Option Pricing Method With Nonparametric Correction: ACE

Parametric option pricing models are largely used in financial area, and nonparametric methods can be used to learn and correct the pricing error of parametric model, so we can combine parametric pricing models with non-parametric approaches to make the pricing performance of parametric models can be significantly enhanced.In this paper, we propose a new nonparametric pricing method to price derivative assets. First, we choose a parametric model to estimate the state price distribution function initially. As distribution function is easier to be estimated than density function, our method relies on the former function of state price rather than the latter in this step. Then, we use some nonpara-metric method to fit the pricing bias of parametric model. By comparting different nonparametric methods, finally we choose local linear regression as the nonparametric method. This model-guided method that nonparametri-cally estimates the state price distribution is called Automatic Correction of Errors (ACE).We develop a nonparametric test based on the generalized likelihood ratio to test the efficacy of ACE, and use SPX 500 index options to do the empirical studies.ACE method can be easily combined with many model-based parametric pricing formula, and get better pricing performance by correcting the system-atic biases of initial model. At the end of this paper, we use ACE method to correct the jump-diffusion model for pricing currency option.