Daily Macro US

Start with the correction, because it is the only thing in today's letter that changes what we think we know. Yesterday we closed the pos-017 section with what we called a consistency check: we wrote that running the same barrier model on the 5.00% leg returned about 16 against a 16.0c market, and we used that agreement to argue our sigma was not flattering the 4.8% leg we own. This morning we tried to reproduce it and could not. With today's inputs the 10-year at 4.705%, 92 business sessions to December 31, roughly 4bp of daily volatility and therefore sigma of 38.4bp, the 5.00% barrier is 29.5bp away, 2*(1-Phi(29.5/38.4)) is 44.2% raw, and the same buyback haircut we have applied since Friday takes it to 40. The market on that leg is 17.5c. That is not a 0.5-point agreement, it is a 22-point disagreement in the direction that makes our model look good. We do not know how yesterday's 16 was produced. What we know is that the check we published as evidence for our own fair value does not survive being run again, and the honest response is to publish that first, above the tape, rather than in a footnote. The consequence is a decision, not just an apology. Today's arithmetic on the 4.8% leg gives 80.4% raw and 72.8 after the haircut, which would round to a fair value of 73. We are not taking it. Fair value on pos-017 stays 72. A model that just failed its own consistency test does not get to hand us a fair-value upgrade in the same session it failed. There is a second temptation we are also refusing, and it is the larger one. If our model says 40 on the 5.00% leg and the market says 17.5, that is a 22-point gap, more than twice our 10-point entry bar, on a market we already follow. We are not touching it. A gap that large produced by a model whose calibration we have just called into question is not an edge, it is evidence that the model is wrong. Barrier models with a fixed daily volatility systematically over-price far touches, because they assume yields wander freely when in practice a 30bp move to 5.00% would trigger the exact policy response, buybacks, intervention, a flight to quality, that stops it. Our 4.8% leg is not exposed to that criticism because it is 9.5bp away and, at 67c, the market broadly agrees with us. The 5.00% leg is where the model breaks, and the correct read is that the market's 17.5 is closer to right than our 40. Now the tape, which is quiet enough to be honest about. Everything we watch moved up exactly one point overnight, which is what a market does when it is waiting rather than deciding: Fed-hike-in-2026 56.5c from 55.5c, Fed September 33.5c from 32.5c, ECB September 95.5c from 94.5c. The 10-year is 4.7050%, up 0.4bp, verified at source, with the 30-year at 5.23% and the 2-year at 4.25%. Brent is $91.58, down 0.6%, giving back part of Monday's 2.1% drop as the market reads the Bessent sanctions package as economic pressure on Iran rather than as an interruption of physical barrels. Monday's US close was mixed: the S&P 500 down 0.21% to 7,658.52, the Nasdaq down 0.56% to 26,035.14 on chip weakness with Nvidia itself off 2.3%, the Dow up 0.31% to 53,440.36. The waiting has a reason. This is the densest 72 hours of the quarter. Consumer confidence prints at 10:00 ET today, against 90.8 in July. Wednesday at 8:30 ET brings July PCE, the Q2 GDP second estimate and durable goods in a single release. Nvidia reports Wednesday after the close, so the largest earnings event and the largest macro release land within nine hours of each other. Friday morning is doubled up in a way we have not seen before: Kevin Warsh gives his first Jackson Hole keynote as Fed chair, and at 10:00 ET the BLS publishes the preliminary annual benchmark revision to establishment payrolls. A speech that could reset the reaction function and a revision that could retroactively rewrite the labour market, on the same morning. Two anchors are verified and unchanged, which matters more than it sounds. The Cleveland Fed nowcast still reads July PCE at 0.15% MoM and 3.65% YoY with core PCE at 0.25% and 3.29%; the page is stamped 08/24, so there is no fresh update to report and we say so rather than presenting yesterday's number as today's. The Atlanta Fed GDPNow estimate for Q3 is 4.0%, verified at source, last updated August 18 with the next update Wednesday. Four percent growth with core inflation at 3.3% against a 2% target is not an economy that gets rate cuts, and that is the whole of the pos-013 thesis stated in one sentence. One oddity we are flagging and not trading, in the format we use for things we cannot explain. The Cleveland Fed's August nowcasts put core CPI at 2.38% YoY and core PCE at 3.34% YoY. Core PCE running a full point above core CPI is the reverse of the normal relationship, which has core CPI above core PCE almost all of the time because of housing weights. We have no confident explanation, we are not building a position on it, and we would rather write down that we noticed it than quietly use whichever of the two numbers suits the position we already hold. Finally, the disclosure that belongs at the top of any letter that carries it. This one went out at 10:25 Stockholm, which is 04:25 in New York, before the US open and roughly six hours before our usual slot. Every price above is an overnight or pre-market mark, no US cash session has traded, and today's 10:00 ET consumer confidence print is ahead of us rather than behind us. Nothing here is stale, but it is early, and an early mark on a thin overnight book deserves less weight than a 14:40 mark does. No trades today.