How a Treasury auction is priced, and what a tail actually tells you

Twice a month the Treasury sells long-dated debt, the result prints at around 13:01 New York time, and within ninety seconds a market that was quiet reprices by several basis points. Reading that result correctly is a small skill with a large payoff, and most of the commentary written about it each month misreads at least one of the five published numbers. This is what each of them measures, in the order they matter.

The when issued market sets the expectation

Nothing about an auction can be understood without the when issued market. From the moment the size and maturity of an auction are announced, usually several days beforehand, dealers trade the security that does not yet exist. That forward market establishes a yield at which the bond is expected to clear. By the time bidding closes, the when issued yield represents the market's collective estimate, refined over days, of where the auction will stop.

The auction itself is a single price affair. Bidders submit the yield at which they are willing to buy and the quantity they want. The Treasury fills the lowest yields first and works upward until the issue is sold. The yield at which the last bid is filled is the stop, and every successful bidder pays that same stop regardless of what they bid.

Why bid to cover is the weakest statistic

Bid to cover is the total value of bids submitted divided by the amount sold, and it is quoted first in almost every account of an auction because it is the easiest number to describe. It is also the least informative. A large share of submitted bids are placed at yields far above where anyone expects the auction to stop, entered to secure an allocation if the result goes badly and never intended to be filled. Those bids inflate the ratio without representing demand at any price the seller would accept.

The ratio also moves mechanically with auction size. Halve the size of an issue and the same nominal demand produces double the cover. Comparing this month's cover to last month's is only meaningful if the sizes match, and through a period of rising issuance they rarely do.

The allocation split is where the information sits

Three categories share every auction. Indirect bidders are accounts bidding through a dealer, historically dominated by foreign central banks and large asset managers. Direct bidders are domestic institutions submitting on their own behalf. Primary dealers take whatever is left, and that last clause is the important one.

Primary dealers are obliged to bid in every auction as a condition of their status. They are not expressing a view when they do it. Whatever they absorb becomes inventory, and inventory has to be financed in the repo market and sold into the secondary market over the following days. A high dealer share therefore predicts selling pressure in the sessions after the auction, which is why the statistic matters more than its position at the end of the release would suggest.

That series is the single best summary of what has happened at the long end of the curve this year. Dealer absorption at the thirty-year point has roughly doubled since October. Each of those auctions cleared, none of them failed, and by the headline measure most looked acceptable. Underneath, a larger share each month was taken by bidders who had no choice, which is why long yields have drifted higher in the sessions following auctions with unusual consistency.

An auction never fails. It simply clears at a yield somebody did not want to see, and then the dealers who were obliged to take it spend a week distributing it.

Ines Marchetti, head of rates strategy at Harrow Lane Capital

Reading a result in sequence

1. 01Note the when issued yield in the minutes before the close of bidding, because everything that follows is measured against it.
2. 02Take the stop and compute the tail. A tail above one and a half basis points on a ten-year, or two on a thirty-year, is a real result.
3. 03Read the dealer share next. A tail with a low dealer share is a pricing accident. A tail with a high dealer share is a demand problem.
4. 04Check the indirect share against its own six month average rather than against the previous auction, which is noisy.
5. 05Ignore bid to cover unless the auction size is unchanged from the prior one, in which case it adds a little at the margin.
6. 06Watch the secondary market for two sessions afterwards, because dealer distribution shows up as a drift in yield rather than a single move.

One structural question makes the coming months different from the past decade. Policymakers meet on Wednesday and are expected to end balance sheet runoff, and the composition of what replaces it is the detail that matters here. Reinvestment spread proportionally across the curve would return roughly $38bn a month of coupon demand to a market absorbing elevated net issuance. Reinvestment into bills alone would return none of it. The dealer share series above was rising before that meeting was scheduled, and the long-dated auctions that follow it are where either answer will show first.

None of this predicts the level of yields. Auctions are a clearing mechanism, not a forecast, and a long run of soft results has preceded rallies often enough to make anyone cautious about extrapolation. What the results do reliably tell you is who is holding the paper a week later, and in a market where the marginal buyer has changed twice in three years, that is the more actionable question.