Fixed Income Duration, Convexity & Risk Models

Under the NY Fed Adrian-Crump-Moench (ACM) affine term structure model, the 10-year belly rallied on downward revisions to the near-term expected FOMC policy path (-11.4 bps) and safe-haven compression of intermediate term premia (-7.8 bps). In contrast, 30-year yields are dominated by the nominal term premium (+148.5 bps), which remained anchored by heavy Treasury refunding supply and a +3.2 bps jump in the long-horizon inflation risk premium driven by $114.20/bbl Brent crude.

As the 10-year Treasury yield broke below 5.05% toward 4.95%, projected homeowner refinancing speeds (CPR) on $1.85 trillion of UMBS 5.5% and 6.0% pools surged from 7.2 CPR to 13.8 CPR. This negative convexity (-340 years^2) shortened effective MBS duration from 5.42 years to 3.88 years, forcing mortgage servicers and OAS desks to buy 10-year Treasury futures and receive 10-year SOFR swaps at a rate of $14.5M–$18.2M DV01 per basis point.

Traditional scalar Modified Duration and single-factor DV01 hedges assume parallel yield curve shifts (PC1) and historical 10Y-30Y correlation near 0.935. On October 23, realized 10Y-30Y correlation dropped to 0.410 as the 10s30s spread steepened +15.4 bps intraday. Portfolios short 10Y DV01 against long 30Y DV01 lost $1.51M per $100k DV01 box, proving the necessity of multi-factor Key-Rate Duration (KRD) and cross-gamma hedging.

FICC GSD computes a 99% 3-day Sensitivity VaR using EWMA volatility scaling (lambda = 0.97) plus a Margin Liquidity Adjustment (MLC) and Intraday Market Move Charge (PD). When the -19.2 bps 10Y rally broke historical cross-tenor covariance assumptions and 1M10Y normal vol jumped to 118.4 bps/yr, FICC GSD and CME SPAN 2 increased required initial margin on 10s30s curve boxes by +28.6% intraday.