Term Premium: Models, Curve Conventions, and Valuation Transmission

Estimate term premium without confusing fitted zero-coupon yields with Treasury par yields, model-implied expectations with forecasts, or rate scenarios with mechanical asset-price effects.

For educational purposes only; not investment advice. Investing may result in loss.

Direct answer

The term premium is an unobserved, model-implied component of a longer-maturity yield relative to the model’s expected path of short rates. For a fitted nominal zero-coupon yield of maturity n, the organizing identity is:

n-year fitted zero-coupon yield = model-implied average expected short rate over n years + n-year yield term premium

The fitted yield, expected-rate component, and term premium must use the same model, maturity, compounding convention, and observation vintage. The premium is a residual under that model, not a directly traded price or a guaranteed positive reward. It can be negative, and a different model can assign a different split to the same market data.

Do not insert a Treasury par or constant-maturity Treasury rate directly into a zero-coupon identity. Treasury’s daily nominal CMT curve reports model-derived par yields based on indicative bid-side inputs and interpolation; a par yield is the coupon rate that prices a hypothetical bond at par. A zero-coupon spot yield discounts one maturity-specific cash flow, and a forward rate is implied from multiple spot yields. Those are related curve objects, not interchangeable observations.

Seven-step reproducible term-premium workflow

Freeze the claim, timestamp, and curve object. Record currency, nominal or real claim, maturity, observation date, source timestamp, frequency, compounding and day-count convention, and whether the input is a security yield, par/CMT yield, fitted zero-coupon spot yield, or forward rate. Preserve the raw vintage before later revisions.
Select the model and estimation vintage. Name ACM, Kim-Wright, or another specification; record model version, sample end, factors, estimation frequency, survey inputs, and real-time or revised status. ACM is a five-factor, yield-only affine model estimated through sequential linear regressions; Kim-Wright is a three-factor Gaussian affine model estimated by maximum likelihood and incorporates survey forecasts of the three-month Treasury bill rate with measurement error.
Reproduce the fitted curve before decomposing it. Match maturity and convention, compare fitted with observed inputs, and retain pricing or fit residuals. The New York Fed ACM release provides fitted zero-coupon yields, model-implied average expected short rates, and yield term premia for maturities from one through ten years; it does not transform Treasury’s published par CMT number into an observed premium.
Decompose the fitted yield on one basis. Apply yield term premium = fitted zero-coupon yield − model-implied average expected short rate. Report levels and changes in basis points, keep full precision until the final display, and treat the result as an estimate with sampling, parameter, specification, and data uncertainty.
Derive forwards and forward premia separately. Under annual effective compounding, 1y1y forward = (1 + z₂)² ÷ (1 + z₁) − 1. Compare that forward with the model-implied expected one-year short rate for that future interval to obtain a forward term premium. A yield term premium averages effects through a maturity; it is not the forward premium at one future horizon. Some affine-model definitions also embody convexity terms, so labels and equations must travel with the series.
Triangulate models, vintages, and real-rate evidence. Compare ACM and Kim-Wright rather than treating either as truth. Survey information can reduce sensitivity to persistent-rate small-sample problems, but survey frequency, sample choice, factor count, estimation method, curve inputs, expanding-sample re-estimation, and methodological revisions still matter. For nominal-real analysis, align comparable zero-coupon maturities and retain inflation-risk, liquidity, indexation, and convention wedges.
Translate the scenario to the actual asset claim. For an option-free bond, use cash-flow repricing plus modified or key-rate duration and convexity. For a company, separately model nominal or real cash flows, refinancing, leverage, margins, currency, equity risk premium, and cost of equity. A term-premium change affects an equity DCF only through an explicit transmission assumption; an equity is not mechanically a fixed-duration Treasury.

At the nominal level, a model can organize nominal zero-coupon yield = expected nominal short-rate path + nominal term premium. A TIPS model can organize a real-yield analogue, but observed TIPS also carry liquidity, indexation-lag, tax, and deflation-floor features. Likewise, nominal Treasury yield − comparable TIPS yield is breakeven inflation, not pure expected inflation: it can include inflation risk premium and relative liquidity or technical wedges. Without a common model, nominal term premium does not mechanically equal real term premium plus inflation risk premium.

Worked examples

Yield decomposition and attribution. A model reports a fitted 10-year zero-coupon yield of 4.50%, an average expected short-rate component of 3.70%, and a yield term premium of 0.80%, so 4.50% = 3.70% + 0.80%. In a later vintage it reports 4.90% = 3.80% + 1.10%. Within that model, the fitted-yield change is +40bp = +10bp + 30bp. This is model attribution, not evidence that an observable security paid a 30-basis-point premium.
Spot-to-forward conversion and forward premium. Suppose one-year and two-year zero-coupon effective annual yields are z₁ = 4.0000% and z₂ = 4.5000%. Then 1y1y forward = (1.045)² ÷ 1.04 − 1 = 5.0024%. If the model’s expected one-year short rate for that future year is 4.6000%, the forward term premium is 5.0024% − 4.6000% = 40.2404bp. This forward premium is not the two-year yield term premium.
Duration, convexity, and exact bond repricing. A 10-year, 4% coupon bond with semiannual payments, face value 100, and nominal YTM of 4.50% compounded semiannually has P₀ = 96.0091, modified duration 8.1152, and convexity 78.0053. For a parallel +40bp yield change, ΔP ÷ P ≈ −8.1152 × 0.0040 + 0.5 × 78.0053 × 0.0040² = −3.1837%. Exact repricing at 4.90% gives P₁ = 92.9516, or −3.1845%. The shock is to total YTM; calling all 40bp a term-premium shock requires a separate model scenario.
Nominal-real DCF consistency and a nonmechanical equity channel. With next-period real FCFE of 100, real cost of equity 6.00%, and real perpetual growth 2.00%, value is 100 ÷ (6.00% − 2.00%) = 2,500. At 2.50% inflation, Fisher-consistent nominal inputs are FCFE₁ = 102.5, Kₑ = 8.65%, and g = 4.55%, preserving 102.5 ÷ (8.65% − 4.55%) = 2,500. If an explicit scenario assumes a +50bp term-premium shock passes fully into nominal cost of equity while cash flow and growth stay fixed, value becomes 102.5 ÷ (9.15% − 4.55%) = 2,228.2609, a −10.8696% change; the corresponding real cost is 1.0915 ÷ 1.025 − 1 = 6.4878%. Full pass-through and unchanged cash flows are assumptions, not laws.

Risks and review controls

Freeze currency, nominal or real claim, maturity, observation date, download timestamp, frequency, compounding, day count, and data vintage.
Label every input as an individual security yield, Treasury par/CMT yield, fitted zero-coupon spot yield, or forward rate.
Do not substitute a Treasury CMT par yield for a fitted zero-coupon yield in a model decomposition.
Record indicative-quote, interpolation, curve-fitting, and observed-transaction limitations in the underlying yield data.
Name the model, version, factor count, estimation method, sample start and end, and survey or macro inputs.
Preserve real-time releases when evaluating historical decisions; revised estimates can embed later data and re-estimation.
Separate a new market observation from a curve-data revision, expanding-sample re-estimation, and a methodology change.
Reproduce the fitted yield and inspect fit residuals before accepting the expectations and premium components.
Report point estimates with uncertainty; confidence intervals can be economically wide even when a series looks smooth.
Compare ACM, Kim-Wright, and other defensible specifications instead of averaging unlike definitions without documentation.
Distinguish a maturity’s yield term premium from a forward term premium for a specific future interval.
Keep annual effective, bond-equivalent, continuously compounded, and other rate conventions consistent in conversions.
Treat breakeven inflation as a nominal-real yield spread containing expectations, risk premium, liquidity, and technical wedges.
Match nominal and real maturities, curve constructions, cash-flow tax bases, and inflation conventions before comparison.
Do not infer that a negative estimated premium makes a long bond riskless or removes duration loss.
Frame supply, volatility, central-bank holdings, inflation uncertainty, hedging demand, and liquidity as hypotheses, not identified causes.
For bonds, test nonparallel shifts with key-rate duration and reprice embedded options rather than relying only on modified duration.
For credit, separate risk-free curve, spread, default, liquidity, callability, and funding changes.
For equities, model cash flows, growth, leverage, equity risk premium, and discount rates separately; do not assign Treasury duration mechanically.
Archive code, inputs, transformations, rounding, outputs, model vintage, and an audit trail sufficient to reproduce every chart and conclusion.

Common misconceptions

“Term premium is the 10-year Treasury yield minus today’s policy rate.” The comparison is with a model-implied average path of future short rates on a matched basis, not one current overnight rate.
“Treasury’s published CMT rate is a zero-coupon spot rate.” CMT rates are par yields from Treasury’s fitted par curve; spot and forward rates are different curve objects.
“A forward rate is the market’s pure forecast of the future short rate.” It can include a forward term premium and model-dependent convexity effects.
“Breakeven inflation is pure expected inflation, so nominal and real term premia reconcile automatically.” Inflation risk, relative liquidity, indexation, floors, taxes, and model conventions can create wedges.
“A term-premium rise mechanically lowers every bond and stock by a fixed percentage.” Bond effects depend on the full curve and cash flows; equity effects require assumptions about cash flows and how the shock enters the discount rate.

Authoritative sources

New York Fed Treasury Term Premia - ACM daily and month-end fitted zero-coupon yields, expected average short rates, and term-premium estimates for one- through ten-year maturities.
Adrian, Crump, and Moench Staff Report 340 - five-factor affine term-structure estimation through linear regressions and the model’s pricing and risk-premium framework.
Federal Reserve Three-Factor Nominal Term Structure Model - Kim-Wright model structure, survey-forecast inputs, and published nominal term-premium estimates.
Federal Reserve Robustness of Long-Maturity Term Premium Estimates - ACM and Kim-Wright methodological differences, sample sensitivity, survey information, and estimate uncertainty.
U.S. Treasury Daily Par Yield Curve Rates - official daily CMT par-yield observations and maturity series.
U.S. Treasury Yield Curve Methodology - indicative bid-side inputs, monotone-convex interpolation, par-yield construction, and methodological scope.
Federal Reserve TIPS Yield Curve and Inflation Compensation - staff estimates of real zero-coupon yields and inflation compensation, including model and revision cautions.
FRED 10-Year Breakeven Inflation Rate - daily spread constructed from nominal and inflation-indexed 10-year Treasury CMT series.