Statement
The unusual decline in long-term yields during the 2004-05 monetary tightening episode was primarily driven by a fall in term premiums rather than changes in expected future short rates. The Kim-Wright model estimates that ~120 bps of the 150 bps decline in the 10-year instantaneous forward rate was due to declining term premiums, with about two-thirds attributable to falling real term premiums and one-third to declining inflation risk premiums.
Evidence summary
Kim and Wright (2005) estimate a three-factor Gaussian arbitrage-free term structure model using weekly yield data (1990-2005) augmented with Blue Chip survey expectations:
- 10-year zero-coupon term premium declined ~80 bps (June 2004 to July 2005).
- 10-year instantaneous forward term premium declined ~120 bps.
- Real term premium fell from 125 bps to 50 bps; inflation risk premium fell ~30 bps.
- The model’s term premium correlates 0.83 with the Cochrane-Piazzesi (2005) return-forecasting factor.
- Expected future real rates edged lower ~10 bps; expected inflation fell ~10 bps.
Conditions and scope
- Gaussian affine model: no stochastic volatility, no regime switching. Time-invariant parameters assumed.
- Stationarity of factors means sufficiently distant-horizon forward rate changes are always attributed to term premiums by construction.
- Survey data plays a key identification role; results may be sensitive to survey specification.
- The Kim-Wright model became the Fed’s standard term premium tool, widely used in policy analysis.
Counter-evidence
- The stationarity caveat means the decomposition is partly a modeling assumption, not purely data-driven.
- Regime-switching models may attribute some of the movement to shifts in long-run expected rates rather than term premiums.
Linked ideas
Open questions
- How would regime-switching extensions change the decomposition during this episode?
- Is the term premium decline structural (reflecting lower macro uncertainty) or cyclical?