Structural and Carry Trades

Carry, generally

Carry is the return you earn from holding an asset assuming its price does not change.

That definition is more general than it first appears, and Koijen, Moskowitz, Pedersen and Vrugt (Carry, JFE 2018) showed it unifies apparently unrelated strategies across asset classes:

In each case, carry has predictive power for returns. This is a genuinely useful unification: it means one conceptual framework covers most of what looks like a diverse set of strategies.

The mechanism, and an important distinction

Carry returns are predominantly a risk premium, not a market inefficiency. You are being paid to bear a risk that someone else wants to shed.

This distinction matters more than it might seem:

• Inefficiencies get arbitraged away as capital finds them. They decay and disappear.
• Risk premia persist indefinitely, because the risk is real and someone has to hold it. What they do instead is periodically pay out - the risk materialises, and everyone collecting the premium loses simultaneously.

So carry strategies do not decay in the way that pure anomalies do. They just occasionally hurt very badly, in a correlated way, across every asset class at once. FX carry unwinds and equity crashes tend to coincide, because both are expressions of the same risk appetite.

The traditional description is picking up nickels in front of a steamroller, and it is accurate rather than merely pejorative. The nickels are real. So is the steamroller.

The retail-accessible instance: crypto funding

Perpetual futures have no expiry. They're anchored to spot by a funding rate exchanged periodically between longs and shorts - typically every eight hours. When perps trade above spot (usually, because retail leverage is structurally long), longs pay shorts.

This creates the cash-and-carry basis trade: hold spot, short the perpetual. The position is delta-neutral, and you collect funding.

def basis_trade_economics(
    spot_price: float,
    perp_price: float,
    funding_rate_8h: float,
    notional: float,
    taker_fee: float = 0.0005,
) -> dict:
    """
    Economics of a delta-neutral long-spot / short-perp position.
    funding_rate_8h is the per-8-hour rate, e.g. 0.0001 for 0.01%.
    """
    periods_per_year = 3 * 365
    gross_annual = funding_rate_8h * periods_per_year

    basis = (perp_price - spot_price) / spot_price

    # Four fee events: open both legs, close both legs
    round_trip_cost = 4 * taker_fee
    cost_pct_of_notional = round_trip_cost

    return {
        "gross_annualised_yield": gross_annual,
        "current_basis": basis,
        "round_trip_cost_pct": cost_pct_of_notional,
        "breakeven_holding_days": (
            cost_pct_of_notional / (funding_rate_8h * 3) if funding_rate_8h > 0 else np.inf
        ),
        "annual_pnl_estimate": notional * gross_annual - notional * round_trip_cost,
    }

The trade is real and has been consistently profitable in bull markets. The risks are also real, and they are not price risk:

• Exchange failure. Your collateral sits on a venue. Several venues have failed, taking customer assets with them. This is the dominant risk and it is not diversifiable within one exchange.
• Liquidation of the short leg. A sharp rally can liquidate the perpetual short even though your net position is flat, because the legs are margined separately unless the venue offers proper portfolio margin. You then hold a naked long spot position in a violently moving market.
• Funding inversion. In bear markets funding goes negative and you pay rather than collect. The trade's profitability is regime-dependent.
• Stablecoin depeg if your collateral is a stablecoin.
• Withdrawal restrictions exactly when you most want to exit.

The general shape should be familiar by now: high win rate, steady returns, occasional total loss. Same distribution as the last chapter, different label.

Structural flows

A related family: predictable, calendar-driven, or mandate-driven flows that move prices for non-informational reasons.

• Index rebalancing. Index funds must buy additions and sell deletions on a known date. Documented price pressure around these events.
• Month and quarter end. Pension rebalancing, mandate-driven flows.
• Options expiry. Dealer hedging around large open interest concentrations.
• ETF creation and redemption. Arbitrage flows between an ETF and its basket.

These have a clear mechanism - someone is required to trade regardless of price - which is exactly the property Chapter 2 identified as the signature of a real edge.

They are also extremely well known, which means the price impact has been substantially front-run into earlier dates and smaller magnitudes. The effect is real; the exploitable remainder is thin and shrinking. Do not assume a published effect size from a 2005 paper still applies.

How it kills you

• Sharpe ratio is the wrong metric for negative-skew strategies, and carry strategies post excellent Sharpe ratios.
• Correlated unwinds. Carry trades across different asset classes fail simultaneously, because they are all expressions of the same underlying risk appetite. Your diversification is illusory.
• Counterparty and venue risk, which does not appear in a price-based backtest at all.
• Leverage, which is tempting precisely because the returns look so smooth.