Goldman Sachs Updates G10 Term Premium Model: Incorporating Survey Data, Premiums Remain Elevated
AI summary card
Goldman Sachs Updates G10 Term Premium Model: Incorporating Survey Data, Premiums Remain Elevated
Goldman Sachs introduces new G10 term premium estimates based on the Kim-Wright method, complementing pure yield curve models; both methods show significant increases in term premiums in recent years, but the survey-based model better captures structural shifts in interest rate expectations.
- Introduces the Kim-Wright survey-enhanced model to supplement existing pure yield curve term premium estimates
- Term premiums indicated by the survey model are generally lower, more stable, and better reflect structural shifts in interest rate expectations
- US 10-year term premium: 70bp under the survey model, 40bp lower than the pure yield curve model
- Japan is the clearest case for the survey model's advantage, better distinguishing between term premiums and rising policy rate expectations
- The pure yield curve model is more volatile but has stronger predictive power for excess returns on medium-term duration holdings
- G10 term premiums have risen broadly in recent years, driven primarily by macro risks, inflation risks, bond supply, and central bank balance sheet reduction
Report interpretation
Overview
This report is an update by the Goldman Sachs Global Rates Strategy team on the estimation methodology for term premiums in G10 economies. The core change is the introduction of a survey-enhanced estimate based on the Kim-Wright (KW) method, alongside the existing pure yield curve model (similar to the ACM approach), utilizing survey data on market expectations for short-term rates to help identify the expectations component in yields. The report argues that both methods have value, but the survey model offers greater economic intuition in capturing structural shifts in interest rate expectations. Regardless of the method used, G10 term premiums have risen significantly in recent years and currently remain at relatively high levels.
Core views
The report's core views center on comparing two term premium estimation methods. The pure yield curve model infers future short-term rate expectations entirely from market yield curve information, defining the term premium as the difference between yields and the risk-neutral expected path. This method is more sensitive to the estimation window and more volatile, but it reflects market changes faster and has stronger predictive power for excess returns on medium-term duration holdings. The survey-enhanced model adds survey forecast data on short-term rates from institutions like Consensus Economics to yield curve information, directly constraining the path of future short-term rates. This allows the model to more accurately distinguish between 'rising expectations' and 'rising term premiums' during structural shifts in interest rate expectations (such as the sustained rise in Japan's policy rate expectations since 2023 and upgraded growth expectations following Germany's debt brake reform in 2025), avoiding the pure curve model's tendency to over-attribute curve steepening to term premiums. In terms of specific data: The US 10-year term premium is 70bp under the survey model, 40bp lower than the pure curve model; Europe is approximately 15bp lower; the UK was previously the highest among G4 in the pure curve model but becomes the lowest in the survey model. Japan shows the most significant divergence—the pure curve model attributes almost all yield curve steepening to rising term premiums, while the survey model indicates this is primarily due to a structural increase in policy rate expectations. The report emphasizes that for investors, understanding the direction and drivers of term premium movements is more important than focusing on precise levels. Currently, term premiums across almost all models and curves are at multi-year highs, reflecting a combination of cyclical risks, inflation risks, rising interest rate levels and volatility, increased bond supply, and central bank balance sheet contraction. Notably, despite the recent decline in implied interest rate volatility, term premiums remain elevated; this combination of 'high premiums and low volatility' suggests a structural increase in duration risk premiums beyond cyclical fluctuations.
Analysis framework
The report's main line of argumentation is 'model comparison and complementarity': it first establishes the technical foundation of two term premium decomposition frameworks, then demonstrates empirical differences and economic implications through cross-country (G4 and broader G10) and cross-event (Germany's debt brake reform, Japan's monetary policy normalization) case studies, and finally evaluates their relative value for investment practice. Methodologically, both models belong to Gaussian affine term structure models, with the core difference lying in how they identify the 'expectations component' in long-term yields. The pure yield curve model (based on the Joslin-Singleton-Zhu method) relies solely on cross-sectional yield information, fitting the curve and extracting the expected path via three latent factors; the ACM variant identifies the price of duration risk directly from realized excess bond returns. The survey-enhanced model (Kim-Wright style) adds survey forecasts as additional constraints in the observation equation, using Kalman filtering for joint estimation of yields and survey data. The report specifically focuses on the trade-off between 'real-time responsiveness' and 'economic intuition': survey data is less frequent (typically monthly) and subject to publication lags, making the survey model slower to react to sudden events; however, this constraint makes the model more stable and avoids misjudgments by pure market models during structural transitions. Furthermore, regression analysis testing the predictive power of both models found that the pure curve model has higher explanatory power (higher R²) for duration excess returns over 1-5 year holding periods, consistent with its higher volatility characteristics.
Methodology notes
Term Premium Decomposition
Long-term bond yields can be decomposed into the 'expected path of future short-term rates' and the 'term premium'. The term premium is the extra compensation investors demand for holding long-term bonds instead of rolling over short-term investments. Accurately decomposing these two is crucial for understanding bond market pricing, but direct observation is impossible, requiring model-based inference.
Affine Term Structure Models
A class of mathematical models that express yields as linear functions of latent state variables, assuming no-arbitrage conditions hold. By estimating parameters under the risk-neutral measure, observed yield curves can be decomposed into expectations and risk premium components, serving as mainstream tools for central banks and investment banks to estimate term premiums.
Three-Factor Decomposition of Yield Curves (Level, Slope, Curvature)
Major variations in yield curves can be explained by three principal components: the first PC (level factor) corresponds to parallel shifts, the second PC (slope factor) corresponds to relative changes between short and long ends, and the third PC (curvature factor) corresponds to relative changes between the medium term and both ends. Differences in term premium sensitivity to different factors are a key dimension for model comparison.
Value of Survey Data as Expectation Anchors
Market participants' subjective expectations of future rates are often dispersed and dynamic. Survey data (e.g., Consensus Economics) aggregates judgments from numerous forecasters, serving as a proxy for 'true expectations' in models, helping distinguish between 'market-priced expectations' and 'actual fundamental expectations', and reducing noise from single-market information.
Key data
- US 10-Year Term Premium (Survey Model)70bp40bp lower than pure yield curve model
- Europe 10-Year Term Premium (Survey Model)~15bp lowerGap relative to pure yield curve model
- UK Term Premium (Survey Model)Lowest in G4Previously highest in G4 under pure curve model; ranking reversed in survey model
- G4 Average Term Premium VolatilitySignificantly lower in survey modelDue to survey data constraining excessive volatility in expectation paths
- Explanatory Power of Pure Curve Model for Excess ReturnsHigher R²Outperforms survey model in 1-5 year holding period forecasts
Impact & implications
The report argues that the core value of term premium estimation lies in helping investors understand the drivers of interest rate movements rather than providing precise trading levels. With G10 term premiums generally at high levels, long-term bond yields contain a significant risk compensation component rather than merely reflecting neutral rate expectations. For asset allocation and duration management, complementary use of both models may be the optimal strategy: the survey model is better suited for judging medium-to-long-term structural trends (e.g., impact of Japan's monetary policy normalization and European fiscal expansion on rate paths), while the pure curve model is better for capturing short-term trading signals and predicting medium-term excess returns. The recent phenomenon of declining interest rate volatility alongside sustained high term premiums warrants particular attention, potentially suggesting that market pricing of duration risk has moved beyond purely cyclical factors into a phase of structural increase. The report previews follow-up research that will delve deeper into G10 term premium drivers and valuation frameworks, implying that the current analysis serves as a foundational step for a series of studies.
What to watch
- Follow-up research providing in-depth analysis of G10 term premium drivers and valuation frameworks
- Continued impact of central bank balance sheet policy changes across countries on term premiums
- Whether sustained low interest rate volatility will eventually drive term premiums lower