Futures carry

Definition

Carry strategies hold futures whose curve shape pays the holder for the passage of time — long markets in backwardation (front above back: rolling long collects the spread) and short or avoiding markets in deep contango (rolling long pays it). The premise: if the spot price ends unchanged, the position earns (or avoids paying) the roll differential; curve shape is the priced compensation.

How it works / structure

• Carry measurement: front-to-next spread annualized (the roll yield), or spot-vs-futures basis; sign and size per market, refreshed each roll cycle (ms-futures-roll).
• Parameters (engine-executable): universe, carry lookback/ smoothing, cross-sectional (rank markets by carry, long top / short bottom) vs time-series (hold each market by its own carry sign), volatility scaling (risk-volatility-targeting), rebalance/roll schedule.
• Why it might pay: storage economics and hedging-pressure theories — producers paying speculators to warehouse price risk; backwardation as a risk premium (ext-commodities covers the theory citations).
• Combination evidence: carry and trend (strategy-futures-trend-following) are lowly correlated and commonly blended.

When it applies

Diversified futures universes (the evidence is cross-market; single-market carry is mostly one commodity’s storage story), FX (rate-differential carry — ext-fx), and as a filter on other futures strategies (trend entries in carry-adverse curves pay double).

Risk profile & failure modes

• Carry crashes: the strategy is structurally short stress — carry-favorable positions unwind violently when the priced calm breaks (FX carry in 2008 is the canonical episode); the return distribution is negatively skewed.
• Curve snapshots lie: carry measured today is not carry received — the curve reshapes continuously; realized roll yield can differ in sign from the entry snapshot.
• Crowding: carry is a known, capacity-limited premium; crowded carry unwinds correlate across markets.
• Single-market traps: one market’s backwardation is often a shortage story ending in a spot break — the cross-market portfolio IS the strategy.

Evidence & limits

Koijen-Moskowitz-Pedersen-Vrugt (2018) documented positive carry-strategy returns across futures, rates, FX, credit, and options universes in their samples — the broadest carry evidence — while measuring its crash-prone skew. Commodity- specific carry (backwardation premia) has an older literature with mixed per-market results. As with all published premia: period-dependent, decay-suspect, and treated by the platform as a hypothesis each replay must re-earn.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Market X, in backwardation at B% annualized today, will deliver positive roll-adjusted return over the next two roll cycles with spot ending within ±2% of today” — falsified by the decomposed replay P&L.
• “A cross-sectional carry portfolio (top vs bottom tercile of 20 markets) will finish the next 6 months positive in replay” — falsified by the replay result.

Cross-references

• The mechanism: ms-futures-roll; the same trade in spread form: strategy-futures-calendar-spread
• Common blend: strategy-futures-trend-following
• Theory and per-asset detail: ext-commodities, ext-fx
• Risk shaping: risk-volatility-targeting, risk-scenario-analysis (skew-aware sizing)