A plain-English guide to the term premium.

US 10Y 4.46%

Any time a strategist tells you the 10-year Treasury yield "rallied on growth concerns" or "sold off on supply," they are making a claim about which piece of the yield moved. The 10-year is two things added together: the market's expectation for where the short rate will average over the next decade, and an extra compensation for taking on the risk of holding a long bond instead of rolling short paper. That second piece is the term premium. This is the explainer on what it is, the standard way it is measured, and why the current cycle's move from deeply negative to firmly positive matters more than the headline yield level.

The decomposition, in one line

Write the 10-year nominal yield as y. Decompose:

y = E[average short rate over 10 years] + TP

The first term is the policy expectation: where the market thinks overnight rates will average over the life of the bond. You can read it directly off the OIS curve, which we covered in a separate reference piece. The second term, TP, is everything else. It compensates investors for inflation uncertainty, real-rate uncertainty, supply risk, and the duration-versus-cash trade-off of locking up money for ten years.

The decomposition is not a fact; it is the output of a model. A raw 10-year yield does not come with a name tag saying "60 basis points of me is term premium." You build the term premium by estimating the expected-rate path with a model and treating the rest as TP. Different models produce slightly different splits. The arithmetic of the splitting is mechanical.

The ACM model is the desk standard

The most-cited measurement comes from the Federal Reserve Bank of New York, built by Tobias Adrian, Richard Crump, and Emanuel Moench. The model is referred to as ACM after their initials. It is a five-factor, no-arbitrage affine term-structure model fit to the off-the-run Treasury curve. It produces a daily, history-matched series of expected-rate paths and term premia for every maturity from one to ten years.

Three things to remember when you read an ACM print:

1. It is a model, not a measurement. The decomposition depends on the assumption that the model class is correctly specified. The NY Fed publishes uncertainty bands; most secondary sources strip them off.
2. The expected-rate piece is anchored by survey forecasts. ACM uses the affine model's own predictions, but cross-checks against the Blue Chip and SPF forecasts. When the affine model and the survey diverge, TP picks up the residual.
3. The cycle range is wide. Over the post-1990 sample the ACM 10-year term premium has run roughly from -100 basis points to +400. It was deeply negative through most of 2019-2023, with the trough around -75 basis points. The current regime is at the other end of the table.

The picture you want in your head

Worked example with today's numbers

The 10-year Treasury yield is currently around 4.46 percent. Suppose the OIS path through the next decade implies an average short rate of roughly 3.5 percent (a reasonable read off the current curve, with cuts in the front and a steady-state neutral around 3 percent in the back). The implied term premium is the residual:

TP ≈ 0.96 percentage points, or about 96 basis points.

That is comfortably in positive territory and consistent with the ACM print. The number you derive yourself this way can drift from the published ACM print by 10 to 30 basis points depending on which short-rate path you assume; the level is roughly right, the direction is robust, and the regime call is unambiguous.

What drives the term premium

Three factors do most of the work historically:

• Net supply. Treasury issuance plus QT minus foreign demand. The market clears at whatever term premium makes investors willing to absorb the float. The 2024-2026 rebuild in TP is mostly here: deficits widened, the Fed kept running off its balance sheet, and foreign Treasury holdings stopped growing in line with issuance.
• Inflation uncertainty. A wider distribution of future inflation outcomes is a real risk to a long bond holder. When inflation prints are predictable, TP compresses; when they are volatile, TP widens.
• Duration-vs-cash preferences. When the flight-to-safety bid for duration is strong, TP falls; when investors prefer cash, TP rises. The 2020-2022 episode is the cleanest example of the duration bid: TP fell to around -75 basis points as central-bank QE and risk-off flows piled into the long end.

How term premium connects to FX and equities

A higher term premium at constant policy expectations means higher long-end real yields, which usually supports the dollar through the rate-differential channel. The recent exception is what we've been writing about in the FX series. Long-end yields have risen while DXY has drifted lower, because the dollar has been driven by relative-growth differentials rather than by the rate side. See the dollar piece and the curve piece for the live application.

Equities care about TP through the discount-rate channel. The standard rule of thumb is that a 50-basis-point rise in TP, with expected rates unchanged, corresponds to roughly a 4 to 6 percent repricing of the S&P. The recent regime change in TP is one of the structural reasons multi-asset desks are running shorter duration on both legs of their book.

What TP does not tell you

1. It is not a directional forecast. A high TP does not mean rates will fall, nor that they will rise. It means investors are demanding more compensation today for holding ten-year duration than they were a year ago.
2. It is path-dependent. TP can stay elevated for years without obvious cause; it can compress in weeks if the cyclical narrative changes. The 2020 compression to -75 bp happened in roughly six months.
3. Model choice matters at the edges. ACM is the most-cited, but the Kim-Wright model (used internally at the Federal Reserve Board) and the d'Amico-Kim-Wei adjustment produce slightly different prints. The level differs by tens of basis points; the regime call almost never differs.

Related reading

• Why the curve un-inverted without a recession, the analysis piece this framework feeds into
• A plain-English guide to OIS, the expected-rate side of the decomposition
• The correlation matrix, where TP and DXY can be read against each other
• Today's analysis: the 7-bp 10Y rally on the Conference Board print is mostly TP compression