Why does a carry trade give back years of gains in a week?
Carry returns have a distinctive and well-documented shape: a long series of small positive months, interrupted rarely by an enormous negative one. Practitioners describe it as going up by the stairs and down by the elevator, and less kindly as picking up pennies in front of a steamroller. Understanding why the shape is like that is the point of this lesson.
The statistical signature
Two words describe it. Negative skew: the losses are far larger than the gains, even though the gains are far more frequent. Excess kurtosis: extreme outcomes happen much more often than a normal distribution would suggest.
This is not folklore. Brunnermeier, Nagel and Pedersen documented in 2008 that high-interest-rate currencies are subject to crash risk that is negatively skewed, that these crashes cluster with rises in equity-market volatility indices, and that they coincide with reductions in speculative positioning — a mechanical, positioning-driven unwind rather than a change in economic fundamentals.
The mechanism
The unwind is self-reinforcing, and each step follows from the last:
- The trade is crowded, because the differential is public and the logic is simple. Many people are in the same position.
- It is leveraged, because the raw carry is only a few points a year.
- Volatility rises for any reason at all — a policy surprise, an equity shock, a credit event.
- Risk limits and margin calls force position reduction. Reducing this position means buying back the funding currency.
- Everyone buys the funding currency at once. It surges, which widens losses, which forces more reduction.
Notice what is absent: nobody has to change their view on Japan or Australia. The exit is mechanical.
The arithmetic of asymmetry
A trade earning a 4.0% differential a year needs a long time to build a cushion. A 30% adverse move in the exchange rate erases
30 / 4 = 7.5 years of carry
and it can happen in weeks. That ratio — years of accumulation against weeks of loss — is the honest description of the risk, and no amount of history in which the trade worked changes it.
Two documented episodes
2008. AUD/JPY traded at roughly 105 in July 2008. By late October 2008 it was near 55. That is close to a 50% fall in under four months, during which the interest differential kept paying its few points a year, entirely irrelevantly.
August 2024. The Bank of Japan raised its policy rate to around 0.25% on 31 July 2024, at a moment when yen-funded positions were widely reported to be large. USD/JPY fell from about 162 in early July to about 142 by 5 August. On that day the Nikkei 225 fell 12.4% — its largest single-day points decline — and US volatility indices spiked intraday above 60. The trigger was a 15 basis point policy move. The reaction was a global de-risking event.
What this means for reading FX
Two habits worth keeping. First, when a funding currency rallies hard and fast without any domestic news to explain it, positioning is a more likely explanation than fundamentals. Second, a strategy's long, smooth track record tells you about the frequency of losses, not their size — and for carry those are two completely different questions.
Nothing here is a view on any currency or a suggestion to hold, avoid, or fade any position. These are factual accounts of documented past episodes.
Try it now
- The longest history this page holds for the carry pair is below. Switch it to Monthly — at this length daily bars are noise — and look at the shape the lesson describes: a long patient climb, then a cliff.
- Measure the largest three-month decline you can find and note the dates it spans. Write down both the percentage and the number of bars.
- Using a rounded 4.0 point annual differential, work out how many years of carry that single decline would have erased. Write one neutral sentence describing the result — no inference about what comes next.