Carry Trades and Tail Risk

The currency carry trade is the second-most-documented style premium in the FX literature, after trend following. It is also the trade with the most distinctive shape of returns: long stretches of small positive nights, punctuated by occasional violent down days that wipe out months of accumulation. The academic literature has a clean phrase for this: “picking up pennies in front of a steamroller”. The metaphor captures the trade better than any equation.

This article walks the structure: what the carry trade is, why the premium exists, what the published evidence on its returns and skewness shows, the funding-currency dynamics that drive its crash episodes, and the August 2024 yen unwind as the canonical recent case study that retail traders should understand before considering the trade at all.

What the carry trade is

The mechanics are simple. Borrow in a low-yielding currency. Convert to a high-yielding currency. Hold. The interest-rate differential accrues to the position every night. As long as the exchange rate does not move against you by more than the accumulated rate differential, the trade is profitable.

The textbook version: in a world where uncovered interest rate parity (UIP) held, the high-yielding currency would be expected to depreciate against the low-yielding currency by exactly the interest-rate differential. The trade would have zero expected return.

The empirical version: UIP fails. The high-yielding currency, on average, does not depreciate by the interest differential. It depreciates by less, or sometimes appreciates. The carry trader keeps most or all of the rate differential as profit. This is the forward premium puzzle, documented in the FX literature since Fama (1984) and confirmed across decades, samples, and methodologies.

In retail practice, a long-USD/JPY position is the classical carry trade: long the higher-yielding dollar, short the lower-yielding yen. Long-AUD/JPY and long-NZD/JPY were even larger carry trades in the 2003-2008 and 2014-2019 eras when those currencies offered substantial rate advantages over the yen.

What the academic record shows

The carry premium is unusually well-studied. A condensed summary of the most-cited findings:

• Fama (1984) documented the original forward premium puzzle: the high-interest-rate currency does not depreciate by the amount UIP predicts.
• Lustig, Roussanov, and Verdelhan (2011) (“Common Risk Factors in Currency Markets”) constructed a cross-sectional currency-carry portfolio (long the high-yield basket, short the low-yield basket). The portfolio produces a Sharpe ratio comparable to or higher than the equity market over their multi-decade sample, with annualised excess returns in the 5-9% range.
• Brunnermeier, Nagel, and Pedersen (2008) (“Carry Trades and Currency Crashes”) is the most-cited paper on the trade’s tail behaviour. They document that carry-trade returns have strongly negative skewness, with the worst drawdowns concentrated in periods of rising risk aversion and funding-liquidity stress. The “currency crash” pattern is the defining property of the strategy.
• Burnside, Eichenbaum, and Rebelo (2011) examine whether the carry premium compensates for risk or represents a true inefficiency. Their conclusion: the premium is real, but its measurable risk loadings do not fully explain its magnitude, leaving open the possibility of structural inefficiency combined with crash-risk compensation.
• Menkhoff, Sarno, Schmeling, and Schrimpf (2012) show that currency volatility innovations price the cross-section of carry returns. High-carry currencies systematically lose value when global FX volatility rises.

The summary that survives across all of this: the carry trade has a positive long-run premium, a sharply negative skew, and a risk profile dominated by occasional violent unwinds. The average is good. The tail is terrible.

The shape of the returns

The picture above is not unique to any one currency pair. It is the structural shape of the strategy. The same pattern shows up in long-AUD/JPY through 2007, in long-MXN/JPY through 2015, in long-NZD/JPY through 2019, and most recently in long-USD/JPY through July 2024.

The arithmetic underneath:

• Per-night carry income, in basis points, is small but positive across a wide range of regimes.
• Per-day price risk is also small in calm regimes, slightly net favourable on average across long-run samples.
• Per-crash event, the price move is several standard deviations larger than the accumulated carry. A two- or three-month carry accrual can be erased in a single day’s unwind.

The expected return calculation is positive across the full distribution. The realised return path, however, depends on whether you are still in the trade when the crash arrives.

Why the crashes happen

The funding-currency mechanic is the same in every episode. Carry trades concentrate in the same low-yielding currencies (historically the yen, more recently also the franc and the euro). When risk-off conditions emerge, several things happen at once:

• Investors reduce leverage across portfolios broadly.
• Carry positions, which are levered by construction, are among the first to be cut.
• Closing a long high-yield / short low-yield position requires buying back the low-yield (funding) currency. The funding currency rises sharply.
• Margin calls and stop-outs accelerate the unwind.
• The funding currency continues to rise as the unwind feeds on itself.

The result is a violent reversal of the carry pair, often 5-10% in a few days. The asymmetry is structural: the carry pair appreciates gradually as the trade is put on by many participants over time, and unwinds rapidly when the same participants need to exit at once.

The 2008 crisis produced the textbook example in AUD/JPY (down roughly 40% in late 2008 from the 2007 peak). The COVID shock in March 2020 produced a smaller version. The 2024 episode in USD/JPY, covered below, is the most recent and the most informative for current retail traders.

The August 2024 USD/JPY unwind

The recent canonical example. Through July 2024, USD/JPY was trading above 161, the weakest yen levels since 1986. Long-USD/JPY had been the dominant retail and institutional carry trade for years, propelled by:

• A wide Fed-BOJ policy gap (Fed funds around 5.25-5.50%, BOJ at -0.10% then 0.10%).
• BOJ’s continued accommodation, including the lingering effects of Yield Curve Control.
• A market consensus that the yen would weaken further.

The unwind began with two near-simultaneous shocks. The BOJ delivered a 15bp rate hike on 31 July that markets had not fully priced. The 2 August US Non-Farm Payrolls printed materially weaker than consensus, repricing the Fed’s expected path lower.

The combination accelerated through 5 August. USD/JPY fell roughly 12% in three weeks, from above 161 to under 142. The move triggered a global cross-asset risk-off, with the VIX spiking to 65 intraday on 5 August. The unwind was not unique to yen pairs: AUD/JPY, MXN/JPY, and other carry crosses fell harder than USD/JPY. The shape was the textbook carry crash: years of slow accumulation, weeks of violent reversal.

The retail trading community absorbed substantial losses in the episode. The brokerage disclosure reports for the relevant quarters reflected the elevated stop-out rates. The lesson, which the published literature had documented for 16 years before the event, is the same after as before: the carry trade is a positive-expectancy strategy with sharply negative skew, and the skew is not theoretical.

Can it be traded at retail

The carry trade can be traded at retail. Whether it should be is a more careful question. The conditions for a defensible retail implementation:

• Position size set against the crash, not the calm. If the position is sized so that a 10% adverse move would liquidate the account, the strategy will be exited at the worst possible time and the long-run premium will not be captured. Sizing for a 15% sustainable drawdown is closer to honest.
• Diversification across multiple carry pairs. A single-pair carry trade is highly concentrated. A small basket (AUD/JPY, NZD/JPY, plus the dominant USD/JPY carry) reduces some of the idiosyncratic risk, though the crash risk is largely correlated and does not diversify away.
• Awareness of the regime. Carry strategies tend to do well in calm, low-volatility environments and badly when global vol rises. Some monitoring of cross-asset volatility (the VIX is a serviceable proxy) gives a small but real edge in reducing position size before regime shifts.
• A multi-year horizon. The carry premium shows up over long windows. In any given quarter, the return is dominated by short-term price movement of the pair.
• Acceptance that the swap income is not free. Some retail traders treat the carry as a quasi-deposit yield. The academic record is unambiguous: the swap income is the premium, and the premium exists because of the crash risk that periodically realises.

If those conditions sound restrictive, that is the right reaction. The carry trade is one of the most empirically well-supported strategies in finance, and it is also one of the most punishing strategies to trade carelessly.

The takeaway

The currency carry trade earns a documented long-run premium of 5-9% annualised in cross-sectional implementations, with a Sharpe ratio comparable to equities. The premium is paid for with sharply negative skewness: long calm periods of gradual accumulation, occasional violent unwinds that erase months of gains in days. The August 2024 yen episode is the most recent demonstration of the pattern; the 2008 AUD/JPY unwind is the larger historical analogue.

For retail traders, the strategy can work in principle but only when sized against the crash rather than the calm, diversified across multiple carry pairs, monitored against the volatility regime, and held over multi-year horizons. The most common retail framing (“long USD/JPY for the swap”) is approximately correct as a description and dangerously incomplete as a risk model.

For the companion piece on the other major style premium in FX, read Trend Following: The Long-Run Evidence. For the position-sizing framework this strategy specifically requires, read Risk Management Basics and Currency Correlation and Hidden Risk.