The Carry Trade: Earning the Rate Gap Every Night, and Why It Can Crash in Days

A carry trade means borrowing a currency with a low interest rate, holding a currency with a high one, and collecting the difference over time. It looks like income that arrives on its own every night, and in calm markets that is roughly what happens. The problem is that the income comes slowly and the losses come fast. This article explains the mechanics, the theory that says the gain shouldn’t exist, what the data actually shows, and works out how small a currency drop it takes to erase a whole year of carry.

How a carry trade works

Institutions do it directly: borrow the low-yield currency, convert it into the high-yield one, and park it in deposits or short-term bonds. Retail traders do it through a currency pair in a CFD account instead. Call the high-yield currency H and the low-yield currency L. Buying the pair H/L means holding H and borrowing L to pay for it.

The income arrives as swap, credited every night the position is held past the broker server’s rollover. The Buy side receives positive swap, the Sell side pays negative swap. How swap converts into money, the triple-swap day and how to read it in MT5 are covered in What is swap and how overnight fees work. This article is about the layer above that: whether carry pays off depends far more on price than on swap.

The rate gap behind swap comes from short-term money-market rates, which track policy rates closely but not exactly. For why policy rates move currencies at all, read Why interest rate decisions move currencies and gold.

Theory says the gain should disappear

If one currency paid 4.5% a year more than another with no strings attached, everyone would pile in. So theory says something must offset it.

Covered interest parity is the part that holds with certainty. The forward price (the price agreed today for a future exchange) of the high-yield currency must sit below today’s price by the rate gap, otherwise anyone could lock in a free profit by pairing trades. Roughly, for H/L: forward = spot × (1 + L rate) ÷ (1 + H rate). With spot at 100.00, H at 5% and L at 0.5%, the one-year forward = 100 × 1.005 ÷ 1.05 = 95.71, about 4.3% below today, close to the rate gap. Holding spot and collecting swap every night is the same deal rolled forward one day at a time.

Uncovered interest parity (UIP) goes further. It says the price you should actually expect in a year is also around that forward price. In plain words, the high-yield currency should weaken by just enough to cancel the extra interest. You earn the interest and lose it on the exchange rate, and you end up where you would have been holding the low-yield currency. If UIP held, a carry trade would have no average profit, only risk.

The data mostly disagrees

A large body of research since the 1980s finds that UIP works poorly over short and medium horizons. A widely cited study is Eugene Fama (1984), which tested whether the rate gap predicts currency moves the way theory says. The result, repeated in many later studies, is that on average high-yield currencies have not weakened as much as the rate gap, and in some periods they strengthened. This is known as the forward premium puzzle, and it is one of the best documented findings in international finance.

As a result, in research, carry trades spread across many pairs show positive average returns in many periods. Read that carefully:

• It is an average across many pairs and decades, not one pair in one year
• It is measured at institutional costs, not retail swap after a broker’s markup
• It includes deep losing stretches that wiped out years of income

The explanation many researchers agree on is that the positive average is payment for a specific risk: the risk of large losses at exactly the moments when everything else is going badly too. Past results do not predict future results.

Picking up pennies in front of a steamroller

Traders describe the carry trade as picking up pennies in front of a steamroller. You collect small amounts steadily until the day the steamroller arrives. In statistics this is negative skew: mostly small gains, occasionally a large loss.

Brunnermeier, Nagel and Pedersen (2008), in “Carry Trades and Currency Crashes”, found that high-yield currencies tend to fall sharply and suddenly rather than spike up, and that the falls tend to come when fear in markets rises. When many traders hold the same carry positions, they have to close them at the same time, and each leg lower forces others to close too. The mechanism goes like this:

1. In risk-on periods, money flows to high-yield currencies. The exchange rate is calm or firm, carry looks easy, and more traders join
2. Something scares the market (risk-off), or the central bank of the funding currency signals a hike. Traders start closing carry trades
3. Closing means selling the high-yield currency and buying back the funding currency. The borrowed currency (often one seen as a safe haven) strengthens sharply
4. Losses tighten margin, forcing more closing, and the loop repeats

Well-known examples are the 2008 financial crisis, when popular yen carry pairs fell hard within a few weeks, and early August 2024, when the yen strengthened fast after the Bank of Japan raised rates, an episode the market called a carry trade unwind. Read more about this kind of market mood in Risk-on, risk-off.

Note that many carry pairs fall together in these periods. Holding several pairs is not the diversification it seems, because they are all the same bet: “markets stay calm”. See Correlation and portfolio risk.

The chart uses the same numbers as the example in the next section. Over the first 600 days, carry adds up to 2.877 × 600 = 1,726 dollars. Then price falls 12.0% in 10 days, and at 300 units the price loss is 3,614 dollars, more than all the carry collected over almost two years.

The carry trade earns like climbing stairs and loses like taking the elevator. Size it for the elevator day, not for the stairs.

Worked example: a year of carry, and the drop that erases it

All figures are made up, not today’s real rates. To keep the arithmetic simple, assume the account and the position are both valued in dollars and ignore currency conversion.

• Equity 10,000 dollars, Buy H/L with a notional of 30,000 dollars at 100.00 (that is 300 units of price). Leverage actually used = 30,000 ÷ 10,000 = 3 times
• H rate 5%, L rate 0.5%, rate gap = 4.5% a year
• Assume the broker takes 1% a year from the receiving side. Net carry = 4.5% − 1% = 3.5% a year

Step by step:

1. Carry before the markup, per year = 30,000 × 4.5% = 1,350 dollars
2. Broker’s share per year = 30,000 × 1% = 300 dollars, or 300 ÷ 1,350 = 22% of the whole carry
3. Net carry per year = 30,000 × 3.5% = 1,050 dollars, or 1,050 ÷ 365 = 2.88 dollars a night
4. Price loss = notional × the percentage fall. Break-even is 30,000 × d = 1,050, so d = 3.5%. A move from 100.00 down to 96.50 eats the whole year of carry
5. Without the markup, break-even would be 4.5%, or a price of 95.50. The markup narrows your cushion by a full percentage point

Check: at 96.50 the loss is 300 × (100 − 96.5) = 1,050 dollars, exactly one year of net carry.

Compare with UIP, which says price should sit around 95.71 in a year. That is below our 96.50 break-even, so if the theory held exactly, a retail trader paying the markup would lose on average. Carry only comes out positive when the high-yield currency weakens less than theory predicts. That happened often in the past, but nothing guarantees it.

A 3.5% move over a year is not far at all for a high-yield pair. On a day the market panics, that distance can disappear within days.

The broker’s share in retail swap

The swap a retail trader receives is not the full rate gap. Three things make it smaller than expected:

• The broker’s markup comes off both sides. The receiving side gets less, the paying side pays more. If the rate gap is narrow, swap can be negative on both sides
• The rate base is short-term money-market rates, not the policy rate you read in the news
• It can change at any time, when a central bank moves rates or the broker changes its terms. And the day a rate is cut is often the day that currency weakens, so carry is hit twice

To check what you actually earn, convert swap per night into an annual percentage:

Annual carry (%) = swap per night in money × 365 ÷ notional

For the example above, 2.88 × 365 ÷ 30,000 = 3.5%. Compare that with the gap between the two countries’ policy rates. The difference between the two numbers is roughly the part that never reaches you. Where to find the values in the symbol specification and how to convert points to money are covered in the swap article above. Terms differ between brokers, so always check the specification of the account you actually use.

Leverage magnifies both the income and the pain

The 3.5% break-even does not depend on leverage, because both the carry and the price loss are calculated on the same notional. What leverage changes is the size of the money relative to your equity. Equity 10,000 dollars, net carry 3.5%:

Leverage used	Notional	Carry per year	Cost of a 10% fall	Fall that costs half the account
1×	10,000	350	1,000	50%
3×	30,000	1,050	3,000	16.7%
5×	50,000	1,750	5,000	10%
10×	100,000	3,500	10,000 (account gone)	5%
20×	200,000	7,000	More than the account (stop-out first)	2.5%

The carry column looks attractive as leverage rises, but it is what you earn if price doesn’t move at all for a year, which never happens. At 20 times, a 2.5% fall costs half the account, a smaller distance than the 3.5% break-even itself. In practice the account is stopped out before equity reaches zero, and in a fast market price can gap through the stop-out level. See how this works in Leverage, margin and stop-out, and try the numbers with the margin calculator.

Leverage doesn’t make a carry trade safer or riskier in terms of break-even. It decides whether you still have enough equity to keep trading on the day price falls.

Using this when you trade

1. If you hold the positive-swap side, count swap as a bonus, not the main reason for the trade. In this example, a 3.5% move against you erases a full year of income
2. Size for the hard drop. Ask what the account loses if this pair falls 10% in two weeks, and whether you can live with that. Use Position size from your Stop loss and Risk 1% per trade as the frame
3. Watch both central banks’ meeting dates. A cut by the high-yielder, or a hike by the funding currency, can change everything in a day. Find the dates in the economic calendar
4. Watch market mood. When fear starts to build, carry pairs tend to fall together. Treat several carry pairs as one big position
5. Know the paying side too. If you Sell a high-yield currency, you pay carry every night, and the trade has to earn enough from price to cover that rent
6. Be careful with backtests. The Strategy Tester uses today’s swap from the specification across the whole data period, so a system that looks good because of positive swap may differ a lot from what really happened

Common mistakes

• Treating positive swap as free money. It is payment for taking the risk of a hard price fall, not interest on a deposit
• Using policy rates to estimate income. What you actually get is swap after the broker’s markup, which in the example takes 22% of the carry
• Raising leverage to make the carry worthwhile. Break-even stays the same, but the distance you can survive before a heavy loss shrinks with leverage
• Thinking several pairs means diversification. Most carry pairs fall together when markets get scared
• Holding on because you don’t want to give up the carry collected. In the example, 600 days of carry vanished in a 10-day fall. Waiting for swap to repay a loss usually takes years
• Believing that what worked in the past must work again. Research averages span many pairs and decades, include deep losses, and do not predict the future