The US Treasury sold $58 billion of 3-year notes at 17:00 UTC yesterday at a high yield of 4.474%, against a when-issued level of 4.475%. It stopped through by a tenth of a basis point. Bid-to-cover was 2.72 against a 2.62 average, direct bidders took 26.9% against a 20.4% average, indirects 64.2% against 65.5%. By every demand metric the auction was fine. The previous 3-year auction cleared at 4.291%. Demand was normal — at eighteen basis points more money. That is the number to carry into today.
Read the auction, not the headline
Auction commentary usually collapses into “strong” or “weak”, and both words describe the wrong thing. A Treasury auction answers two separate questions, and they are worth separating.
- Did buyers show up? Yes. A 0.1bp stop-through means the auction cleared slightly better than where the market was quoting it beforehand — buyers paid up marginally rather than demanding a concession. Bid-to-cover above its average and directs at 26.9% against a 20.4% average say real money, not just dealers, was in the book. Dealers took 10.9% against a 14.1% average, which is what you want: the dealer take-down is the residual, and a small residual means the auction did not need rescuing.
- At what price? 4.474%. The previous one cleared at 4.291%. Same instrument, same issuer, same three-year horizon, eighteen basis points apart.
Those two answers are not in tension and people keep treating them as if they are. The market cleared the paper cleanly because the price had already moved to where it needed to be. That is a functioning auction and a repriced curve at the same time.
Eighteen basis points at the front end is the actual event
The 10-year yield topped 4.8% on Tuesday for the first time since October 2023 on Trading Economics' series, easing back to around 4.78–4.79%. BabyPips marks it at roughly 4.80% into the close. The 3-year clearing at 4.474% is the same story told at a maturity where Federal Reserve policy dominates and term premium does not.
What changed to justify it? Two things, both dated inside the last five sessions.
- August payrolls at +162,000 against a consensus near 56,000, which moved September Fed hike odds from roughly a coin flip to about two-thirds.
- Brent at $99.00 and WTI at $94.41, after Houthi strikes on Saudi energy facilities halted operations at some sites. Brent's high of $99.22 is its best since 24 July; WTI's $94.60 its best since 8 June.
Note what those two have in common: both push the same direction on the Fed, and Friday's August CPI at 12:30 UTC is the release that has to absorb them. The energy component of an August print will not contain a September oil spike — that arrives in the September data, published in October. So Friday is a clean read on where inflation was, delivered into a market that has already repriced for where it is going. A market that has repriced ahead of the data is a market with room to be disappointed in both directions.
The dollar did not get paid for any of it
This is the part that should bother anyone running a rate-differential signal.
Front-end US yields rose. Fed hike odds are near two-thirds. Oil, which is normally dollar-supportive through the terms-of-trade channel for a net exporter, spiked. And USD/JPY still closed lower on the day, printed a seven-month low at 152.89 in Tokyo hours, and failed to reclaim 155 — a level it broke and has not recovered.
Yes, the dollar bounced off the low. It bounced roughly a hundred pips inside a 650-pip four-day slide. If widening differentials in the dollar's favour buy you a hundred pips of retracement, differentials are not what is setting the price. Positioning is. The speculative yen short stood at 92,227 contracts on 4 September and the market has spent the week grinding it out.
The practical consequence: a carry or differential-based signal on the yen crosses is currently being paid by something it is not measuring. That is not an argument to switch it off. It is an argument to stop treating its recent P&L as evidence that its logic is working, because it isn't — you are on the right side of a trade for the wrong reason, which is the least durable kind of edge there is. The Friday 11 September COT release is the direct test. If the net short has collapsed from 92,227, the fuel is spent.
What is actually scheduled, in UTC
The auction cycle is not finished, and it is the part of this week most systems do not have in their event filter.
- Today 01:30 — China August CPI and PPI. Consensus 0.9% y/y on CPI from 0.5% prior, and 3.6% on PPI from 3.5%, on Investing.com's calendar. A CPI rebound from a six-month low, if it lands, is an AUD and NZD input more than a CNY one.
- Today 17:00 — US 10-year note reopening. Same slot as yesterday's 3-year.
- Thursday 10 September 12:15 — ECB decision, 2.25% to 2.50% expected.
- Thursday 12:30 — US August PPI, consensus +0.4% m/m, and weekly jobless claims, consensus 208,000.
- Thursday 12:45 — Lagarde press conference and staff projections.
- Thursday 17:00 — US 30-year bond reopening, the last leg of the refunding.
- Friday 11 September 12:30 — US August CPI.
- Wednesday 16 September — FOMC. Friday 18 September — Bank of Japan.
Thursday remains the day that stacks three tier-one events into thirty minutes, and we have already argued it should be configured as a single exclusion window rather than three. Today, by contrast, has one dated item that matters for the dollar and it is at 17:00 — well after London has done its business.
What this does not tell you
An eighteen basis point gap between two auctions is not eighteen basis points of repricing in a day. We do not have the settlement date of the prior 3-year auction in front of us, so we cannot tell you over how many weeks that gap opened. Treat it as the size of the move between two comparable clearing prices, not as a velocity.
We are also not claiming the auction was good because yields are high. Both statements are true at once and the causal arrow is not something an auction result can settle. A cleanly cleared auction at a higher yield is consistent with confident buyers and with buyers who simply required more compensation; the metrics do not distinguish them.
We do not know whether the oil move persists. Aramco had not commented and damage assessment was ongoing at the time of the reports we read. If the supply impairment turns out to be minor, the inflation input behind part of this week's yield move goes away, and it goes away faster than it arrived.
And we have no forecast for Friday's CPI. The Cleveland Fed nowcast circulating this week — roughly 3.38% headline, 2.38% core — comes to us through a single carrier and we have not been able to read it at source. We have not built anything on it and neither should you.
Related
- FX event calendar — the full week in UTC, including the 17:00 auction slots.
- Signals
- Expert Advisors
