The dollar, the last refuge from an oil shock

The reversal occurs between 9 and 11 September. On 9 September, the 10-year auction clears at 4.834%, its highest since August 2007; on 10 September, the ECB raises rates; on 11 September, August CPI exceeds consensus at 3.4%. From 9 September to 7 October, the dollar index moves from 98.61 to 102.41 and EUR/USD from 1.1652 to 1.1177. Over the same period, the US 10-year yield gains 45bp, peaking at 5.31% on 5 October, its highest in two decades, while the 2-year gains 34bp: the 2–10-year spread rises from +40 to +51bp.

The 10-year breakeven stays between 2.31% and 2.36%; the entire increase comes through the real yield, which reaches 2.93%, its highest since the snapshot's history began in late 2021. Our reading is that investors demand more real yield to hold US duration, with a federal deficit of $2,000bn over the first eleven months of fiscal 2026 according to the CBO. A stable breakeven suggests this premium compensates for securities supply, rather than questioning the Fed's inflation credibility.

This real yield is offered by an economy relatively protected from oil. The Chicago Fed estimates that the oil shock subtracts 81 to 166bp from US growth in 2026, while domestic production and energy's smaller share of spending cushion the blow. Franklin Templeton notes that this is the first time since 1980 an oil supply shock has coincided with restrictive monetary policy and a strong dollar. This regime's risk lies in Treasury quarterly refunding in early November: if long-dated supply increases, the term premium could become a fiscal risk premium and weigh on the dollar.

The FOMC unanimously raised rates by 25bp to 3.75–4.00% on 16 September, the first increase since 2023. Median projections put fed funds at 4.1% at the end of both 2026 and 2027, against 3.8% and 3.6% in June, implying one additional increase this year. PCE is expected at 3.7% in 2026 and core at 3.4%. Kevin Warsh called the decision “serious” and “responsible” without promising anything further, consistent with his Jackson Hole speech judging financial conditions insufficiently restrictive.

On prices, the month's data supports the committee. August core PCE stands at 3.0% year on year, while manufacturing ISM prices paid jumps to 77.9 in September, against 72.3 expected and 71.1 in August. Tariffs are cited by 34% of respondents and the Iran war by 30%. Growth holds: second-quarter GDP was revised to +2.2% annualised, August nominal consumption rose 0.9% and ISM remained at 54.5.

The labour market, however, slows. Average job creation over twelve months falls to +45,000 per month, while unemployment at 4.2% remains within its 4.1–4.3% range since March. Average hourly earnings growth slows to +3.0% year on year from 3.7% in February, below headline CPI of 3.4%: real wages fall. Our reading is that US inflation comes from imported prices and tariffs, and the Fed tightens against a supply shock, limiting the scale of tightening. Projections of 4.1% for 2027 nevertheless rule out cuts over this note's horizon: the Fed limits the dollar's potential decline without being what drives it higher.

The 14 June ceasefire brought Brent spot back to $69.70 in early July. Its expiry without extension on 17 August renewed hostilities: an Iranian attack on the Saudi tanker Sidr, the first US strikes on Iranian tankers on 2 September, then IRGC missiles against US Navy ships the following weekend. Brent rose from $89.73 on 28 August to $124.15 on 18 September, ending at $120.03 on 2 October. UKMTO still recorded 12 incidents between 28 September and 5 October.

The crude spread widened. Brent exceeded WTI by around $5 in late August; the gap reached $23.5 on 25 September and $22 on 2 October, beyond April's maximum ($20.6). With oil freight at record levels in September according to the EIA, seaborne delivered oil diverges from oil produced and consumed in America, where commercial crude stocks of 427.3 million barrels stand 2% above their five-year average.

Supply will provide no near-term relief. The IEA estimates 2026 global supply falling by 4.3 million barrels per day and demand by 1.56 million barrels per day, with an implied deficit of 1.27 million barrels per day drawing on stocks (report vintage to be confirmed). The EIA raised its fourth-quarter Brent forecast by $14 to $105, with 4.5 million barrels per day of constrained Middle Eastern production. On 4 October, OPEC+ rolled September targets forward to November. Demand destruction or a reopening of Hormuz would be needed to lower prices.

The bill falls on seaborne importers: the euro area, Japan, Korea, whose won is at a 17-year low according to the press, India, whose rupee reached 96.30 on 5 October, and Indonesia, whose rupiah trades below its level in the 1998 crisis. The IMF, whose adverse scenario estimated a 0.6 percentage-point reduction in global growth with oil averaging $110 in 2026, will publish new forecasts in Bangkok between 12 and 18 October. Spot prices already exceed that level.

The ECB raised its deposit rate to 2.50% on 10 September, its second increase in 2026. September flash HICP caught up three weeks later: 3.8% year on year against 3.2% in August, with energy at +18.8% and core at 2.5%. The deposit rate therefore remains 1.3 percentage points below headline inflation and level with core. Activity does not call for caution: September flash composite PMI reached 53.1, its highest in around three and a half years, while manufacturing was at a 55-month high.

An increase on 29 October was around 60% priced on 24 September, before HICP; we have no later measure. Our reading is that another 25bp increase would not move the euro: it would leave the short-rate differential with the United States around 200bp, while the long-end gap widens. In monthly averages, the 10-year Bund rose 28bp in September to 3.47%, and the US 10-year 54bp to 5.29%.

Winter and France add two euro-specific risks. TTF exceeded €80/MWh in September for the first time in three years. European storage was only around 71% full in late September, against a five-year average of 84% in mid-September, with reduced Qatari LNG deliveries; Christine Lagarde herself cited gas risk in a cold winter. The 2027 draft budget targets a deficit of 5.0% of GDP after 5.4% in 2026, with debt close to 119%. Every ECB increase makes funding more expensive for a state whose Bund spread is at its highest since 2011–2012. The ECB therefore cannot match US real yields without increasing that cost.

Outside the Fed, four G10 central banks raised rates over the period, and all four currencies fell against the dollar: the euro by 3.69%, the yen by 2.58%, the Australian dollar by 3.76% and the New Zealand dollar by 4.34%, the G10's worst performance. The Bank of England, holding at 3.75%, is the exception: sterling falls only 2.50%, and EUR/GBP declines 1.22%.

The RBA raised its rate to 4.60% on 29 September, its highest since 2011, but AUD fell the next day despite August CPI at 4.0%. A trimmed mean holding at 3.6% and the end of fuel-excise relief explain part of that figure; Michele Bullock said the Board considered a hold, rather than a 50bp increase. Traders read this as a peak approaching. The RBNZ, back up to 2.75%, itself signals an October pause. A hold on 28 October is around 97% priced, leaving NZD with no increase to expect before December.

At the BoE, three members voted for an increase to 4.00% in September (secondary source, to be confirmed). The bank expects CPI around 3.75% in the fourth quarter, then slightly above 4% in the first quarter of 2027. With August core CPI at 2.6%, Bank Rate offers a real rate of around 1.15 percentage points, while the ECB's is zero on core inflation.

Our reading is that, in a supply shock, a reactive increase presented as the last does not support the currency: traders see a central bank suffering imported stagflation. A currency holds when its real rate is positive and the door remains open to further increases. That is true of the Fed, with projections of 4.1%, partly of the BoE, but not of the other four.

Joint Japanese–US intervention on 31 July took USD/JPY from 163.9 to 156.7 by 3 August. The pair returned to 160.16 on 1 September. The 3.6-yen drop on 3 September, which the Ministry of Finance did not confirm as intervention, opened a decline to 153.27 on 9 September; USD/JPY has since returned to 158.28. Confirmed interventions of ¥11,700bn in April–May and ¥15,399bn in July–August did not change the trend.

The BoJ raised its rate to 1.25% on 18 September, a 31-year high, with two dissenting votes weighing on the yen. August CPI excluding fresh food slowed to 1.7%, below the 1.8% consensus, and remained below 2% for an eighth month because of fuel subsidies. Bets on an October increase approximately halved in late September; the prevailing scenario places the next one in December. Oxford Economics and Citi believe yen weakness could bring it forward.

The 10-year JGB gained 11bp in September's monthly average to 3.05%, while the Treasury gained 54bp, and Japan pays seaborne prices for its oil. Fiscal policy adds to this: Sanae Takaichi asked on 5 October that investors be reassured, while the 30-year JGB is at a record and Washington urges Tokyo to reduce spending to stabilise the yen. US pressure targeting the budget suggests the Treasury would prefer adjustment through Japanese fiscal policy to another joint intervention.

Beyond 160, the level on 1 September, two risks accumulate before late October: another intervention and a BoJ bringing its increase forward (meeting date to be confirmed). Below that threshold, the long-rate differential continues working against the yen.

The eighth round of US–China consultations, from 20 to 23 September, produced reciprocal tariff reductions on around $30bn of goods on each side and extended the truce until 10 January 2027, ahead of Xi Jinping's state visit on 24 September. The PBoC prepared the ground with eight consecutive stronger fixings and offshore yuan at 6.6967 on 18 September, its strongest since July 2022. The yuan reached that high as the dollar index began recovering, resilience no G10 currency showed.

China's foreign trade gives room for this appreciation: August exports grew 25% year on year, driven by AI-related demand, for a surplus of $119.09bn and $806bn over eight months. Official manufacturing PMI returned to 50.1 in September, with input prices at 60.8 and output prices at 54.0, their highest in five months. Our reading is that China stops exporting disinflation, removing a cushion for importers of Chinese goods. Third-quarter Chinese GDP on 19 October will show whether growth supports this currency choice.

The Canadian dollar followed the opposite path. Section 338 tariffs of 50% on Canadian cars, alcohol and dairy have applied since 19 August. Ottawa retaliated on 8 September on C$27.6bn of US imports, with steel and aluminium taxed at 50%. USD/CAD rose from 1.3768 on 9 September to 1.4262 during a month in which Brent gained 34%: Canada sells its crude at the WTI price.

Canadian inflation does not push the BoC to follow the Fed. August CPI stays at 3.0% because of petrol, but CPI-trim is 1.9% and CPI-median 2.0%, with unemployment at 6.4%. The BoC, at 2.25% for almost a year, has little reason to raise rates on 28 October; the gap with a Fed at 3.75–4.00% exceeds 150bp. Employment on 9 October could confirm this reading; only trade de-escalation with Washington would change the loonie's regime.