Treasury yields: two different numbers wearing one price

Decompose a ten-year yield and you get two parts. The first is the average overnight rate the market expects over the next ten years - essentially a forecast of monetary policy stretched across a decade. The second is the term premium: the extra compensation investors demand for locking money up for that long instead of rolling short-term paper. The two are added together and quoted as one number, which is why the number so often behaves in ways the policy story cannot explain.

The proportions matter. With the ten-year near 4.7 per cent and the term premium around 0.80 per cent in mid-August 2026, roughly five sixths of the yield is a policy expectation and one sixth is compensation for duration risk. That ratio is not fixed. The term premium spent years near zero and at times below it, and its return to clearly positive territory is one of the more consequential shifts in this market - because it means the long end can rise while the policy outlook is unchanged.

The practical reading is by tenor. The front of the curve is almost entirely policy expectation; the middle carries inflation expectations, visible as the breakeven between nominal and inflation-linked debt; and the long end is where the term premium lives. A market on a two-year yield and a market on a thirty-year yield are asking about different things, even though both are called interest rate markets.

The ten-year is the reference rate for long-duration cash flows across the economy. Mortgage rates track it, corporate borrowing is priced as a spread over it, and equity valuation models discount future earnings at rates derived from it. That is the mechanism behind the otherwise odd observation that good economic news can lower share prices: strong data raises the expected policy path, raises the yield, and raises the rate at which distant earnings are discounted.

It is also the government's own cost of borrowing, which makes it politically reflexive. Higher yields raise interest expense, which widens the deficit, which increases issuance, which - through the supply channel - pushes yields higher again. That loop is slow and rarely dominant, but it is the reason the composition of issuance between bills and coupons has become a market-moving decision in its own right.

For markets, the tradeable objects are thresholds and spreads. Contracts on whether the ten-year ends a period above a level, whether the curve inverts or steepens, and whether a specific policy path is realised are all in the inventory. They resolve against published daily series, which puts this subject among the most cleanly settleable in the whole library.

Treasury publishes the daily par yield curve, which is the official source for what a given tenor yielded on a given day. That is the series a well-written contract names, and it has an important property: it is a par curve constructed from market quotes at a fixed time, not a last-trade print, so a contract written against it is insulated from a spike in the final minutes of a session.

The St. Louis Fed's database carries the constant-maturity series and the term premium estimate, which are the two things you need to decompose the number rather than just observe it. The term premium is a model estimate rather than an observation - there is no instrument that trades it directly - and different models produce different levels while broadly agreeing on direction. A contract on a term premium level would be a contract on a model, which is a reason none exist.

The supply side is announced in advance. The quarterly refunding announcement sets out how much of each tenor Treasury intends to sell, and the split between short bills and long coupons changes how much duration the market must absorb. Those announcements are scheduled and have moved the ten-year by five to fifteen basis points, which makes them one of the few reliably tradeable calendar events in this subject.