Guide

Term premium explained

Harbor Capital's fixed-income desk ran a post-mortem on 2022: the 10-year Treasury yield jumped from 1.5% to 4.3% while the Fed's own dot plot moved only 300 basis points. The team had sized duration using implied forward rates from the yield curve as if long yields were pure expectations of future overnight rates. They were not. Roughly 120–150 bps of the 2022 sell-off was rising term premium — compensation for inflation uncertainty, supply, and the convexity risk of holding 10-year paper while the Fed fought price shocks. After the refactor, duration targets now subtract a model term premium before comparing market yields to the desk's rate path. Drawdowns in rate-volatility regimes improved 35 bps annualized with similar carry.

Every long bond yield embeds two ideas: what the market thinks short rates will average over the life of the bond, and an extra spread investors demand to commit capital for years instead of rolling T-bills. That spread is the term premium. It can be negative (as in 2010s Japan-style environments), spike during fiscal worries, and move independently of the next Fed meeting. This guide covers yield decomposition, how researchers estimate term premium (ACM, surveys, simple spreads), what drives it, the Harbor Capital duration sleeve refactor, a technique decision table vs curve trades and inflation breakevens, pitfalls, and a production checklist.

What term premium is

The term premium (also called bond risk premium or term spread premium) is the expected excess return of holding a long-term bond over repeatedly investing in short-term instruments at the rolling market rate. Intuitively: why accept a 4.5% 10-year yield when you could buy 3-month T-bills and reinvest? Because future bill rates are uncertain, long bonds can rally or sell off, and you may want compensation for that risk.

Formally, decompose the n-year zero-coupon yield y(n):

y(n) ≈ average expected short rate over n years + term premium(n)

The first term is the expectations component — the market's implied path of overnight or short policy rates, rolled forward and averaged. The second is everything else: inflation uncertainty, supply/demand for duration, preferred habitat effects, and risk aversion. Under the pure expectations hypothesis, term premium is zero and long yields are perfect forecasts of future short rates. Empirically, that fails often enough that serious allocators treat term premium as a first-class state variable.

Decomposing nominal yields

Spot curve vs forward curve

The Treasury spot curve shows yields on bonds of different maturities today. The instantaneous forward curve implied by those spots answers: “What one-period rate does the market embed for a future date?” Forward rates mix expectations and term premia at each horizon. A steep 10-year spot yield with flat forwards far out can mean high term premium near the belly, not necessarily higher policy rates forever.

Example decomposition (illustrative)

Suppose the 10-year Treasury yields 4.20% and a model estimates:

Expected average short rate (10y) — 3.50%
10-year term premium — +0.70%

A duration bet is partly a bet on the path of Fed funds (expectations) and partly a bet on whether 0.70% is too high or too low (term premium). In 2022, both legs rose; in 2016, term premium was often negative while expectations were stable, making long bonds look “expensive” on a risk-adjusted basis even when the curve looked flat.

Separate from inflation breakevens

Nominal yields split into real yields plus inflation expectations (see breakevens). Term premium applies within each leg: there is a real term premium and an inflation risk premium embedded in breakevens. Confusing the two leads to misreading TIPS vs nominal trades. This guide focuses on the nominal side; pair it with TIPS analytics when hedging CPI.

How term premium is estimated

Term premium is not directly observable; it is a residual after modeling expectations. Common approaches:

Method	Idea	Pros / cons
ACM / Kim–Wright (Fed NY)	Affine term-structure model separates expected rates from premium	Standard benchmark; model-dependent, revised with new data
Survey-based	Compare 10y yield to Blue Chip or SPF long-rate forecasts	Intuitive; surveys are sticky and may lag markets
Forward-spread rules	10y minus rolling average of realized short rates	Quick and dirty; backward-looking, noisy
Macro factor models	Regress bond excess returns on growth, inflation, supply	Explains drivers; less useful for daily marks

Reading the Fed's ACM chart

The Federal Reserve Bank of New York publishes ACM term premia for 1- through 10-year horizons. When the 10-year ACM premium is near zero or negative, long bonds offer little extra compensation beyond expected hikes — duration is “cheap” in risk terms but vulnerable if uncertainty returns. When ACM 10-year premium exceeds +100 bps, investors are paid more for bearing rate risk; historically that coincided with better forward excess returns on Treasuries, though timing is imprecise.

Harbor Capital ingests ACM 10-year weekly, cross-checks against the desk's internal OIS-implied policy path, and flags disagreements larger than 50 bps for the risk committee.

What drives term premium

Inflation uncertainty — wider CPI forecast dispersion raises the premium for locking nominal coupons; links to monetary policy credibility.
Fiscal supply — heavy Treasury issuance of long bonds can lift term premium even if the Fed is on hold (2020–2023 debates).
Foreign demand — reserve managers and pension funds absorbing duration suppress premium; repatriation or hedging flows reverse it.
Volatility regime — high rate vol increases convexity and hedging costs, often widening premium; see duration and convexity.
Business cycle — late-cycle inflation scares and early-cycle cuts both move premium differently from expectations.
Central bank QE/QT — large Fed balance-sheet duration removal (QT) can add term premium by forcing private holders to absorb supply.

Term premium is countercyclical in many samples: it tends to fall in recessions (flight to quality compresses spreads) and rise in uncertain expansions. Do not assume “higher yields always mean higher premium” — sometimes yields rise because expectations moved while premium fell.

Term premium across the curve

Premium is not uniform. ACM publishes estimates at 1y, 2y, … 10y horizons:

Front end (1–2y) — dominated by near-term policy expectations; premium is usually small.
Belly (5y) — sensitive to medium-term growth and inflation paths; often the most volatile premium segment.
Long end (10y+) — reflects secular supply/demand and inflation uncertainty; drives aggregate bond index duration risk.

Curve trades (2s10s steepeners, 5s30s butterflies) mix expectations and premium changes. A bear steepener can mean higher long-term premium with stable front-end expectations — painful for passive long-duration funds even if the Fed is done hiking.

Harbor Capital duration sleeve refactor

The pre-2023 rule: target portfolio duration to match a liability horizon using raw 10-year yields vs an internal Fed funds forecast. Failure mode: when term premium rose independently of the policy path, the fund added duration into a widening risk premium — double loss when yields spiked for non-Fed reasons.

New rules:

Estimate fair 10y yield = OIS-implied average policy rate over 10 years + ACM 10y term premium (or +75 bps floor if model unavailable).
Rich vs cheap duration: if market 10y yield > fair + 25 bps, add duration up to policy max; if < fair − 25 bps, trim toward neutral.
Premium shock filter: if ACM 10y premium rises > 30 bps in a month, cap duration additions until supply calendar and vol normalize.
Implement via Treasury ETFs (IEF, TLT) and futures overlays; separate nominal duration from TIPS allocation governed by breakeven rules.

Backtest 2005–2025: term-premium-aware duration reduced max drawdown in 2013 taper tantrum and 2022 episodes vs expectations-only targeting, with 0.15% lower annualized return in strong bull steepening rallies — an acceptable trade for pension-style mandates.

Technique decision table

Goal	Term premium lens	Alternative	When alternative wins
Forecast Fed hikes	Strip premium; read OIS forwards	Fed funds futures	Near-term meeting bets; cleaner policy signal
Decide long-duration allocation	Buy when ACM premium is elevated	Yield curve steepness alone	Steepness mixes expectations and premium without separating them
Hedge inflation uncertainty	Watch inflation risk premium in breakevens	TIPS when breakevens cheap	Pure CPI hedge without duration view
Trade supply shocks	Monitor premium after Treasury refunding	Curve butterflies on auction weeks	Tactical desks with auction flow data
Match pension liabilities	LDI with premium-aware discount rates	Immunization on government curve only	Small plans without macro overlay budget

Common pitfalls

Equating yield moves with Fed expectations — premium can dominate multi-month moves.
Using one model as truth — ACM revisions can shift historical premium by 50+ bps; triangulate.
Ignoring negative term premium — long bonds can look attractive on yield alone while offering poor risk-adjusted carry.
Confusing breakeven and term premium — inflation expectations and nominal premium are different constructs.
Static duration from liability date only — Harbor's lesson: macro premium regimes matter as much as calendar matching.
Over-trading ACM noise — weekly premium changes under 15 bps are often model noise, not signals.
Forgetting convexity at high premium — elevated premium often coincides with high vol; duration cuts may still be warranted for risk limits.
Applying U.S. premium to foreign bonds — ECB/JGB supply dynamics differ; use local estimates.

Production checklist

Track ACM (or equivalent) 5y and 10y term premia weekly alongside spot yields.
Decompose major yield moves into expectations vs premium using OIS forwards.
Compare model premium to survey-implied long-rate gaps quarterly.
Document Treasury refunding calendar and link to premium monitoring windows.
Set duration bands that reference fair yield = expectations + premium, not yield alone.
Separate nominal duration decisions from TIPS/breakeven allocation rules.
Stress-test portfolios for +100 bps premium shock independent of Fed path.
Log model revision dates; re-backtest rules when Fed NY updates ACM methodology.
Report premium contribution in monthly fixed-income attribution.
Review Harbor-style rules annually against realized excess returns of long bonds.

Key takeaways

Long bond yields = expected average short rates + term premium — not just a Fed forecast.
ACM and similar models estimate premium; treat them as benchmarks, not oracle numbers.
Term premium rises with inflation uncertainty, supply, and vol — often independent of the next FOMC meeting.
Elevated premium historically compensated duration risk; negative premium made long bonds look cheap but fragile.
Harbor Capital's duration sleeve improved drawdowns by sizing against fair yield that includes premium, not raw 10y levels.