The risks have become harder to separate
Cheap money and low inflation helped investors through much of the 2010s. When growth weakened, central banks could often cut rates, supporting both bond prices and equities. The 2022 inflation surge showed how that response changes when prices are rising too fast.
Conflict adds to the problem. There were 65 state-based armed conflicts in 2025, up from 31 in 2010. 1 An energy supply disruption can lift costs while reducing the amount households have left to spend. Companies face weaker demand, and central banks may still need to keep rates high.
Recent market episodes show how quickly losses can spread:
- In January 2025, news about a low-cost Chinese AI model helped send Nvidia down 17% in one day. 2
- During the February-April 2025 tariff sell-off, the S&P 500 fell 18.9% from its closing peak. The euro-listed world fund used in our comparison fell 21.6%, with a different trough date and exposure to currency moves.
- From 27 February to 30 March 2026, our US equity proxy fell 7.7%, the 7-10 year Treasury proxy 2.5%, and gold futures 13.8%. Broad commodities rose 16.6% over that window.
The historical appendix identifies each series, date window and currency. These episodes reveal different exposures that a single crash scenario can miss. They give no reliable schedule for the next one.
What happened to the stock-bond cushion?
Government bonds tend to help when a recession reduces demand and inflation. Investors seek reliable payments, central banks can cut rates, and existing bonds become more valuable. In 2008, US 7-10 year Treasuries gained about 18% while the S&P 500 lost 37%.
Inflation changes that relationship. If energy prices or tariffs push costs higher, interest rates may rise while company profits weaken. In 2022, the S&P 500 lost about 18%, 7-10 year Treasuries 15%, and long Treasuries 31%.
Our calculation of rolling three-year stock-bond correlations averaged -0.37 in 2011-2021 and +0.43 from 2022 through September 2026. The negative figure means they tended to offset one another; the positive figure means they more often moved in the same direction. It uses monthly returns from US equity and Treasury fund proxies, in dollars.
Higher bond yields now provide more income, changing the return comparison with other assets. Their ability to cushion an equity sell-off still depends on the cause of that sell-off. Holding fewer equities mechanically reduces participation in their losses; any additional bond gain depends on yields. Neither creates a contractual equity-crash payout, and the whole allocation must still justify its effect on compounding.
How do growth and inflation affect the comparison?
Two questions organise the analysis: is growth stronger or weaker than expected, and is inflation higher or lower than expected? The combination affects company earnings, interest rates and the purchasing power of fixed payments.
Copio framework. Directions are relative to expectations; valuations, real yields and the cause of the shock affect returns. How to read the matrix.
In a demand collapse, falling inflation gives central banks room to cut rates, helping high-quality bonds. In a supply squeeze, scarce energy or materials can rise in price while shares and nominal bonds both fall. Strong growth with easing inflation supports earnings without the same pressure on borrowing costs. Strong growth with rising inflation can support cyclical companies and commodities, while creating more risk for long fixed-rate bonds.
Gold needs a separate judgment about real interest rates, currencies and confidence in money. An inflation-linked bond can outperform a nominal bond and still lose value if real yields rise. Crash options respond to the size and speed of a market fall, whichever quadrant caused it.
The matrix groups exposures by the economic conditions that affect them. It is too coarse to serve as a rule for trading the whole portfolio whenever an economic release changes.
Government bonds
High-quality government bonds are useful when falling growth brings lower interest rates. Longer-duration bonds respond more strongly, in both directions. They can produce large gains in a recession and large losses when inflation drives yields higher.
Duration therefore changes both the possible recession gain and the exposure to rising yields. A withdrawal before maturity can turn a temporary bond-price decline into a realised loss.
Gold
Gold adds an asset whose value does not depend on a company's profit or a government's promise to repay. It can benefit when real interest rates fall, the dollar weakens or confidence in financial institutions deteriorates.
It pays no income and can fall sharply. Gold futures fell almost 30% between March and November 2008 before finishing the year higher, and about 12% in ten days during March 2020. They also fell 13.8% in the first month of the 2026 Gulf war.
Investors sometimes sell gold to raise cash during a panic. Rising real yields also increase the appeal of interest-paying assets alongside it. Our analysis treats gold as a conditional diversifier and includes periods when it falls with equities.
Currency changes the experience. In calendar 2022, gold futures were roughly flat in dollars, at -0.1%, but gained about 6.1% in euros because the euro fell 5.9% against the dollar, in our calculation. The historical dollar result and the return on a French investor's holding can differ substantially.
Commodities
Commodities can help when the shock comes from scarce energy or materials. Broad commodity exposure gained 19.3% in 2022 and 16.6% during the March 2026 window in our analysis. It lost 12.2% in the April 2025 growth scare.
That pattern gives commodities a role against supply-driven inflation, with a clear weakness when demand collapses. Fund structure, futures costs and the weight in energy affect how much of that behaviour an investor receives.
Trend-following
Trend-following strategies adjust positions as markets rise or fall across equities, bonds, currencies and commodities. They can take short positions in falling markets, allowing them to profit during a sustained decline.
The SG Trend Index gained 27.35% in 2022. 3 Falling bonds and rising commodity prices gave these strategies persistent trends to follow. The weakness is a sudden reversal: the strategy may still hold yesterday's positions when the market changes direction. The retail replicator in our comparison lost 7.2% in the April 2025 crash and 3.6% in the March 2026 shock.
A fund's live record, fees and trading approach can differ from those of a broad strategy index. An index result therefore does not establish the protection provided by a particular product.
Convex crash protection
A put gives its holder the right to sell at an agreed price before or at expiry, depending on the contract. Its payoff can rise sharply as the underlying market falls through the strike. A small premium can therefore buy a much larger payment during a covered crash, while the buyer's loss is limited to the premium.
Copio's approach evaluates the protection and the growth portfolio together. The comparison is between a growth allocation with a cost-controlled hedge and allocations with permanently lower equity exposure. The protection earns its place if the combined portfolio compounds better after premiums, trading costs and taxes, within the investor's capacity for loss.
The rebalancing rule is central. When a crash makes the protection valuable, some of that value can fund purchases of productive assets at lower prices. Without a plan to realise and reinvest the payoff, a hedge can gain during a crash and give much of it back during the recovery.
Premium budget, covered exposure, expiry, renewal and gain-realisation rules define the strategy being analysed. The percentage spent on premiums each year is different from the percentage of the portfolio held in a protection fund. Confusing them misstates the amount of capital committed and the recurring cost.
Protection can expire before a fall, miss the assets that suffer, or cost too much through a long quiet period. Ordinary equity losses can occur without a large hedge payout. A protection model that omits those losses cannot establish the sustainability of the allocation.
Which risks do the different responses address?
| Risk to the portfolio | Potential response | Main limitation |
|---|
| Recession and falling demand | High-quality government bonds | Higher inflation can push yields up |
| Energy or materials shortage | Commodities | Losses when demand falls |
| Loss of confidence in money or institutions | Gold | No income; vulnerable to rising real yields and forced selling |
| Sustained bear market | Trend-following | Sudden reversals and trading costs |
| Severe equity crash | Cost-controlled convex protection | Covered exposure, premiums, expiry and realising the payoff |
| A bill arriving during a sell-off | Cash reserved for that bill | Inflation and the return forgone |
The table compares mechanisms rather than specifying holdings or weights. An assessment of their geometric growth includes the return forgone through lower equity exposure as well as the premiums paid for explicit protection.
Does a smoother allocation establish better geometric growth?
Copio's long-term assumptions let us compare the return and volatility of several mixes in euros. The estimates use maintained portfolio weights over ten to fifteen years, before product costs and personal taxes. These examples contain no options or convex protection sleeve. Lower estimated volatility alone does not establish maximum geometric growth or cost-effective downside protection. They show the diversification within the allocation; they cannot establish the result of the protection strategy described above.
| Portfolio | Estimated annual compound return | Estimated annual volatility |
|---|
| 100% US large caps | 7.0% | 17.0% |
| World equities | 7.0% | 15.5% |
| 60% world equities, 40% euro aggregate bonds | 6.2% | 10.0% |
| 50% world equities, 40% euro aggregate bonds, 10% gold | 6.1% | 8.6% |
| 60% regional equities, 30% bonds, 10% gold | 7.0% | 9.5% |
The last example holds 30% US large caps, 15% euro area large caps, 15% Japan, 30% US aggregate bonds hedged to euros and 10% gold. It has the same rounded compound-return estimate as US equities alone, with about 44% lower estimated volatility. The selected regional weights, bond return and gold correlation all contribute to that result.
The simpler gold comparison is useful too. Moving ten percentage points from world equities to gold reduces modelled volatility from 10.0% to 8.6%, while the annual return estimate falls from 6.2% to 6.1% after rounding.
These results depend on average relationships holding well enough over time. When we move all cross-asset correlations halfway towards 1, the last example's volatility rises from 9.5% to 11.4%. The model does not establish a maximum loss or a probability of surviving a crash without losses. The calculation appendix gives the assumptions and sensitivity tests.
How do withdrawals and rebalancing affect compounding?
A portfolio that falls 50% needs to gain 100% to recover. Large losses interrupt compounding, and assets sold to meet a bill no longer participate in a recovery. Cash can fund withdrawals without selling shares, while carrying its own inflation and opportunity costs.
Re-entry is difficult after a panic sale. In our 2006-2025 sample, eight of the S&P 500's ten best days came within fifteen trading days of one of its twenty worst days. On 9 April 2025, it rose 9.5% in a single session. A recovery can begin while the news still looks bad, leaving an investor who sold with a difficult decision about when to buy back.
Portfolio drift creates another problem. A US 60/40 allocation left unrebalanced since 2002 would have grown to roughly 90% equities in our calculation. The owner would enter the next downturn with much more equity risk than originally chosen.
Rebalancing assumptions and withdrawal dates therefore affect the comparison. A calculation with fixed weights and no withdrawals describes a different return path from a household that sells assets during a decline.
What would establish cost-effective protection?
A smaller drawdown is one observation; copio's standard is the effect on expected geometric growth after costs. A complete comparison depends on several open questions:
- Which losses does the hedge cover, and which remain?
- How much do premiums, trading, financing and taxes reduce the result?
- What happens during long quiet periods and slow declines?
- How does the assumed reinvestment of gains affect compounding?
- How sensitive is the comparison to withdrawals and different market paths?
The evidence here supports analysing these mechanisms together. It does not establish optimal weights, a household's suitability for options or a proven protected-return estimate. Our conservative asset allocation comparison applies the same distinctions across ten investor case studies.
Information only; no personal investment advice. Long-term estimates and illustrative allocations can differ substantially from realised results. Past performance does not guarantee future returns.
