US 10-Year Treasury: Fair Value Model

Model Methodology

This model decomposes the 10-year US Treasury yield into three theoretically motivated components:

Fair Value = α + r* + E[π] + TP

where α is a calibration constant equal to the historical mean of (Actual − raw composite), ensuring the residual is mean-zero over the full sample. Without this intercept, GDP-based r* chronically overstates the true neutral real rate — particularly post-GFC — producing a persistent level bias in the raw model.

Residual = Actual Yield (DGS10) − Fair Value

• Residual > 0: Yield trades above fair value → market is cheap / undervalued relative to fundamentals
• Residual < 0: Yield trades below fair value → market is rich / overvalued relative to fundamentals

Notes & Caveats

• r* uses a 10-year backward-looking window, making it a lagging proxy; structural breaks in potential growth may not be captured promptly.
• The model carries no free parameters — all three components are pulled directly from FRED without calibration.
• Data availability limits the sample: the earliest fully-populated observation depends on the ACM term premium series (available from 1961 on a limited basis; EXPINF1YR from ~1978).
• Daily series (DGS10, ACMTP10) are averaged to monthly; quarterly GDP is forward-filled.

Sources: Federal Reserve Board (Kim-Wright term premium, THREEFYTP10), Federal Reserve Bank of New York (inflation expectations, EXPINF1YR), Bureau of Economic Analysis (Real GDP, A191RO1Q156NBEA), Federal Reserve H.15 (DGS10). All data via FRED.