Real-World Yield-Curve Dynamics and Dynamic Nelson–Siegel
A portfolio exposed to several maturities can lose money when the yield curve changes shape, even if the average rate barely moves. Historical factor models describe many curve movements with a small number of components associated with level, slope and curvature. This supports scenario design and forecasting under the real-world measure. Check factor stability across estimation samples and distinguish fitted historical dynamics from those required for derivative valuation.
How it works
The dynamic Nelson–Siegel model represents yields with maturity-specific loadings on three time-varying factors. Level contributes at all maturities. Slope has its largest loading near the short end. Curvature primarily affects intermediate maturities. Forecast the factors as a time series and then map their forecasts back to each maturity.
Worked example
A one-percentage-point rise in the level factor adds one point at every maturity under the standard loading. A one-point rise in the slope factor raises short yields much more than very long yields. Apply both factor shocks to key-rate durations so a portfolio with offsetting exposures is not treated as a single-duration position.
Validation and limitations
Check walk-forward forecast accuracy against a no-change benchmark and track changes in the estimated factors as the curve sample changes. Distinguish yields from zero-coupon forward rates and specify the maturity units used in the decay term. The fitted real-world factor dynamics have a different purpose from an arbitrage-free pricing model.
Further reading
- Forecasting the Term Structure of Government Bond Yields
Francis Diebold and Canlin Li · Journal record
- The U.S. Treasury Yield Curve: 1961 to the Present
Refet Gürkaynak, Brian Sack and Jonathan Wright · Federal Reserve working-paper PDF