Estimating random variance option pricing models: an empirical analysis

Abstract

This paper deals with the estimation of continuous time stochastic volatility models, which have been introduced in finance to price options on stock and other derivative assets. I start by showing that the indirect inference estimators do not suffer from the bias coming from the approximate discretization of the model, this operation being necessary due to the discrete time frequency of observations. Moreover, I suggest to use observed option prices to estimate the parameters of interest (those of the volatility process) because this avoids a possible mispecification of the drift of the price process. The sample I analyse refers to the CAC 40 index, and it includes 775 daily observations of the price of the index, along with an incomplete panel of corresponding option prices. The results are encouraging, but they also suggest the need of further generalizing both the model and the estimation method.