Reference 08 September 2026 · 9 min

A plain-English guide to when higher yields stop supporting the dollar.

The yield-differential model (higher UST yields pull dollar higher) works most of the time and breaks in three specific regimes. Regime 1: the yield move is term-premium rather than real-rate, so foreign capital does not rebalance into the dollar; signature is long-end steeper than front-end and TIPS breakevens widening. Regime 2: foreign policy repricing narrows the forward differential even when spot US yields rise; signature is a sharp move in the single pair with the policy news, other DXY components lagging. Regime 3: a crossflow shock (oil, terms of trade, reserves policy) changes the balance-of-payments arithmetic for the paired currency. The three regimes usually run together; separating them is the diagnostic skill. Tuesday September 8 delivered roughly 40/55/5 across the three.

By Igor Komolov

Tuesday's analysis piece is the live example. The 10-year US Treasury yield closed at 4.806 percent, a new thread high. DXY closed at 98.92, below the 99 line for the first time in two weeks. USD/JPY dropped 251 pips to a seven-month low at 153.53. Under the textbook yield-differential model these three moves do not fit together: higher US yields should support the dollar, and a stronger dollar should support USD/JPY. That did not happen. This reference piece is the framework for reading when the standard rate-differential mapping breaks, why, and what the tape looks like in each of the three breakdown regimes.

The standard model

The textbook says: higher US yields, relative to the rest of the developed-market curve, attract marginal capital into dollar-denominated assets. The rebalancing flow raises the dollar. The specific number the framework uses is the two-year yield differential: the 2Y US yield minus the 2Y JP yield (for USD/JPY), the 2Y US yield minus the 2Y DE yield (for EUR/USD). See our yield-differentials framework for the full mechanic.

The standard model works most of the time. Over the trailing five years, the correlation between the 2Y US-JP differential and USD/JPY sits around 0.75; the correlation between the 2Y US-DE differential and EUR/USD sits around 0.65. When the standard model breaks, it breaks for one of three specific reasons, and the diagnosis matters because each reason has a different tape signature and each maps to a different trade.

Regime 1: the yield move is term-premium, not real-rate

Nominal yields decompose into expected real short rates plus a term premium. See the term-premium framework for the ACM decomposition. Foreign capital chasing yield rebalances on the real-rate component; it does not rebalance on the term-premium component (the term premium compensates for duration risk, and a foreign buyer already bears that risk plus FX risk).

So when nominal yields rise but the move is fully in term premium, the yield-differential model over-predicts the dollar move. The specific footprint on the tape is (a) long-end steeper than the front-end (30Y up more than 2Y), (b) TIPS breakevens rising alongside nominal yields (real yields flat), (c) dollar failing to extend against the DM basket, and (d) gold typically holding rather than selling off with the nominal yield.

Tuesday September 8, 2026 delivered the Regime 1 signature partially. The 30Y moved +3.8bp, the 2Y moved flat, the long-end steepened, and the Jazan supply shock produced the inflation news that widened the term-premium band. The Regime 1 read on Tuesday: roughly 40 percent of the day's rate move was term-premium, and the dollar's failure to extend against the DM basket is proportionate to that decomposition.

Regime 2: foreign policy repricing narrows the forward differential

The yield-differential model uses the current spot differential, but FX prices the forward differential (spot plus the expected path over the next six to twelve months). When a foreign central bank prices in hawkish news that narrows the forward differential, the FX pair can move even when the current US yield is rising. The relevant currency-pair examples are USD/JPY and BoJ pricing, EUR/USD and ECB pricing, USD/CAD and BoC pricing.

The specific footprint is a divergence between spot yields and forward-implied rate paths. When BoJ hike bets for the next twelve months move from 25 basis points priced to 50 basis points priced, the forward 2Y US-JP differential compresses by roughly 15-20bp, and USD/JPY can drop 150-250 pips even if the US 2Y is flat or up. The tape signature is a sharp move in the single pair while the other DXY components lag.

Tuesday September 8 delivered a clean Regime 2 signature on USD/JPY. Takata's follow-through commentary pushed OIS-implied BoJ hike odds for the September 17-18 MPM from roughly 30 percent Friday close to 45-50 percent Tuesday close; the forward differential moved 12-15bp; USD/JPY dropped 251 pips. That is 15-20 pips per basis point of forward-differential move, which sits inside the framework's typical Regime 2 range of 12-25 pips per basis point.

Regime 3: a crossflow shock dominates the pair

Some events change the balance-of-payments story enough to move the FX pair independent of the rate differential. Middle East supply shocks on oil are the archetypal example: a $3-5 Brent move driven by a supply disruption changes the current-account arithmetic for oil-importing DM economies (Japan, the euro area, the UK) and can move their FX pairs in the opposite direction of the standard yield-differential prediction. Trade-policy shocks, terms-of-trade shifts, and central bank FX-reserves policy changes are the other main sources.

The signature is (a) the crossflow catalyst is dated and identifiable, (b) the move in the pair is proportionate to the shock (not to the rate move), and (c) other pairs without the crossflow exposure do not move in tandem. If the Regime 3 signature is genuine, EUR/USD does not have to follow USD/JPY, and cable does not have to follow either.

Tuesday September 8 had a partial Regime 3 layer: the Jazan attack pushed Brent to $99.38 (up $3.10 on the session), and Japan is a large net oil importer whose current-account position deteriorates on higher oil. But the yen strengthened rather than weakened, so oil is not the Regime 3 driver here. The Regime 3 dimension is instead the BoJ policy repricing itself (which is technically Regime 2), and the Jazan shock is showing up in the rates (Regime 1 term-premium) rather than in the FX crossflow.

The three regimes usually run together, and separating them is the skill

In practice a given tape delivers a mix of the three regimes rather than a clean single-regime signature. The skill is diagnosing the mix. Tuesday September 8 delivered roughly 40 percent Regime 1 (term-premium yield move), 55 percent Regime 2 (BoJ forward-differential repricing on the yen leg), and 5 percent Regime 3 (Jazan supply shock feeding into oil-inflation term premium rather than into FX current-account flows). The Regime 2 dominance on USD/JPY is what produced the 251-pip move; the Regime 1 dominance on the aggregate US-DM differential is what produced the muted DXY move (only 24 pips lower on the session).

A trader who reads the tape as "the yield-differential model broke" misses the signal. The signal is that the yield-differential model is telling the truth about the current yield differential, but the FX is pricing the forward and pricing the composition of the yield move. Both are correct. The tape looks confused; it is not.

The four checks worth running before concluding the standard model has broken

Decompose the yield move. Read the 2Y-30Y curve shape and the TIPS breakeven move on the same session. If the long-end is up more than the front-end and breakevens are widening, you are in Regime 1 and the FX under-response is expected, not anomalous.
Check the OIS-implied policy path for the foreign central bank. If the pair that broke correlation is USD/JPY, pull the BoJ OIS path for the next three MPMs. If it is EUR/USD, pull the ECB OIS path for the next three ECB meetings. A visible shift in the foreign OIS is the Regime 2 diagnostic.
Cross-check the other pairs in the DXY basket. If USD/JPY broke correlation but EUR/USD and GBP/USD did not, the mechanism is pair-specific (Regime 2 or 3). If all three broke correlation together, the mechanism is dollar-general (Regime 1, plus positioning).
Read the term-premium proxy against the FX response. The single best diagnostic for Regime 1 is the ACM ten-year term premium: if it moved more than 3bp on the session while the two-year real rate moved less than 1bp, Regime 1 is the driver.

What this framework says about the September 15-16 FOMC and the September 17-18 BoJ pair

The setup coming out of Tuesday has all three regimes stacked in the same direction. If the FOMC delivers the hawkish-hike on September 15-16 and the BoJ delivers a hike on September 17-18, the forward differential compresses further and USD/JPY has room to run toward 150. If the FOMC hikes and the BoJ holds, the pair reverses violently toward 158-160 as the Regime 2 pricing unwinds. The two-day FOMC-then-BoJ window is where the current three-regime stack gets resolved; the resolution shape is doubly asymmetric because the market has priced both meetings in the same direction and a break in either half changes the setup materially.