Multi-Asset Futures Carry

Thesis (edge)

A futures price is not a forecast of where the market will be. It is today's price adjusted for the cost of getting there: financing, storage, dividends, interest rate differences. That adjustment shows up as the slope of the curve, and the slope tells you what happens to your position if the spot price simply stays where it is.

If the nearest contract is more expensive than the one behind it, a long position drifts upward toward the cheaper contract as time passes. That is positive carry. If the nearest contract is cheaper, a long position bleeds. That is negative carry.

The strategy is to systematically own the positive-carry markets and short the negative-carry ones, across every asset class you can trade. The same idea powers the FX carry trade, the commodity backwardation trade and bond curve roll-down, but running them together gives a much better diversified book than any one of them alone.

Why does it pay? Partly because carry compensates you for taking a risk somebody else wants to shed. A commodity in backwardation is one where consumers are anxious about supply and will pay up for immediate delivery. A high-yielding currency is often one with higher inflation or political risk. You are being paid, and the payment is real, but so is the risk.

Where it works (regimes)

Carry works in calm, risk-on markets. It grinds out steady gains for long stretches and looks like a wonderfully smooth strategy, right up until it is not.

It fails in risk-off shocks, and it fails everywhere at once. The high-yield currencies collapse, the backwardated commodities crash, the credit-sensitive markets gap. This is the defining feature of carry: many small gains and occasional large losses. Any backtest that does not include 2008 and March 2020 is not telling you what you need to know.

The cross-asset version is genuinely better diversified than single-asset carry, because bond curve carry and commodity carry are not driven by the same thing as FX carry. But under real stress, the diversification thins out just when you need it.

Signals

• Carry is the annualized slope between the two nearest liquid contracts, adjusted for the time gap between them.
• The economic meaning differs by asset class. In bonds it is roll-down along an upward-sloping yield curve. In equity index futures it reflects dividends minus financing cost. In commodities it reflects storage costs and how badly people want the physical product now. In currencies it is essentially the interest rate difference.
• Normalize by volatility. A market with 5 percent carry and 60 percent volatility is not a better bet than one with 2 percent carry and 8 percent volatility.
• Rank and take a long-short position. Some implementations also keep a directional element by comparing each market's carry to its own history, which is time-series carry rather than cross-sectional.

Portfolio construction

Balance risk between the long and short sides and between sectors. Left alone, a carry ranking will happily put your whole book into commodities in one year and into currencies the next, because that is where the extreme values were.

Keep the sector budgets fixed. Rank within energy, within metals, within bonds, within currencies, and allocate a set risk share to each sector. This costs a little return in backtest and protects you from the concentration that turns a bad quarter into a fatal one.

Risk model

The distribution is skewed. Model the tail explicitly rather than relying on volatility, which will look reassuringly low right before the unwind. Expected shortfall is a better guide than standard deviation here.

The most useful control is a trend filter. If the carry signal says be long a currency and the price has been falling for six months, that is the classic value trap: you are collecting interest while the currency devalues underneath you. Overriding or reducing carry positions that fight a strong trend removes some of the worst episodes. It costs you some carry in the good years, and it is worth it.

Costs & implementation

Turnover is low, since carry signals change slowly. That is a genuine advantage over trend strategies.

The real implementation cost is in the data. Getting clean curve data across dozens of futures markets, handling contract months that trade thinly, and dealing with markets where the next contract is not liquid, is more work than the signal itself. Many carry backtests are wrong because they used a curve point that nobody could actually trade.

Capacity is decent in the large markets and poor in the small ones. Since the carry ranking often puts extreme values in the thinner markets, check whether your top-ranked longs are the ones you can least afford to trade in size.

Failure modes

• Correlated unwind, where every carry position loses simultaneously.
• Value traps: high carry that exists precisely because the market is about to fall.
• Curve data that is stale or untradeable, especially in deferred contracts.
• Sector concentration that hides in a naive cross-sectional ranking.
• Being lulled by a smooth equity curve into using too much leverage.

Our Notes & Suggestions

Run carry and trend together rather than separately. They are close to the two most reliable systematic premia in futures, and their bad periods do not line up. Trend usually makes money in the exact shocks that hurt carry, because a shock is a large sustained move.

Size the book by the tail, not by the volatility. If you set leverage so that normal-day volatility feels comfortable, you have almost certainly built a portfolio that will lose far more than you expect on the bad day.

Be sceptical of any carry backtest with a Sharpe ratio above 1.5. It usually means the sample missed a crisis, or the deferred contract prices were not real.