Friends, the carry trade is the most seductive strategy in all of finance. Borrow in a currency with a low interest rate. Invest in a currency with a high interest rate. Collect the difference. Go to sleep. Wake up richer. It sounds like free money. For months, sometimes years at a time, it behaves like free money. And then, in a span of days, it gives everything back and more. The carry trade has produced fortunes for those who understand its risk structure and bankruptcies for those who do not. The difference between the two groups is not intelligence or even skill. It is whether they understood what they were actually being paid for. Here is what the academic literature documents consistently: carry trade returns are not compensation for holding a high-yielding currency. They are compensation for bearing volatility risk. The interest rate differential is just the delivery mechanism. The actual product is insurance against a crash that, when it comes, will arrive without warning and with extraordinary violence.
The Textbook Says This Should Not Work
Before we get to the institutional evidence, it helps to understand why carry trades are theoretically puzzling. Standard economic theory predicts they should not be profitable. Uncovered interest parity (UIP) says that a currency with a higher interest rate should depreciate over time by exactly enough to offset the yield advantage. If the US pays 5% and Japan pays 0%, the dollar should weaken 5% against the yen over the year, leaving the carry trader with zero net profit. In practice, the opposite happens. High-interest-rate currencies tend to appreciate in the short run, not depreciate. This is the forward premium puzzle, one of the oldest and most persistent anomalies in international finance. Chaboud and Wright at the Federal Reserve examined the same anomaly at the opposite end of the horizon range, over intraday windows; the multi-year result often attached to them is Meredith and Chinn’s, and neither paper tests the one-to-twelve-month range where carry trades are usually held: www.federalreserve.gov This means carry traders are exploiting a genuine market anomaly. The question the academic literature has spent decades answering is: why does this anomaly persist? If it is genuinely free money, why hasn’t it been arbitraged away? The answer, as it turns out, is that it is not free money. It is cheap insurance. And the premium is collected slowly until the moment the policy pays out.
The 90% Finding: Volatility Is the Carry Trade
The definitive quantitative answer comes from Menkhoff, Sarno, Schmeling, and Schrimpf, who published “Carry Trades and Global Foreign Exchange Volatility” in the Journal of Finance in 2012. They constructed a global FX volatility measure and tested whether it could explain the cross-sectional returns of currency portfolios sorted by interest rate. The paper is available on SSRN: papers.ssrn.com Their finding: their global FX volatility proxy accounts for more than 90% of the spread across five carry-sorted portfolios — the gap between the highest- and lowest-yielding baskets, not 90% of any one currency’s return. High-yield currencies earn positive excess returns during periods of low and stable volatility. Low-yield currencies provide a hedge during periods of high volatility. The carry premium is, at its core, a volatility risk premium. This paper is not an outlier. Lustig, Roussanov, and Verdelhan (2011), in “Common Risk Factors in Currency Markets” published in the Review of Financial Studies, identified an HML factor (high-minus-low interest rate) in exchange rates that is closely related to global equity volatility. Currencies with high interest rates load positively on this factor. Currencies with low interest rates load negatively. The factor explains most of the cross-sectional variation in currency excess returns: www.nber.org
Those two results point the same way and are worth keeping apart rather than merging: Menkhoff and co-authors measure global FX volatility, while the Lustig, Roussanov and Verdelhan factor is related to global equity volatility. Different factors, different papers, different data. One caution on the citation as well — the equity-volatility wording is in the published Review of Financial Studies abstract, while the 2008 working-paper version linked above says only “a global risk factor”.
The retail trader translation: when you go long USDZAR or USDMXN or any high-yielding EM currency, you are not simply collecting a yield premium. You are selling volatility insurance. You will collect small, steady premiums during calm periods. And you will pay out a large, sudden claim when volatility spikes. The interest rate differential is the premium income. The crash is the claim.
How Carry Trades Crash: The Brunnermeier Evidence
The dynamics of carry trade crashes were formalized by Brunnermeier, Nagel, and Pedersen in “Carry Trades and Currency Crashes,” published in the NBER Macroeconomics Annual in 2008. The paper is available from the University of Chicago Press: www.journals.uchicago.edu Their findings established three properties of carry trade returns that every FX trader should understand: First, carry trade returns are negatively skewed. The distribution has a fat left tail. Most of the time, returns are small and positive. Occasionally, returns are large and negative. This is the statistical signature of a short-volatility position. Second, crash risk increases with the size of the interest rate differential. The higher the carry, the larger the potential crash. This is not a coincidence. Larger rate differentials attract more capital into the trade, which creates larger positions that must be unwound during stress. The size of the carry is proportional to the size of the risk. Third, carry trade losses are correlated with increases in VIX. When equity volatility rises, carry trades lose money systematically. This is the leverage constraint mechanism we discussed in Parts 1 and 2 of this series. Rising VIX tightens intermediary balance sheets, forces deleveraging, and triggers the carry trade unwinds that produce crash returns. The self-reinforcing nature of the crash is critical to understand. When carry positions begin to unwind, the funding currency (typically JPY or CHF) appreciates. This appreciation causes further losses for remaining carry traders, triggering more stop-losses and margin calls, which causes more unwinding, which causes more appreciation. The feedback loop accelerates until enough positions have been liquidated to exhaust the selling pressure.
Bond Volatility: The Early Warning System
Here is the pearl of this article, the finding that connects carry trade risk directly to the MOVE index discussion in Part 1 of this series. A 2025 paper in the Journal of Financial and Quantitative Analysis, “Currency Carry, Momentum, and Global Interest Rate Volatility,” demonstrated that exposures to global interest rate volatility (a MOVE-type measure applied globally) explain 92% of the cross-sectional return variations in both carry and momentum currency strategies (Ming Zeng (2025), Journal of Financial and Quantitative Analysis 60(2), 839–873). The paper is available through Cambridge University Press: www.cambridge.org This finding is why this monitor reads bond volatility first for carry trade risk. It treats MOVE in three states, as a convention rather than an observed sequence: low and stable, which it reads as the premium-collection phase; rising, which it reads as the claim approaching; and spiking, which it reads as the unwind in progress. The ordering below is this site’s working heuristic, not a measured lead time. Part 1 describes MOVE moving ahead of VIX in stress episodes; the analysis cited there establishes only conditional predictive power when both indices are already elevated, and reports no interval in days or hours. The sequence this monitor watches for is: MOVE rises → intermediary VaR constraints tighten → deleveraging begins → carry positions start unwinding → VIX rises → the crash becomes visible in price action. Read it as a sequence of conditions rather than a countdown, and as this monitor’s convention rather than a measured order: it treats a move in the carry pair itself as the last condition in that sequence, not a leading one.
The August 2024 Unwind
The most violent recent carry trade unwind occurred in August 2024, and the BIS provided a detailed post-mortem in Bulletin No. 90: www.bis.org The trigger was a Bank of Japan rate signal suggesting further policy normalization. JPY began appreciating. Carry traders who had borrowed in yen to fund positions in the Mexican peso, Australian dollar, and other high-yielders faced losses on the funding leg. As the unwinding accelerated, VIX briefly exceeded 60, a level previously seen only during the 2008 financial crisis and the 2020 COVID crash. The BIS put a number on the stock that had built up over years of zero-rate policy rather than leaving it at “enormous”: “Various estimates based on both on- and off-balance sheet activity yield a rough middle ballpark of ¥40 trillion ($250 billion) going into the event.” The unwinding compressed into three days. The Mexican peso lost nearly 10% against the yen. The Australian dollar was hit hard. Even the US dollar weakened against JPY, despite being the other major safe haven.
The sequence matched the academic predictions precisely. The initial JPY appreciation triggered stop-losses. The stop-losses triggered further JPY buying. The further JPY buying triggered margin calls. The margin calls triggered forced liquidation. The forced liquidation triggered more JPY buying. The feedback loop ran until enough positions were gone to break the cycle.
EM Carry Pairs: The Extreme Asymmetry
Emerging market carry pairs, USDZAR, USDMXN, USDBRL, USDTRY, exhibit the most extreme version of the carry trade asymmetry. The interest rate differentials are large, often 5 to 10 percentage points or more. The carry income is substantial. And the crash risk is proportionally enormous. The ECB’s Economic Bulletin (2019) decomposed EM FX movements into four factors: a dollar factor, a carry factor, interest rate differentials, and idiosyncratic domestic shocks: www.ecb.europa.eu The critical difference between EM carry and G7 carry is the asymmetry of returns. G7 carry pairs (like USDJPY or AUDUSD) produce moderate carry income with moderate crash risk. EM carry pairs produce high carry income with extreme crash risk. The distribution of returns is far more negatively skewed for EM pairs because their markets are less liquid, their capital accounts are more sensitive to global flows, and the positions are more crowded during good times. Hambuckers and Ulm (2023) measured the relationship between the rate differential and crash risk directly, and not on emerging currencies. From the paper’s abstract: “the larger the difference between interest rates, the more likely the high-yield currency appreciates on average but also exhibits greater risk of a large depreciation.” The currencies for which they find the differential significant are the euro, the Swiss franc, the Swedish krona, the yen and sterling. None is an emerging market currency. A sharper version of the same mechanism in emerging pairs is a reasonable expectation from their market structure, and it is not something this paper measures: www.sciencedirect.com For retail traders, the practical implication is sobering. The pairs with the most attractive carry are the pairs with the most violent reversals. USDZAR, USDMXN, and USDTRY produce beautiful equity curves during calm markets and catastrophic drawdowns during stress. The six months of steady gains can evaporate in a single week.
The Carry Trade Is Not a Position. It Is a Regime Bet.
Here is the framework shift that institutional carry traders understand and most retail traders do not: entering a carry trade is not a bet on the interest rate differential. It is a bet on the volatility regime.
When you go long USDZAR or short AUDJPY or take any other carry position, you are making a specific prediction: the volatility regime will remain calm long enough for the carry income to accumulate. If you are right, the carry accrues steadily and the position is profitable. If the regime shifts before you have accumulated enough carry to absorb the drawdown, you lose. This is why the volatility gate sits at the top of the transmission chain. It is not enough to know the rate differential. It is not enough to know the yield curve shape. It is not enough to have a view on the dollar. If the volatility regime is stressed, none of those things matter for carry trades. The regime determines whether carry accrues or carry crashes. The Federal Reserve’s FEDS Note on monetary policy and exchange rates quantified the non-carry component of FX returns. Their model combined 2-year OIS differentials with VIX and high-yield credit spreads, and the risk measures added significant explanatory power beyond rate differentials alone. About half of the dollar’s appreciation during the 2021-2024 tightening was attributable to rates. The other half was attributable to risk appetite: www.federalreserve.gov This means carry returns are roughly 50% rate differential and 50% volatility regime. Ignoring either half leaves you exposed to the half you missed.
What the Evidence Conditions On
Read together, these findings point one way: the carry trade’s return distribution is conditional on the volatility regime, and not on the size of the rate differential alone. Menkhoff and co-authors attribute most of the cross-sectional variation in carry returns to global FX volatility. Brunnermeier, Nagel and Pedersen document that the losses arrive as crashes rather than as drift. August 2024 ran the sequence in public. Three things follow, and they are readings of the evidence rather than instructions.
First, a wide rate differential is not by itself information about the return. It is the compensation being offered; the volatility regime is what determines whether it survives being collected. The literature keeps those two readings separate, and so does this site.
Second, the crash risk is conditional rather than a property of exchange rates in general. Brunnermeier, Nagel and Pedersen are explicit that currencies have no unconditional skewness — the skewness of a randomly chosen pair is zero, because country A’s positive skewness is country B’s negative skewness. The negative skewness belongs to the carry position, not to the exchange rate.
Third, the horizon matters, though not on the authority this article previously gave it. Chaboud and Wright examine uncovered interest parity over intraday horizons, in their own words “going in exactly the opposite direction, examining UIP over extremely short horizons.” The multi-year result belongs to Meredith and Chinn, whom Chaboud and Wright cite in their literature review. Neither paper tests the one-to-twelve-month range where the puzzle is usually stated, and this article previously attributed that range to Chaboud and Wright.
This is the ordering the 4xForecaster FX Monitor renders. The volatility environment is read first, the rate structure second, and the pair-level bias last. Carry-sensitive pairs — USDZAR, USDMXN and the commodity and EM complex — carry explicit regime conditioning, so their published conviction is capped while volatility is elevated. The monitor reports a condition. It does not size anything, and it does not tell you what to do about it. Because the carry trade is not a strategy. It is a regime bet. And the regime comes first.
This is Part 9 of the Macro-to-FX Transmission Series from 4xForecaster. Next: Sector Rotation as a Currency Signal: What XLU, XLE, and XLF Tell You About the Dollar.