2026-Q3 revision: the energy shock becomes a regime
Quarter reviewed: 2026-Q3What happened
The US–Iran ceasefire expired on 17 August. In mid-September three simultaneous disruptions (the East–West pipeline, a vessel struck in the Strait of Hormuz, the blockade of Bab el-Mandeb) pushed Brent to about $107 on 13 September; at quarter-end it remains close to $100, up 29% over the quarter. Inflation turned up again: euro area 3.8% in September with energy at +18.8%; United States 3.4% in August with energy at +16.3%. The ECB (10 September, deposit rate 2.50%) and the Fed (16 September, 3.75–4.00%, the first hike in three years) raised rates; the 10-year Treasury ended the quarter at 5.29%, up 88 basis points. On the technology side, Nvidia reported quarterly revenue of $96.2 billion (+106% y/y) with guidance of $108 billion; the S&P 500 gained 2.0% and the Russell 2000 lost 7.5% over the quarter: a concentrated rally. China shows a stimulus-supported stabilisation (manufacturing PMI 50.1 in September after two months of contraction, Q2 GDP +4.3%). COFER 2026Q2: the dollar at 56.7% (from 57.2%), the euro at 20.6%. The ECB raised its 2027 euro-area growth projection to 1.4%.
What we missed
On 12 September we left S3 at 20%, treating the energy shock as “already happened”: its transformation into a regime (two central banks hiking, the 10-year above 5%) was already readable in the July and August data. We underestimated the fiscal and rates channel: +88 basis points on the 10-year in one quarter is a signal of sub-case S5B (fiscal stress), which the model does not measure with a column of its own. The three registered questions remain open; the criticism received on their calibration (priors too cautious) is well founded, but by rule priors are not changed after registration: it will remain visible in the Brier score.
Decision on the weights
S1 from 35% to 30%; S3 from 20% to 25%; S2, S4 and S5 unchanged. The duration and transmission of the shock (double-digit energy inflation, Fed and ECB hikes, higher real yields) meet the persistence criterion the model requires for S3; the five points are taken from S1 because selective cooperation is the most compromised condition. S2 stays at 25%: AI capex and revenues confirm capacity, but diffused productivity is not yet verified and higher rates weigh on valuations. S4 at 10%: China’s stabilisation still depends on stimulus. S5 at 10%: the yield signal is recorded but not yet accompanied by credit stress; separating S5A and S5B with their own weight is the first model change planned for the Q4 revision.
To verify next quarter
Brent below or above $90 for two consecutive quarters and the state of the Hormuz talks; the October WEO (question 1, deadline 31/10); US Q3 productivity (BLS, November); AI capex versus revenues in the October–November earnings season; China’s Q3 GDP and consumption; the term premium and the 10-year above 5%; credit spreads and the ECB’s November Financial Stability Review.