The Auction Costs Most at the Front End. It Costs Again at 20 Years.
When the US government sells a bond it does not get the screen price. It gets whatever the buyers in the room will pay at that moment, and the gap between the two is the concession — the small discount an auction has to offer to clear size in one afternoon. At the long end traders call a bad version of this a tail: the auction went badly enough that the stop, the highest yield accepted, printed above where the bond had been trading. A concession is the clearest read there is on whether the marginal buyer showed up.
This piece measures that gap for every coupon auction of 2026 — the notes and bonds that carry duration, as opposed to bills, which mature inside a year. The result is not a slope. It is a U with a spike at the very front and a second bump at 20 years. The 2yr pays about 16 times what the 10yr pays. The 10yr clears almost exactly where it had been trading. Then the 20yr — a tenor nobody has a liability in — charges again. That shape is the argument: the concession is compensation for warehousing risk, and it shows up wherever the buyer is a dealer rather than an owner.
It matters this week because the vote is 2 days away and the Treasury announces its next round of coupon sizes on 4 November. If the deficit is going to be funded at duration, that announcement is the only place the election can become a supply number — and the concession has to reappear at 10 and 30 years first. Here is what the auction record says, and the level at which it stops being true.
The ladder is a U, and the second bump is the interesting one
Measured as the auction's high yield against the par curve on the same date: the 2yr at +3.54bp over 9 auctions, the 3yr +2.56 over 10, the 5yr +1.88, the 7yr +0.20, the 10yr +0.22, the 20yr +1.51 and the 30yr +0.48. The fall from 2 to 10 years is 3.3bp. The rise from 10 to 20 is 1.3bp, on 9 auctions.
The 2yr cleared above its close in all 9 auctions and the 5yr in all 9. The 7yr did so in only 5 of 9, and its mean of +0.20bp sits below its median of +0.30 because the misses were larger than the hits. That asymmetry is the signature of a tenor met by buyers who are not price-sensitive, interrupted by the odd auction where they stay home.
Two honest caveats. This is the stop against the 3pm par close, so it carries an afternoon of market drift as well as the concession; across 66 auctions the drift should largely cancel, but it is not a clean when-issued tail and free data does not carry when-issued levels. And the latest usable auction in these averages is 2026-10-08: the fiscal feed still reports null results for the 26 to 29 October coupon auctions, which TreasuryDirect publishes as the 2yr at 4.708%, the 5yr at 4.845% and the 7yr at 4.894%. Those 3 prints sit outside the averages rather than inside them.
Two different crowds, and the 20yr has the worst of both
Indirect bidders — the catch-all for foreign official accounts, domestic funds bidding through a dealer, and pensions — take 65.5% of the 10yr and 61.0% of the 30yr. At the 2yr the same group takes 51.9%, at the 3yr 55.3%, at the 5yr 56.2%. The 10yr and the 2yr are bought by different crowds, and the gap between them is 13.6 percentage points.
Primary dealers take the mirror image: 7.6% of the 10yr and 8.4% of the 30yr, against 10.7% of the 2yr and 11.6% of the 3yr. A dealer is the residual buyer, the firm that must take what is left when the bid is done. High dealer share means the auction cleared through inventory; low dealer share with high indirect share means it cleared through end buyers who had already decided to own it.
The 20yr is the case that proves it. It pays +1.51bp — more than the 30yr, more than the 10yr — on 8 of 9 auctions clearing above the close, with a dealer share of 10.8% against 7.6% at the 10yr. Nobody has a 20yr liability. A pension matching 30yr promises buys 30yr paper; a bank hedging a short rate buys the front. The 20yr has to be cleared by whoever is willing to warehouse it, and it charges for that. The two tenors where dealers take least, 7.6% at the 10yr and 8.4% at the 30yr, are two of the three that cleared within half a basis point of the close.
Bills did the funding, so duration was never asked
Over the 90 days to 2 November the Treasury settled $7,932bn of gross auction supply. Bills were $6,960bn, or 87.8%, and coupons $972bn, or 12.2%. In the prior 90 days coupons were $1,232bn on almost identical gross supply of $7,936bn: coupon supply fell $260.8bn while total supply was flat, and the mix moved +3.3pp toward bills.
Monthly coupon settlements have not moved much all year — 363bn to 411bn a month through September — and the 4-week bill has been the shock absorber. That is why the 10yr has almost no concession to pay. Duration supply is a policy choice made at the refunding, and the choice so far has been to keep it small and fund the marginal dollar at the front.
But aggregate scarcity does not explain the 20yr. Duration supply is small in total and the 20yr still charges +1.51bp, which says that bump is not about how much the Treasury issues — it is about who is willing to own that specific maturity. The bill share is also a constraint rather than a gift: the 4-week is refunded every 4 weeks, so a rising bill share is a rising refinancing wall. It buys the long end time and charges the front end a concession for it.
Term premium is expensive for 2026 and ordinary for 1990
The ACM 10yr term premium — the extra yield investors demand for holding a long bond instead of rolling short ones, estimated by the New York Fed and distinct from the market's inflation expectations — closed at 0.9358% on 2026-10-23. Against 2026 it is the 86th percentile of 204 prints, inside a range of 0.4627% to 1.0965%. Against the whole series back to 1990 it is the 60th percentile of 9189.
Both readings are true and they say different things. Against this year the long end is expensive: term premium has roughly doubled from its 0.4627% low. Against 3 decades it is unremarkable. A fiscal risk premium that only looks expensive against a base of the last 10 months is not yet a fiscal risk premium — it is the normalisation of a very compressed starting point.
The vote reaches this number through exactly one pipe: the coupon sizes announced at the refunding. Nothing else about the election touches term premium directly. A sweep, a split Congress, a contested count — none of them change how much duration the Treasury sells next quarter. The refunding does.
Wednesday is the only channel the vote has
TreasuryDirect's calendar has the next refunding announcement on 4 November, with the 3yr auctioning 9 November, the 10yr 10 November and the 30yr 12 November, all settling 16 November. The announcement lands 2 days after the vote and sets coupon sizes for the quarter. That is the first moment a political outcome can become a supply number.
The test is mechanical. If the 4 November statement raises 10yr and 30yr sizes, the long end has to absorb more duration, the dealer becomes the residual buyer again, and the concession should reappear at the 10 November and 12 November auctions. If sizes are unchanged and the bill share stays near 88%, the U holds: cheap at the front, cheap at the 20yr, almost free at the 10yr.
The curve as it stands on the 29 October par close: 2s10s at 37.0bp, 20s30s at +3.0bp with the 20yr above the 30yr, and the 3-month bill to the 20yr at 117.0bp. A curve 37bp steep from 2 to 10 and then inverted at 20 is not pricing a growth shock or an inflation shock. It is pricing a buyer who exists at 10 and is scarcer at 20.
Where this is wrong
If the 4 November refunding lifts 10yr and 30yr sizes, and the 10 November 10yr then clears more than 2bp above its same-day close with primary dealers above 12%, the concession has moved out to duration and this read is dead. The 2026 mean dealer share at the 10yr is 7.6%, so 12% is a clean break rather than a wiggle.
The second falsifier is term premium above 1.0965%, the 2026 high. That would say the fiscal path is already in the long end and the auction ladder is measuring the wrong thing. A third, quieter one: if the 20yr stops clearing above its close for three consecutive auctions while the 2yr keeps paying, the U was a dealer-inventory artefact rather than a buyer-structure fact.
The Vote Cannot Move the Breakeven. It Moves the Real Yield. — where the political risk premium has actually been paid this year
The Corridor Reopened 7.1 Basis Points. The Front End Did All of the Work. — Monday's curve read, which found the front end doing the repricing