 0.76 in 2022-2023 and losses alongside equities in the 2022 fall.
- Test growth and losses together. The 80/20 gold mix grew 9.6% a year with a 35.5% largest fall; the trend mix grew 8.6% with a 36.5% fall, versus equities' 8.7% and 49.0%, before most product costs.
- A research result is not a product return. The trend series is a factor before fees, and the diversification score is a statistical proxy; inspect an actual fund's rules, costs and exposure before treating it as the same investment.
Over 20 years of monthly returns in euros, trend following, euro money market, gold and high-quality government bonds had correlations with world equities close to zero, while listed real estate and emerging-market equities moved with them. A correlation of 1 means two assets move in lockstep, 0 means no linear link, and a negative value means they tend to move in opposite directions.
Correlation with world equities, all months and falling months: trend following -0.14 and -0.21, money market -0.12 and -0.18, gold -0.09 and -0.02, global government bonds 0.01 and -0.07, euro government bonds 0.18 and 0.06, listed real estate 0.81 and 0.73.
Source: Copio calculation on MSCI World NR in EUR, AQR's TSMOM factors and the index series listed below; results by asset class (CSV).
The groups are clear:
- Moved with equities: listed real estate (0.81) and emerging-market equities (0.72). Owning them spreads company and country risk; it does not add a separate return source.
- Partly linked: inflation-linked bonds (0.40) and commodities (0.35).
- Close to zero or negative: euro government bonds, global government bonds hedged into euros, gold, euro money market and trend following.
Our falling-together study shows that most of these values hold up when only the falling months are counted. That is a 20-year average, though, and the average hides when a correlation changed.
Does a low correlation protect you in a crash?
Not reliably: over 20 years government bonds had a low correlation with world equities, yet in 2022 they fell with them. The correlation of each asset with world equities changed from one stress period to the next.
Correlation in 2008-2009, 2020-2021 and 2022-2023: euro government bonds -0.19, 0.15 and 0.76; global government bonds -0.37, -0.11 and 0.68; gold -0.29, -0.14 and -0.20; trend following -0.54, -0.38 and -0.45.
Source: Copio calculation; results by asset class (CSV), monthly returns (CSV).
What each class returned while world equities fell shows the same break:
| Return in EUR, from peak to trough of world equities | 2008 (May 2007 to Feb 2009) | 2020 (Jan to Mar 2020) | 2022 (Dec 2021 to Jun 2022) |
|---|
| World equities (MSCI World NR) | -49.0% | -19.8% | -13.5% |
| Gold | +53.7% | +2.5% | +8.9% |
| Trend following (AQR TSMOM factor) | +42.8% | +10.4% | +26.7% |
| Global government bonds (EUR-hedged) | +13.4% | +1.4% | -10.1% |
| Euro government bonds | +12.0% | -2.1% | -12.3% |
| Euro money market | +6.8% | -0.1% | -0.2% |
| Euro inflation-linked bonds | +5.8% | -7.2% | -6.1% |
| Commodities | -21.7% | -13.3% | +28.2% |
| Emerging-market equities | -46.3% | -20.1% | -10.5% |
| Listed real estate (MSCI World IMI Core RE) | -66.3% | -27.2% | -14.1% |
Source: Copio calculation; results by asset class (CSV).
Three readings follow from the table:
- Government bonds protected in a growth scare, not in an inflation shock. They gained 12% to 13% in 2008 and lost 10% to 12% in 2022, when rising rates hit both bonds and equities. Our 2008 and 2022 bond study explains the mechanism.
- Commodities did the reverse. They lost 21.7% in 2008 and gained 28.2% in 2022: a 0.35 average correlation hides which kind of crash they help in.
- Listed real estate fell harder than world equities in all three falls, including 66.3% in 2008.
The same pattern appears month by month: in the 20 months world equities fell from January 2022 to April 2026, euro government bonds had a correlation of 0.37 with them, against -0.01 in the 67 falling months before.
Which assets kept earning when equities fell?
Gold and trend following were the only classes with a clearly positive average return in the 87 months world equities fell.
| In the 87 months world equities fell | Months it rose | Average monthly return | 2006-2021 average | 2022-2026 average |
|---|
| Gold | 52 | +1.61% | +1.83% | +0.86% |
| Trend following (AQR TSMOM factor) | 47 | +0.91% | +0.66% | +1.76% |
| Euro money market | 50 | +0.09% | +0.08% | +0.15% |
| Global government bonds (EUR-hedged) | 48 | +0.09% | +0.52% | -1.34% |
| Euro government bonds | 47 | -0.08% | +0.40% | -1.71% |
| Commodities | 39 | -0.65% | -1.09% | +0.82% |
| Emerging-market equities | 26 | -2.67% | -2.74% | -2.45% |
| Listed real estate (MSCI World IMI Core RE) | 22 | -3.43% | -3.27% | -3.96% |
Source: Copio calculation; monthly returns (CSV).
Money market funds earned 0.09% in an average falling month: their small gains offset little equity loss, while their low volatility preserved liquidity. Government bonds earned in falling months before 2022 and lost after it. Gold and trend following earned in both periods, and they did not always earn together: both fell in 21 of the 87 months.
Different return patterns can reduce portfolio risk, but they need not raise realised compounded growth over every period. The result depends on expected returns, volatility, covariance, weights and rebalancing; growth and drawdowns do not scale at a fixed common rate.
Over 2006-2026, the gold mix raised annual growth from 8.7% to 9.6%, while the trend mix produced a similar but slightly lower 8.6% before implementation costs. Both cut the largest fall by about 13 percentage points. Money market and government-bond mixes cut it by about 9 points, with annual growth 1.2 to 1.4 points lower than equities alone.
80% world equities with 20% in gold: 9.6% a year and a largest fall of 35.5%; with trend following 8.6% and 36.5%; with money market 7.3% and 40.5%; world equities alone 8.7% and 49.0%.
Source: Copio calculation, hypothetical mixes of index series rebalanced monthly; results by asset class (CSV).
The squared diversification ratio from our ETF diversification study describes the historical reduction in volatility from combining holdings. It divides their weighted individual volatilities by portfolio volatility and squares the result. A value of 1 is the single-asset reference; 1.63 is a score under this formula, not an observed count of independent economic return sources.
| 80% world equities plus 20% of | Diversification ratio squared | Annual growth | Largest fall |
|---|
| Nothing (world equities only) | 1.00 | 8.7% | -49.0% |
| Gold | 1.63 | 9.6% | -35.5% |
| Trend following (AQR TSMOM factor) | 1.57 | 8.6% | -36.5% |
| Commodities | 1.26 | 7.9% | -44.0% |
| Euro government bonds | 1.15 | 7.5% | -39.9% |
| Global government bonds (EUR-hedged) | 1.14 | 7.4% | -39.7% |
| Euro money market | 1.02 | 7.3% | -40.5% |
Source: Copio calculation; results by asset class (CSV).
Money market has the second-lowest correlation here but a ratio near 1: its 0.5% annual volatility is small against equities' 13.3%. Most of that mix's risk reduction comes from holding less equity; the ratio does not measure the value of cash for spending or rebalancing. Gold and trend have similar volatility to equities (16.0% and 13.2%) and low measured correlation with them, producing higher scores. Low correlation in this sample does not establish statistical independence or a reliable crash payoff.
Two limits apply. Gold's 11.4% a year over this window is high by its own history: it took 24 years to regain its 1983 peak in euros. The trend figures come from a research factor before fees; AQR researchers' own comparison with put options uses simulated trend returns as well.
Why did trend following and gold behave differently from bonds?
They respond to different drivers. High-quality government bonds gain when growth fears push interest rates down, as in 2008, and lose when inflation pushes rates up, as in 2022.
AQR describes its factor as "a 12-month time series momentum strategy with a 1-month holding period" across equity indices, bonds, currencies and commodities: it holds each market in the direction of its past 12-month move, so a fall that lasts months becomes a short position, whether the falling market is equities or bonds. Gold has no cash flows tied to company earnings or interest payments, and in this sample rose in 52 of the 87 months equities fell.
Neither is a promise. A trend strategy follows moves that have already happened and can lose when markets reverse sharply; gold and trend both fell in 21 of the 87 falling months. Our Sharpe ratio guide explains why combining uncorrelated returns improves a risk ratio without, by itself, establishing higher compounded growth.
What does copio take from the comparison?
Low average correlation is a starting point, not a protection promise. Euro government bonds combined a low twenty-year correlation with losses alongside equities in 2022. Gold and the trend factor improved the tested growth/drawdown balance, but their historical windows and product costs matter. Cash supplied liquidity and reduced exposure to shares, a useful role that this volatility-based score does not fully measure.
Three checks for your portfolio
- Compare the same holding in different regimes. Use returns in your spending currency over matched dates for 2008, 2020 and 2022, alongside the full-period and falling-month correlations. Record both the sign and size of the gain or loss; a small positive cash return and a large hedge payoff solve different problems. Our monthly CSV supplies the study inputs.
- Test the proposed mix, not just the extra asset. Compare annual growth, worst drawdown and the cash available for spending at the weight you are considering. Use the same dates and rebalancing rule, and repeat the comparison over shorter subperiods. Treat the squared diversification ratio as a historical risk statistic, not a count of independent economic bets or a forecast.
- Check what you can buy and what it costs. Read the product's underlying exposure, currency hedge, leverage, dealing terms and charges. Use a fund's return after costs when available; a gross trend factor cannot establish what a retail fund would have delivered. Write which risk the holding is meant to address and review whether it still performs that role when correlations change.
The portfolio checker can help inspect your fund mix; the crash-havens study tests the cost of protection across longer windows.
FAQ
Which assets are uncorrelated with stocks?
In euros, from January 2006 to April 2026, the monthly correlation with world equities (MSCI World) was -0.14 for a trend-following research factor, -0.12 for euro money market, -0.09 for gold, 0.01 for EUR-hedged global government bonds and 0.18 for euro government bonds (Copio calculation). Emerging-market equities (0.72) and listed real estate (0.81) moved largely with world equities.
Do uncorrelated assets protect a portfolio in a crash?
Not reliably. Euro government bonds had a correlation of 0.18 with world equities over 20 years, but lost 12.3% while world equities fell 13.5% from December 2021 to June 2022, and their correlation over 2022-2023 was 0.76. Of nine asset classes tested, only gold and trend following rose in all three falls of 2008, 2020 and 2022.
Is gold uncorrelated with stocks?
Over January 2006 to April 2026 in euros, gold's monthly correlation with world equities was -0.09, and it stayed negative in the 2008-2009, 2020-2021 and 2022-2023 stress periods (-0.29, -0.14 and -0.20). It rose in 52 of the 87 months world equities fell and gained 53.7% while they fell 49.0% from May 2007 to February 2009. Its 11.4% annual growth over this window is not a forecast.
Is cash a diversifier?
Cash can reduce equity exposure and provide liquidity for spending or rebalancing. In our euro sample, an 80/20 equity/money-market mix lowered annual growth from 8.7% to 7.3% and the largest fall from 49.0% to 40.5%. Its squared diversification ratio was 1.02, a volatility-based score that does not measure all the uses of a cash reserve.
Does trend following diversify an equity portfolio?
In AQR's time-series momentum research factor, converted to a euro investor's return, the correlation with world equities was -0.14 from 2006 to April 2026 and -0.21 in the months equities fell, and the factor rose in the 2008, 2020 and 2022 falls. It is a research factor before fees, not a fund; funds' results after fees and trading costs differ.
