Estimating a General Stochastic Variance Model from Option Prices 

Methods have recently been devised for estimating from option prices a restricted class of stochastic variance models as well as deterministic variance models where the variance of the underlying asset is a deterministic function of the level of the underlying asset and time.  Although the option prices generated by both of these types of models improve upon Black-Scholes values, their remaining empirical shortcomings motivate the development of a procedure for estimating more general models.  This paper presents a method for estimating from option prices a general stochastic variance model.  This model permits correlation between innovations to the level and the variance of the underlying asset as well as flexible nonparametric specifications of the market price of variance risk and the drift and diffusion functions of the variance process.  A simulation experiment shows that the estimation procedure performs well on an artificial data set of the size available from the options markets.  Estimation of the model from S&P500 index option prices over the period June 1, 1988 through December 29, 1995 provides evidence of misspecification in the restricted stochastic variance model