A plain-English guide to yield differentials and FX carry.

USD/JPY closed at 162.64 yesterday, its highest print in more than two years. The paired analysis piece, USD/JPY through 162 as yields push, reads the move as one leg of a rate-driven carry trade. That reading depends on a specific understanding of what a "yield differential" is and what it drives. Casual commentary uses the phrase for anything that resembles a rate gap; the desk usage is more constrained. This piece explains the constraints.

Yield differentials are, on the professional desk, the single cleanest driver of every G10 cross. The correlation of the two-year swap-implied differential to spot in USD/JPY is above 0.85 on a rolling-90-day window through most of the last decade. In EUR/USD it sits around 0.6 to 0.7. In AUD/USD, 0.7 to 0.8. When a G10 pair moves and the differential does not, the move fades; when the differential moves and the pair does not, the pair catches up. That empirical stability is the point of the framework.

The differential that matters

The version that drives price is the swap-implied yield differential, not the spot cash-rate gap. The two are different because a currency pair's carry is priced against the curve of expected future rates, not against the current level. When the market expects the Fed to hold and the BoJ to hike, the swap curve prices that in ahead of the actual moves, and USD/JPY responds to the swap gap even before either central bank has changed policy.

The mechanics: take the two-year OIS-implied yield in each currency (see the OIS reference). Subtract to produce the two-year swap-implied differential. This is the number the desk marks. The two-year (not the five, ten, or thirty) is picked because it captures both the current stance and the near-term rate path expectation, which is where the FX carry pays or does not.

For pairs where the desk trades the front-end more aggressively (AUD/USD, NZD/USD, CAD/USD), the one-year swap-implied differential often works better. For pairs where the back-end matters (EUR/USD, GBP/USD in certain regimes), the five-year can be more informative. The two-year is the default because it is the horizon at which almost every central bank's policy forecasts start to matter.

What "carry" actually is

Carry is the yield you earn (or pay) for holding a currency-pair position. For USD/JPY long, carry is the US short rate minus the JPY short rate, adjusted for the roll on the forward. When US rates are higher, a long USD/JPY position pays; when they are lower, it costs. The current annualised carry on a long USD/JPY position is roughly +4.0% for the desk's overnight funding tape, or about 1.1 basis points per day.

Carry decomposes into two components: hedged and unhedged. The unhedged component is what a speculative trader captures. The hedged component is what a real-money account (a Japanese life insurance company, say) captures after paying for the FX forward. When the hedged carry is negative, real money hedges its dollar exposure and buys yen back on the forward, which supports the yen. When hedged carry is positive, real money can hold the position without hedging, and the yen bid disappears.

In 2025-2026 the hedged carry for a Japanese account holding US Treasuries has been negative to marginal because the JPY-USD FX forward premium (basis) has priced through the differential. The hedge is expensive, so real money hedges less, so the yen has less structural bid. That real-money mechanic is why the 2026 yen carry trade has been the trade of the year even for accounts that do not usually speculate.

The three regimes yield differentials produce

Understanding the mechanics is one thing. Trading them is another. Yield differentials drive spot cleanly in three regime archetypes and produce nothing but noise in a fourth.

1. Stable-differential regime. The 2Y differential is unchanged week over week; policy expectations are set. Spot grinds around the implied fair value with a low realised volatility. Carry-earning positions pay their theoretical annualised return with minimal noise. This is the boring regime; it is also the one the yen carry has been in for most of 2025.
2. Widening-differential regime. The 2Y differential is expanding; one central bank is being priced hawkish while the other is being priced dovish. Spot trends cleanly toward the receiving-side currency. Realised volatility picks up on the trend side. The June 2026 push through 161-162 is a widening-differential episode; the US 10Y has climbed 25bp from the June low while the JGB 10Y has been flat.
3. Compressing-differential regime. The 2Y differential is shrinking; policy expectations are converging. Spot reverses toward the paying-side currency, usually with a lag as positioning unwinds. This is the regime the yen bulls are waiting for; it typically follows a hawkish BoJ event or a dovish Fed shift.
4. Noise regime. The 2Y differential is unchanged, but spot is moving. This is the case where the framework does not explain the tape and something else is in control (intervention, risk-off, positioning liquidation, a geopolitical shock). When you see spot move and the differential does not, the correct next step is not to trade the pair against the yield gap; it is to ask what else is happening.

Where this shows up in practice

Recent analysis pieces that lean on parts of this framework:

• USD/JPY through 162 as yields push. The widening-differential read for the current break.
• The dollar is weaker on the data, not the dots. The stable-differential read for the DXY drift lower in May.
• The plain-English guide to the carry trade. The financing side of the differential trade, including the leverage math.
• The plain-English guide to OIS. The instrument the swap-implied differential is priced from.

What the framework misses

Yield differentials do not explain intervention risk. They do not explain positioning-liquidation events. They do not explain step-changes in cross-currency basis or in the swap-cash gap. In each of those regimes, spot moves in ways the differential cannot price. The framework is a workhorse, not a universal explanation. When the paired analysis piece says "the rate leg did the work," it means "the yield differential explains the move today"; it does not mean the yield differential is the only variable that matters.

The single test we run before deploying the framework on a fresh move: is the 2Y differential change of the same sign and roughly the same magnitude as the spot move? If yes, the framework applies. If no, we treat the move as being driven by something else and go looking for it.