2Y 4.800 ▼5Y 4.900 ▼10Y 5.100 ▼30Y 5.460 =VIX 25.46 ▲MOVE 101.88 ▲DXY 103.60 ▲USD/JPY 159.49 ▼GOLD 3,852 ▼COPPER 6.01 ▲WTI 85.84 ▲BTC 64,856 ▼ETH 1,819 ▼S&P 7,341 ▼NDX 27,002 ▼MSTR 96.50 ▼
Rates · Liquidity · Scarcity · Credit · Power · 2026-11-17

The 20yr Costs 5 Times as Much to Sell as the 30yr. The Treasury Sells It 40% Smaller.

Read the plain-English version

An auction tail is the difference between the yield the Treasury had to pay and the yield the market printed in the minutes before the sale. It is a toll: the compensation a buyer demands for absorbing a specific bond at a specific hour, rather than buying it in the market whenever it suits. A tail of 2 basis points on a $40bn sale is small change in yield terms and real money in dollars, and it is the cleanest read available on who is unwilling to own a maturity. This piece asks a narrow question. Strip the auction results of 2026 by tenor, divide each tail by the dollars actually raised, and the answer to which maturity is hardest to place is not the longest one. It is the 20yr, by a factor of 5 against the 30yr, in a year when the Treasury has funded almost everything with bills and left duration supply flat. So the toll is not being set by the deficit. It is being set by the balance sheet of the firm that has to warehouse the bond when nobody else bids, and by a size small enough to hide it.

Per dollar raised, the 20yr is the most expensive point on the curve

There are 66 coupon auctions with a completed result in the 2026 file, 12 January to 8 October, and averaging tails across them is misleading because a $70bn 2yr and a $13bn 20yr are not the same transaction. Divide the tail by the dollars offered instead and the ranking changes shape. The 2yr pays 0.051 basis points of tail per $1bn raised, the 3yr 0.044, the 5yr 0.027, and then the two maturities that institutions treat as interchangeable, the 7yr and the 10yr, cost almost nothing: 0.005 and 0.006. The 30yr is 0.021. The 20yr is 0.108, which is 5.1 times the 30yr and 20 times the 10yr.

auction tail in basis points per $1bn offered, 2026 coupon sales66 coupon auctions, 12 Jan to 8 Oct 2026; amber marks the 20yr000.10.10.10.0512yrn90.0443yrn100.0275yrn90.0057yrn90.00610yrn100.10820yrn90.02130yrn10tail = auction high yield minus same-day par close; divided by dollars offered, not per auction

Read the plain averages and the same fact appears with less force: the 20yr's mean tail is +1.51 basis points against +0.48 at the 30yr, a gap of 1.03. That gap looks like noise until it is divided by size, and size is the point. The Treasury sold $14bn of 20yr paper per auction on average and $23bn of 30yr. Per dollar, the market charges 5 times as much to take the smaller, shorter bond. Two instruments a decade apart in maturity, and the one 10 years shorter is the harder sale.

Note what this is not. It is not a claim that the 20yr auctioned badly. The mean tail is positive but under 2 basis points, and only 1 of the 9 sales stopped through the pre-auction market (22 July, -0.7). It is a claim about the price of a marginal buyer's balance sheet, and about what happens when the Treasury sells a maturity in a size too small for the dealers to lay off efficiently and too large for the end buyers to absorb in one print.

The scarcity is on the buy side, and the dealer is where it lands

Auction results split the accepted bids into 3 books. Indirect bidders are the ones who submit through a dealer, which is where foreign official money and most real-money duration sits. Direct bidders are domestic institutions bidding for themselves. Primary dealers are the firms obliged to bid, and what they take is the residual nobody else wanted. Across the 2026 sales, indirect bidders took 64.1% of the 20yr's private book and 68.6% of the 30yr's; primary dealers took 11.6% of the 20yr and 9.6% of the 30yr. Two points of dealer share is not a scandal on its own. What makes it matter is the combination: the 20yr is the smallest print on the coupon calendar, it carries the most expensive tail per dollar, and the dealer is taking a larger share of it than of the 30yr the Treasury sells at twice the size.

The 15 September sale is the clean illustration. It was $13bn, it tailed +2.0 basis points, indirect bidders took 52.1% and primary dealers took 16.7%. That is the lowest indirect share and the highest dealer share of any 20yr sale this year, and it is the only one where the dealers ended up with more than a sixth of the paper. Compare the 30yr on 8 October: +1.8 basis points of tail, 72.1% indirect, 6.8% dealers, on a $22bn print. Twice the size, less than half the dealer share, and a tail 0.2 basis points tighter.

The mechanism is a maturity mismatch inside the dealer community. A 20yr bond has the interest-rate risk of a long bond and the liquidity of neither end of the curve. It is not a benchmark, so it does not finance cheaply in repo. It is not a 30yr, so it does not satisfy the pension funds and insurers whose liabilities are long enough to want it. When the Treasury sells $13bn of it, the dealers underwrite the print and then have to find a home for a bond that no client's mandate specifically asks for. The tail is the fee for that search, and per dollar it is the highest fee on the curve.

The inversion has held for 2 months, which is why it is a price and not an accident

The 20yr has yielded more than the 30yr on 47 of the 51 sessions since 1 September, against 44 of 167 sessions before it, and the current run is unbroken: 45 consecutive sessions from 09/10/2026. The spread has not closed at or below zero since that date. On the 13 November par close it is 5 basis points, the 217th highest of 218 sessions against a median of 0 and a full-year range of -7 to 6. Only one session in the file has been wider.

20yr minus 30yr par spread, basis points, 218 sessions of 202613 Nov close 5bp; 217th of 218 against a 0bp median-8-6-4-20246JanMarMayJulSepNov5above the line: the 20yr yields more than the 30yr. It crossed in 09/10/2026 and has not gone back.

A 5 basis point inversion is not a market accident, because an accident would not survive 50 sessions. Nor is it a spectacular one: the widest print of the year is 6 basis points, and a 20yr costing 5 more than a 30yr is well inside the range that a pure expectations model would tolerate. What makes it useful is its stability. A spread that sits at the 99.5th percentile of its own year, for 2 months, while the average long-end term premium is falling, is telling you about structure rather than about a single auction failing.

The level chart shows the same thing from the other side. The 20yr closed at 5.52% on 13 November, the 203rd highest of 218 sessions, while the 30yr closed at 5.47%, the 204th. Both are near the top of the year's range; the 20yr is nearer. The curve's peak has moved inside 30 years and stayed there.

20yr and 30yr par yields, 202613 Nov: 20yr 5.52% against 30yr 5.47%; the 20yr is the higher yield4.64.85.05.25.45.620yr30yrJanMarMayJulSepNovpar closes; the crossover is inside the 20yr-30yr spread chart above, seen as a level

Bills carried the funding, so the long end was never asked

The supply story cannot explain a 20yr-specific toll, and the issuance data says why. Over the 90 days to 16 November the Treasury settled $5,742bn of bills and $815bn of coupons, so bills were 87.6% of $6,557bn gross. In the prior 90 days coupons were $1,234bn on a 84.7% bill share. Coupon settlement is down $419bn, or 34%, while the bill share rose +2.9 points.

gross coupon settlement against gross bill settlement, 90-day windowsbills 87.6% of $6,557bn in the last 90 days6,829bills1,234couponsprior 90d5,742bills815couponslast 90d02,0004,0006,0008,000$bn of gross auction settlement; windows 2026-08-18 to 2026-11-16 and 2026-05-20 to 2026-08-17

This is the political channel, and it runs the opposite way to the obvious one. A government that cannot pass new spending does not need to sell new duration, so the refunding keeps 3yr, 10yr and 30yr sizes flat for a fourth consecutive quarter and the funding need is met at the bill line. Duration supply does not rise, so the term premium on duration does not rise. The ACM 10yr term premium, the New York Fed's model estimate of the extra yield a buyer demands for holding a 10yr bond rather than rolling short paper, fell 21.1 basis points to 0.8738% between 6 October and 6 November (the series lags, so that is the latest print, 10 days stale on this piece).

So the average long-end compensation is falling while the 20yr's specific premium over its neighbours is the widest all year. Those two facts only reconcile one way. What is being priced at the 20yr is not the size of the government's borrowing. It is the cost of placing a bond the government insists on selling in a size the dealer community cannot distribute.

One measurement that cuts against this desk's own story

The obvious test is whether the spread tracks the auction result. Plot the 20yr tail against the 20s30s spread on each of the 9 sales and the correlation is -0.22: nothing. The 19 August sale tailed +3.4 basis points with the spread then at -2; the 15 September sale tailed +2.0 with the spread at +4. If a bad auction were driving the inversion, the two would move together. They do not, on 9 observations, which is far too few to settle anything but enough to refuse the easy version of the argument.

20yr auction tail against the 20s30s spread on the day, 9 sales of 2026correlation -0.22; the curve spread does not track the auction result-6-3+0+302-1808-1909-15tail, basis points20s30s, bpamber: 15 Sep, the only sale where dealers took more than 16% of the private book

What that leaves is a spread set by the standing conditions of the 20yr rather than by any single print: the size, the repo treatment, the client base, and the dealer's own capital cost. Last week the spread widened 2 basis points, from 3 on 6 November to 5 on 13 November, but the 20yr butterfly - the 20yr's yield against the average of the 10yr and the 30yr, the same maturity priced against both neighbours instead of one - went the other way, from 46 to 42 basis points, against a 52 basis point median. Neither measure moved the way the bad-auction story requires. The 20yr is not being progressively penalised. It is being penalised at a roughly constant rate, which is a different and much more durable claim.

The October sale exists in the feed and has no price in it

One data caveat has to sit on the record before the call. The Treasury's Fiscal Data auctions feed carries a 19-Year 10-Month row dated 21 October with every result field null: offering, accepted, bid-to-cover, high yield, bidder splits. The row exists and the price does not. So the 20yr statistics in this piece are 9 sales through 15 September and not 10, and the October auction, the most recent one, is absent from every average quoted above. If that sale was the widest of the year, the per-dollar figure for the 20yr is understated; if it was the tightest, the ratio is flattered. The 217th-of-218 spread rank and the 5 basis point level are unaffected, because those come from the par curve rather than the auction feed.

The same feed gap is why the 20yr cadence matters for what comes next. The 2026 sales fell on 21 January, 18 February, 17 March, 22 April, 20 May, 16 June, 22 July, 19 August and 15 September, which is a third-week-of-the-month rhythm on a Tuesday or Wednesday, and the next one in that cadence lands in the third week of November. That is the nearest test of everything in this piece, and it is 2 days away.

The mechanism, the instrument, the falsifier

The mechanism, stated plainly. Gridlock means no new duration supply: coupon settlement fell $419bn in 3 months and bills carry 87.6% of gross funding. No new duration supply means no new supply pressure on term premium, which is why the 10yr term premium fell 21.1 basis points in the month to 6 November. But the Treasury still sells $14bn of 20yr paper 9 or 10 times a year, into a client base that does not specifically want it, through dealers who must warehouse it. The price of that sale does not rise with the deficit. It rises with the dealers' balance-sheet cost and falls with the size of the print. That is why the toll per dollar is highest exactly where the print is smallest.

The instrument. The 20yr-minus-30yr par spread, at 5 basis points on the 13 November close, and as the tradable expression, the 20yr tail per $1bn raised at 0.108 basis points against 0.021 at the 30yr. A reader who wants one number: 5 basis points, 217th of 218 sessions.

The falsifier. This read is wrong if the 20s30s spread closes below zero for 3 consecutive sessions. A 20yr that yields less than a 30yr again means the market has stopped charging the maturity a separate price, and the whole construction collapses to a rounding error. The second, weaker falsifier: a 20yr auction that tails 0.5 basis points or less while the spread is still 4 basis points or wider would say the toll is a standing feature of the maturity's liquidity rather than of any single sale, which is a smaller claim than the one made above.

Where this is wrong

First, 9 observations. Every auction statistic here rests on 9 completed 20yr sales and 10 30yr sales in one calendar year, and the correlation between tail and spread is -0.22 on that sample, which is indistinguishable from no relationship. The per-dollar ranking is arithmetic, not inference, but the story attached to it is inference and it is thin.

Second, the repo point is asserted rather than measured. The claim that a 20yr finances worse than a 30yr in the repo market is central to the mechanism and this desk has not pulled a general-collateral series for either maturity. If a reader has that data, it would settle or kill the section on dealer balance sheets faster than anything in this piece.

Third, the 20yr is not the only maturity with a stale print, and the whole long end is carrying a term premium that fell 21 basis points in a month for reasons this piece treats as background. If the next refunding raises 20yr size rather than holding it flat, the size argument inverts: a bigger print should lower the toll per dollar if the mechanism is distribution cost, and raise it if the mechanism is simply that nobody wants the maturity. Those are the two outcomes to watch on the next sale, and they point in opposite directions.

CALL & INVALIDATION
Levels2Y 4.88 / 3Y 4.92 / 5Y 4.98 / 7Y 5.05 / 10Y 5.15 / 20Y 5.52 / 30Y 5.47 / 2s10s 27bp
CallThe 20yr costs 0.108 basis points of auction tail per $1bn raised, 5.1 times the 30yr's 0.021 and 20 times the 10yr's 0.006, across 66 coupon sales in 2026, and it is the highest yield on the curve at 5.52% against 5.47%, a 5 basis point inversion that ranks 217th of 218 sessions. The mechanism is a closed fiscal channel: coupons are down $419bn over 3 months, bills are 87.6% of gross funding, and the 10yr term premium fell 21.1 basis points to 0.8738% in the month to 2026-11-06. No new duration supply means the 20yr's premium is not a supply price; it is the dealer's price for warehousing $14bn of a bond no client mandate asks for, and that is why it is highest where the print is smallest.
InvalidationWrong if 20s30s closes below zero for 3 consecutive sessions, which undoes the premise that the market charges this maturity a separate price. Second test: a 20yr auction tailing 0.5bp or less while the spread holds 4bp or wider would demote the claim from structural to liquidity-driven.
RELATED
The Curve Peaks at 20 Years, and the Dealer Is the Buyer There — the same maturity, measured on the concession rather than the auction tail
The Auction Costs Most at the Front End. It Costs Again at 20 Years. — the U-shaped concession ladder this piece re-tests per dollar raised
The Long End Got Cheaper While 12 Months Got Dearer. That Split Is the Political Map. — why a closed fiscal channel does not add duration supply
DISCLOSURE — NOT INVESTMENT ADVICE
NoticeMacroTape publishes macroeconomic analysis and market commentary for information and education only. Nothing published here is financial, investment, legal or tax advice, or a personal recommendation, and no author is a registered investment adviser. Calls, Levels and invalidation points are analytical views about how a market mechanism works, not instructions to trade. Readers act on their own responsibility and carry their own losses.