FX Claims, Tested · #3

Do carry trades pay?

“Borrow in a cheap currency, hold an expensive one, and get paid the difference while you wait.”

Verdict: mixed
Eight major currencies · 2004–2026 · BIS policy rates · last run 1 Oct 2026
← All claims tested

The claim

If one currency's interest rate is 4% and another's is 0%, being long the first against the second earns roughly 4% a year, as long as the exchange rate doesn't move against you. Our central bank rates page shows those gaps every day. The question is whether collecting them actually pays once the exchange-rate moves are counted.

The rules, set before we looked at any results

CurrenciesUS dollar, euro, pound, yen, Aussie, Canadian, Swiss franc, Kiwi
Each month-endBuy the three with the highest policy rates, sell the three lowest, equal amounts
What countsExchange-rate moves plus the interest earned or paid every day
Costs0.02% on every change of position (about two pips)
Retail versionAlso takes 1% a year off the interest, roughly what a broker's swap mark-up costs

This is the textbook "G10 carry" strategy that academic studies and fund managers use as the benchmark.

The two halves of every carry trade

Carry has two parts, and they pull in opposite directions.

Interest collected
+2.70% a year
Currency moves
−1.19% a year
Net, after costs
+1.49% a year
Retail version
−0.51% a year

The interest is real and it's reliable: it came in steadily. But the high-yielding currencies, on average, lost some of their value against the low-yielders, giving back 1.2% a year of the 2.7% collected. What's left is small. And once a typical swap mark-up is taken off, a retail trader running the same strategy would have lost money.

Growth of the strategy

Main rule, after costs, before any swap mark-up.

-10%+0%+10%+20%+30%200420082012201620202024
Yearly return
+1.49%
Sharpe
+0.17
Worst drawdown
-36%
2004–14 / 2015–now Sharpe
+0.21 / +0.12

Up the stairs, down the lift

Carry is famous for small, steady gains followed by sudden crashes. In 2008 it lost 25%, then recovered 20% in 2009. When markets panic, everyone unwinds the same trades at once: the high-yielders get sold and the yen and Swiss franc get bought.

2004: +5.3%042005: +14.1%+14%2006: +1.2%062007: +2.5%2008: -24.9%08-25%2009: +19.6%+20%2010: -2.6%102011: +0.1%2012: +6.2%122013: +0.0%2014: +4.2%142015: -8.6%2016: -2.3%162017: -1.8%2018: -3.4%182019: +3.0%2020: +1.7%202021: +6.5%2022: +0.8%222023: +0.3%2024: +6.7%242025: -0.9%2026: +6.1%26

Worst single months

Oct 2008−13.8%
Jan 2015−7.5%
May 2010−7.4%

A month like the worst one here wipes out years of interest.

Who it bought and sold

CurrencyTime long · time short
Kiwi dollar91% · 0%
Aussie dollar77% · 4%
Pound37% · 0%
Canadian dollar36% · 6%
US dollar43% · 33%
Euro16% · 63%
Swiss franc0% · 94%
Yen0% · 100%

Right now the rule would be long US dollar, Pound, Aussie dollar, and short Yen, Canadian dollar, Swiss franc.

Variations, including the classic: AUD/JPY

Buying just the single highest-yielder against the lowest, or two against two, and the trade retail traders know best: long AUD/JPY and never let go.

VersionYearly returnRetail versionSharpeWorst drawdownSharpe by half
Top 1 v bottom 1+3.39%+1.39%+0.23-52%+0.38 / +0.03
Top 2 v bottom 2+3.50%+1.50%+0.30-48%+0.38 / +0.20
Top 3 v bottom 3 (main rule)+1.49%−0.51%+0.17-36%+0.21 / +0.12
AUD/JPY, always long+5.49%+4.49%+0.35-45%+0.39 / +0.32

AUD/JPY looks best at +5.5% a year, but three warnings. It's the famous carry pair precisely because it worked, so choosing it today is hindsight. About 2.5% a year of that came from the yen's long slide rather than interest, which is luck of timing, not carry. And it fell 45% at its worst, including 21% in Oct 2008 alone.

What this means for you.
  • The interest is real. Rate gaps between currencies do pay, and over twenty years the textbook strategy came out slightly ahead.
  • But it's small next to the crash risk. Traders call carry "picking up pennies in front of a steamroller". A month like October 2008 takes back years of gains, and it always comes when everything else is falling too.
  • Check your swap rates. The difference between the wholesale and retail versions here is the whole edge. If your broker's overnight financing is poor, the trade may never have paid you at all.
  • If you do run carry, size it for the crash, not for the quiet months.
What this test doesn't prove.
  • It uses central bank policy rates. The rates banks actually lend at, and the forward prices that set real carry, differ a little from those.
  • Smarter versions exist: cutting positions when volatility rises, or mixing carry with other signals. Those are different claims.
  • The Bank of Japan had no policy-rate target between 2001 and March 2006 (it held short-term rates at effectively zero), so we use 0% for the yen before then.
Risk warning: CFDs and spread bets are complex instruments and come with a high risk of losing money rapidly due to leverage. Most retail accounts lose money when trading them. You should consider whether you understand how they work and whether you can afford to take the high risk of losing your money. Satdish provides education, not financial advice.
Data: BIS central bank policy rates (WS_CBPOL); Yahoo Finance daily closes. Positions rebalanced at month-end using rates known two trading days earlier. Re-run monthly.