Since the seminal work of Joslin, Priebsch, and Singleton (2014), a puzzling phenomenon has persisted in the macro-finance literature: the existence of “unspanned macro risks.” These are macroeconomic variables that have predictive power for future bond prices but are not spanned by the contemporaneous yield curve.

This presents a challenge to standard asset pricing theories. If a macroeconomic variable contains useful information for forecasting future returns, rational investors should immediately trade on it, causing the current yield curve to span that information. The existence of unspanned risks, therefore, implies a potential inefficiency in the bond market or a puzzle regarding how information is processed.

In our latest working paper co-authored with Kyu Ho Kang, “When Are Macro Risks Unspanned?: Evidence from a Regime-Switching Term Structure Model”, we provide a granular and theoretically grounded explanation for this phenomenon. We find that unspanned risks are not a permanent feature of the term structure but reflect episodic frictions in information transmission triggered by conflicting economic signals.

Using a regime-switching affine term structure model and Gaussian Process (GP) regression, we identified that real activity and inflation expectations become “unspanned” under very specific, distinct economic conditions.

1. Real Activity: The “Peak Growth Disconnect”

Figure 1. Probability of Real Activity Becoming Unspanned Risk and GP Regression Results

We find that real activity risks (proxied by the CFNAI) tend to become unspanned during a period we term the “Peak Growth Disconnect”.

Our analysis using SHAP values reveals that this regime emerges when:

• Equity prices (S&P 500) are rising,
• Valuation multiples (P/E ratio) are contracting (implying an earnings boom), and
• Real housing activity remains robust

Why does the bond market ignore such strong real activity signals? This connects directly to the theoretical mechanism proposed by Tanaka (2025) regarding imperfect information. Investors observe these robust fundamentals (“supply-side improvements”) but misinterpret them as transitory cyclical peaks or demand shocks that will not persist.

Because investors view this strength as temporary—a “Peak Growth” that will soon revert—they delay incorporating this information into the yield curve. Consequently, the signal remains unspanned until the ambiguity resolves.

2. Inflation Expectations: “Sectoral Decoupling”

Figure 2. Probability of Inflation Expectations Becoming Unspanned Risk and GP Regression Results

In contrast, unspanned risks in inflation expectations (IE) arise during periods of “Sectoral Decoupling”.

This occurs when the economy sends conflicting signals from different sectors:

• The persistent shelter sector (indicated by robust housing starts and permits) suggests inflationary pressure, while
• The cyclical goods sector (indicated by contracting manufacturing hours) signals weakness.

This divergence creates a “signal-to-noise” problem. As argued by Ulrich (2013), such conflicting signals generate Knightian uncertainty (ambiguity) regarding the true trend of inflation.

Faced with this ambiguity, the yield curve tends to span only the “average” expectation, effectively under-weighting the persistent inflationary pressure emanating from the housing sector. As a result, the survey-based inflation expectations contain predictive information that the bond market fails to capture immediately.

Conclusion: Information Frictions in Complex Economic States

Our findings offer a new perspective on the “unspanned puzzle.” It is not necessarily evidence of permanent market inefficiency or irrationality. Instead, the phenomenon reflects episodic information frictions that arise when the economy enters complex states.

As we documented, unspanned risks are not random noise. They are systematic responses to specific economic conditions:

During the “Peak Growth Disconnect,” investors struggle to distinguish between persistent supply-side improvements and transitory demand shocks, leading to under-reaction to real activity news.

During “Sectoral Decoupling,” the divergence between the sticky shelter sector and the cyclical goods sector creates Knightian uncertainty, causing the yield curve to miss the true inflation trend.

In both cases, the bond market does not simply “ignore” the economy. Rather, it faces a severe signal extraction problem. When economic signals become noisy or conflicting, price discovery is delayed. Therefore, to fully understand the yield curve, we must look beyond standard pricing models and account for the structural heterogeneity and informational constraints that investors face during these critical turning points.

For more details, please check out our full paper.

References

• Lee, S. and Kang, K. H. (2026), “When Are Macro Risks Unspanned?: Evidence from a Regime-Switching Term Structure Model,” Working Paper.
• Joslin, S., Priebsch, M., and Singleton, K. J. (2014), “Risk premiums in dynamic term structure models with unspanned macro risks,” The Journal of Finance, 69, 1197–1233.
• Tanaka, H. (2025), “Equilibrium yield curves with imperfect information,” Journal of Monetary Economics, 149, 103621.
• Ulrich, M. (2013), “Inflation ambiguity and the term structure of U.S. Government bonds,” Journal of Monetary Economics, 60, 295-309.