The term premium is the extra yield investors demand for lending at a fixed rate for many years instead of lending for a short period and reinvesting each time the money comes back. When a 10-year Treasury yields more than the average of the short-term rates investors expect over the next decade, that excess is the term premium. The New York Fed defines it as compensation for bearing the risk that interest rates change over the life of the bond, and its researchers have found that it tends to rise with uncertainty and disagreement about where yields are heading. No security pays it on its own and no screen quotes it, so every figure you see is a model estimate. All the numbers below are for US Treasuries, the securities covered on the US Treasuries hub.
How is the term premium calculated?
Take a bond's yield and subtract the average short-term rate investors expect over its life; the remainder is the term premium.
Only the first input can be looked up; expected short rates are not traded on their own. If investors asked nothing extra for committing their money, long yields would equal the expected average of short rates, the textbook expectations hypothesis, and the premium would be zero. The Federal Reserve Board's research staff note that a large body of evidence shows the hypothesis fails, which is why models are needed to separate the two pieces. Those models work with zero-coupon yields, the return on a single payment due on one date, so their yields differ slightly from the coupon-bond par yields on the Treasury yield curve page.
The same split works one year at a time. A forward rate is the rate an investor can lock in today for a period that starts later, and the Board defines the forward term premium as that forward rate minus the short rate expected for the same period.
Worked example: Locking in two years or rolling short bills, October 29, 2026
The New York Fed's ACM model gave these zero-coupon figures for October 29, 2026:
- Fitted yields: 4.440% for 1 year and 4.614% for 2 years
- Expected average short-term rate: 4.272% over the first year and 4.367% over both years
Working it through:
- 2-year term premium: 4.614 − 4.367 ≈ 0.25 point a year, the expected extra over rolling very short bills for both years
- Rate locked in today for the second year, the forward rate: about 2 × 4.614 − 4.440 = 4.788%
- Short rate expected during the second year: about 2 × 4.367 − 4.272 = 4.462%
- Forward term premium for the second year: 4.788 − 4.462 = 0.326 point, against 0.17 for the first year; the 2-year premium is roughly the average of the two
The forward rate is also a break-even. Someone who buys a 1-year bill now and plans to buy another next year comes out behind the 2-year note unless next year's 1-year rate tops 4.79%.
How does the New York Fed's ACM model estimate it?
It fits a five-factor statistical model to the whole Treasury curve, infers the expected path of short rates from it and counts the rest of each yield as term premium.
The model is named after Tobias Adrian, Richard Crump and Emanuel Moench, current and former New York Fed economists, who released it as a staff report in 2008 and published the final version in the Journal of Financial Economics in October 2013. It belongs to the affine, no-arbitrage family of term structure models, which keep yields at every maturity consistent with one another. Its five factors are principal components of yields, statistical summaries of how the whole curve moves, and it is estimated with a three-step linear regression that the authors describe as computationally fast. The inputs are zero-coupon yields from the Gürkaynak, Sack and Wright curve, which the Federal Reserve Board publishes daily.
The New York Fed posts the results as an Excel file: for each maturity from 1 to 10 years, a fitted yield, the expected average short rate and the term premium, daily from June 14, 1961 and as month-end values. It re-estimates the model once a month, which it warns can make minor changes to past figures, so a number copied today may differ slightly from the same date in an older download. The figures on this page come from the file with data through October 29, 2026.
| Maturity | ACM fitted yield | Expected average short rate | Term premium, points |
|---|---|---|---|
| 1 year | 4.44% | 4.27% | 0.17 |
| 2 years | 4.61% | 4.37% | 0.25 |
| 3 years | 4.67% | 4.39% | 0.28 |
| 5 years | 4.75% | 4.38% | 0.37 |
| 7 years | 4.83% | 4.34% | 0.49 |
| 10 years | 4.95% | 4.29% | 0.66 |
The middle column barely moves, staying between 4.27% and 4.39%, while the premium nearly quadruples from 1 year to 10. Of the 0.51-point rise in fitted yields between those maturities, 0.49 came from the premium. Rounded parts may not add exactly to the yield.
What is the term premium now?
The latest New York Fed estimate, for October 29, 2026, put the 10-year term premium at 0.66 percentage point, up from 0.50 a year earlier.
Worked example: Splitting the 10-year yield a year apart
| ACM, 10-year maturity | October 29, 2025 | October 29, 2026 | Change |
|---|---|---|---|
| Fitted zero-coupon yield | 4.118% | 4.952% | +0.834 |
| Expected average short rate | 3.617% | 4.287% | +0.670 |
| Term premium | 0.501 | 0.664 | +0.163 |
- Check: 4.287 + 0.664 = 4.951, matching the 4.952% yield to rounding.
- About four-fifths of the year's 0.83-point rise came from a higher expected path of short rates and about a fifth from the premium.
- On $10,000, a 0.664-point premium is worth about $66 a year, or roughly $660 over ten years before compounding: the extra the market pays for the 10-year commitment rather than rolling short Treasuries.
Within 2026 the 10-year estimate has ranged from 0.46 point on June 29 to 0.98 on October 7. The longer record puts those readings in context.
| ACM 10-year term premium | Points |
|---|---|
| Average, June 1961 to 1979 | 1.07 |
| Average, 1980s | 3.41 |
| Average, 1990s | 2.12 |
| Average, 2000s | 1.49 |
| Average, 2010s | 0.47 |
| Average, 2020 to October 29, 2026 | −0.21 |
| Highest day, January 11, 1982 | 5.28 |
| Lowest day, March 9, 2020 | −1.66 |
The premium was below zero on most trading days from 2016 through 2024, years in which investors took a 10-year yield below the short-term rates they expected to average. The last negative reading came on October 16, 2024. Why it climbed after that, and what moved it in particular weeks, is the subject of what moves Treasury yields.
Why do term premium estimates differ?
Each model needs its own assumptions to pin down expectations nobody can observe, so identical market yields can produce different premiums.
The Federal Reserve Board's staff publish a second estimate based on the Kim-Wright model. It uses three unobserved factors instead of five and adds survey forecasts of the 3-month bill rate to help keep its expected path of short rates realistic. Its definition also folds in a small negative convexity effect, so the Board notes its level runs mechanically slightly lower than a pure premium. The Board calls the series a staff research product rather than an official statistical release and usually posts it weekly, on Tuesday for data through the previous Friday.
| 10-year term premium, points | ACM (New York Fed) | Kim-Wright (Fed Board) |
|---|---|---|
| March 9, 2020 | −1.66 | −0.61 |
| October 29, 2025 | 0.50 | 0.48 |
| October 23, 2026 | 0.67 | 0.94 |
The two estimates were 0.02 point apart a year ago, 0.27 apart in late October 2026 and more than a full point apart in March 2020. Their direction agrees more often than their level: between month-ends since January 2000, the two 10-year series moved the same way in 81% of months, by our calculation from both files. The ACM authors made the same point in 2014, writing that the estimates co-move over long periods but can differ markedly at times.
Where do you see the term premium?
Mostly in Federal Reserve data and meeting minutes, and in market commentary explaining moves in long yields that Fed rate decisions do not account for.
- Fed research data. The New York Fed's Treasury Term Premia page, with its ACM Excel file, and the Fed Board's three-factor model page, with the Kim-Wright CSV.
- FOMC minutes. The minutes of the September 15-16, 2026 meeting recorded market commentary linking higher term premiums and Treasury yields to geopolitical developments, uncertainty over Treasury's buyback program and heavy private borrowing to build artificial intelligence infrastructure.
- Explanations of the yield curve. The inverted yield curve explained shows how a negative premium deepened the 2022 to 2024 inversion, and short-term vs long-term Treasuries weighs the extra yield against the price risk. In valuation work, a long Treasury yield used as the risk-free rate carries a term premium on top of expected short rates.
How is the term premium different from the term spread?
The term spread is a quoted gap between two yields, while the term premium is an estimated slice of one yield, so the two can even have opposite signs.
| Measure | What it is | Observable? | October 29, 2026 |
|---|---|---|---|
| Term premium, 10-year (ACM) | Yield minus the average short rate expected over 10 years | No, a model estimate | 0.66 point |
| Term spread, 10-year minus 3-month | Gap between two quoted par yields | Yes | 0.80 point (4.98% − 4.18%) |
| Expected average short rate, 10-year (ACM) | The path of short rates investors anticipate | No, a model estimate | 4.29% |
| Credit spread | Extra yield a company or other borrower pays over a Treasury of similar maturity | Yes | Measured against Treasuries, so not applicable |
| Yield to maturity | Annual return on a bond held to maturity at today's price | Yes, from the price | Contains both the expected rate and the premium |
On October 29 the spread and the premium happened to be close, because expected short rates sat only slightly above the 3-month bill yield. That is not the rule. On July 8, 2016, the 10-year par yield was 1.37% and the 3-month 0.28%, a positive spread of 1.09 points, while the ACM premium was −0.72. The expected-rate part follows what investors think the Fed will do with the federal funds rate, the same force that sets bill yields, as how Fed rate decisions affect T-bill rates explains.
What do people get wrong about the term premium?
The common errors are reading the curve's slope as the premium, quoting an estimate as if it traded and expecting a positive premium to make long bonds beat bills.
- Using the spread as the premium. As the 2016 example shows, a steep curve can coexist with a negative premium.
- Treating an estimate as a price. On October 23, 2026, two Federal Reserve models differed by 0.27 point, and monthly re-estimation revises ACM history. Give the model and the date with any figure, and compare changes within one series.
- Reading a positive premium as a promise. It is the extra yield expected if short rates follow the modeled path. If the premium itself jumps, existing long bonds lose value: a new 10-year note loses about 7.5% of its price when its yield rises 1 point, as bond duration shows.
- Confusing it with default risk. Treasuries are backed by the full faith and credit of the US government. The premium pays for the risk that rates change over the bond's life, and a corporate bond adds a credit spread on top.
- Expecting the Fed to set it. The FOMC sets an overnight rate range, and a decision moves the premium only if it changes how much investors want to be paid for uncertainty. Over the year to October 29, 2026, most of the rise in the 10-year yield came through expected rates instead.