How to improve swaption calibration? A first-order correction to the HJM model.

A factor HJM stochastic volatility model helps price swaptions. Usually, the model "freezes" the nonlinear swap‑rate loading along a deterministic path. This freezing removes the dependence of conditional swap‑rate variance on the current yield‑curve state. Researchers add a first‑order Taylor correction to the loading. The correction introduces no new calibration parameters, keeping the model simple. Keeping the frozen annuity‑measure drift unchanged, the swap‑rate variation becomes affine in centered rate states. This reduces the problem to one‑dimensional equations for volatility. For quadratic‑drift lognormal stochastic volatility, these equations give a finite‑dimensional ODE representation and a direct log‑volatility formulation. Tests on independently generated prices show substantially lower calibration bias and out‑of‑sample pricing error. The model fits future prices accurately while remaining locally identifiable. Imagine the yield curve as a river that changes shape each day. A swaption is an option that gives you the right to exchange cash flows along that river. The HJM model captures how the river’s shape and its volatility evolve over time. Freezing the loading means assuming the river follows a predictable path. The first‑order correction is like a small push that improves the path without changing the rules of the game. Keeping the annuity drift unchanged preserves the model’s internal stability. Swap‑rate variability becomes more manageable, almost like a magic formula to reduce uncertainty. One‑dimensional equations simplify calculations, making the model faster on computers. The ODE representation lets you simulate many scenarios in real time. The log‑volatility formulation makes the time‑behavior of volatility intuitive. Calibration tests show the corrected model fits real prices better than earlier methods. This means more reliable prices for investors and lower risk for banks. The research paves the way for more accurate HJM models for other financial instruments.

The factor HJM model extends traditional yield‑curve models by incorporating stochastic volatility. Here, the swap rate is a nonlinear transformation of several latent factors. The swap‑rate loading quantifies how each factor influences swaption prices. The original method freezes this loading along a deterministic path, simplifying conditional variance. Freezing the loading removes the variance’s dependence on the current level of the yield curve. The proposed first‑order Taylor correction approximates the nonlinear loading with a linear term plus a small higher‑order remainder. Because the remainder is tiny, no new calibration parameters are needed. Keeping the frozen annuity‑measure drift unchanged preserves the swap‑rate dynamics’ stability. The first variation of the swap‑rate transform becomes affine in centered rate states. This affinity collapses the system to one‑dimensional volatility equations. For quadratic‑drift lognormal stochastic volatility, these equations yield a finite‑dimensional ODE representation. The ODE can be solved analytically, giving a closed‑form description of swap‑rate dynamics. The log‑volatility formulation converts stochastic equations into deterministic variables. This makes calibration easier, as volatility can be estimated directly from observed data. Simulations show the corrected model reduces bias in stochastic volatility parameters. Out‑of‑sample pricing errors drop markedly compared with the original model. Local identifiability stays intact, maintaining the model’s robustness. The first‑order correction is especially valuable when the yield curve is flat or steep. In those regimes the nonlinear loading is stronger, so the linear approximation is more accurate. Lower bias helps avoid over‑estimating volatility during market stress. Reduced conditional variability also improves interest‑rate risk management. This approach can be extended to other multi‑factor HJM frameworks.