The reported story: yields rose Wednesday because the economy ran hot. S&P Global’s flash September readings beat forecasts on both manufacturing and services, and the ten-year Treasury yield closed at a nineteen-year high.
The auction record: indirect bidders took 54.3% of the five-year sale, the smallest share since March 2020. The bid-to-cover ratio was the weakest since December 2018. The sale priced 3.1 basis points above where it was trading.
Almost every account of Wednesday’s selloff led with growth. Strong purchasing-manager surveys, a stubborn economy, higher yields, a tidy causal chain.
Then the Treasury tried to sell $70 billion of five-year notes, and the tidy chain fell apart.
The sale cleared at 5.033%, the highest yield on a five-year note since June 2006. That is the number that made the headlines.
The number that did not is 2.212. That was the bid-to-cover ratio, meaning $2.21 of bids arrived for every dollar on offer, the thinnest showing since December 2018.
A hot economy does not do that. A hot economy produces buyers who want a 5% risk-free coupon.
To see what actually happened, you have to look at who was left holding the paper.
There are roughly two dozen primary dealers. They are the banks and broker-dealers that the New York Fed designates as counterparties, and in exchange for that designation they carry an obligation almost nobody outside the plumbing thinks about: they are expected to bid at every single Treasury auction.
They are the buyer of last resort, by arrangement. When the rest of the world declines, they absorb the difference onto their own balance sheets.
On Wednesday they absorbed 15.8% of the offering.
Run the arithmetic on a $70 billion sale and that is roughly $11.1 billion of five-year notes that the market did not want, parked at firms that did not want them either and took them because the rules say they must.
That is the story. Not the PMI.
The tail makes it plainer still. An auction “tails” when it prices at a higher yield than the pre-auction trading level, which is the market’s way of saying the Treasury had to pay up to get the deal done.
Wednesday’s tail was 3.1 basis points, among the largest ever recorded at this maturity.
Three basis points sounds like a rounding error. On $70 billion, it is real money, and it repeats every four weeks for the life of the fiscal path.
Compare it with August, when the same tenor cleared at 4.393%. Sixty-four basis points of deterioration in a single month, on the note that anchors corporate refinancing and auto lending.
Brookings keeps a checklist for telling apart a government paying more because growth is strong from one paying more because buyers are thinning out. Wednesday hit nearly every marker in the second column.
The distinction matters for one reason.
A growth story is self-correcting. The Federal Reserve tightens, activity slows, yields come back down. That mechanism has worked for forty years, and it is the implicit assumption behind last week’s hike to 4%.
A demand story is not self-correcting. If foreign central banks and institutional buyers are stepping back from US paper, a slowing economy does not bring them back. It gives them one more reason to stay away, because a slowdown widens the deficit that made them nervous to begin with.
That is the fight Stanley Druckenmiller picked with the Treasury in August, and Wednesday’s indirect bid handed him the better of the argument.
Here is where BusinessTech lands, and it is not where the wires landed.
Calling Wednesday a growth story is the more comfortable of two available readings, and the auction record does not support it.
The Treasury’s debt-management office has spent two years extending issuance into a market that has been signalling, sale by sale, that it would rather not be asked. Indirect participation at a five-and-a-half-year low is a price signal, and it is being reported as a headline about purchasing managers.
The consequences are already priced elsewhere. Mortgage rates cleared 7% in the daily trackers before the weekly surveys admitted it, and SoftBank raised $20 billion in a book that was, as we wrote this week, a bet on yield rather than on OpenAI.
What should happen next is specific: shorten the issuance calendar, stop leaning on a dealer community absorbing $11 billion a sale to keep the optics clean, and say out loud that the marginal buyer has changed. The record-setting auction data is public.
Equity desks are still trading this as an inflation scare, with the S&P 500 off 0.8% and stagflation the fashionable comparison. Strong data got the billing, and the nineteen-year high in the ten-year got the chyron.
The dealers holding $11 billion of unwanted five-year paper know which story they are in.