Computing option values by pricing kernel with a stochatic volatility model

To use a wider range of information available on the market, we propose a parameter estimation and option pricing procedure which involves a two step approach: in a first step real world parameters are estimated from time series data of the underlying financial asset, and in a second step the so called pricing kernel is computed from option data. For the first step we compare two likelihood based estimation procedures, namely the particle filter and the SEM algorithms. For the second step we use an adapted version of the so called asset specific pricing kernel. The results are then analyzed in a simulation study and implemented in a real dataset of the FTSE Mib Index, and compared with the classical calibration approach, which makes use of the option data only.