The Rates That Could Not Agree

The government's own average cost of borrowing is now two stories at once. The average rate paid across all interest-bearing debt stands at 3.447, up 0.1 from 3.352 a year earlier. Notes moved the same way, to 3.309 from 3.073, a rise of 0.24, and bonds to 3.442 from 3.309, a rise of 0.13. Bills went the other way, down to 3.758 from 4.314.

Four series, one ledger, and the signs do not match. The easy account is that debt service is still climbing and the bill line is the exception. That account requires the largest single move in the file, the fall in bills to 3.758 from 4.314, to be noise. Don's position spends exactly that: if the split is real, his confirmation is one instrument wide, and the widest move in the stack is the one he has set aside.

Franc's position spends differently. A changed relationship between the bill rate and the coupon rates would explain the whole table, but the claim must name the relationship, with a before and an after: bills still carry the highest average in the stack, yet their cost falls while every other piece rises. What the data cannot settle is whether the split is composition, the mix of the debt shifting beneath the averages, or condition, the pieces themselves pricing apart.

The Treasury's month-end average rates, next landing at September's close, test both sides of the split at once. If bills and coupons resolve in the same direction, the split was arithmetic, Don's concession is refunded, and Franc owes back the changed relationship he never named. If they split again, composition stops being available: Don concedes his exception was the story, and Franc's claim survives only if he finally says which relationship broke, and when.