Using GMM to flatten the option volatility smile

By using an over-identified generalized method of moments (GMM) estimation procedure with careful consideration for data biases existing in the previous literature, parameters are estimated for a stochastic volatility jump diffusion option pricing (SVJ) model. The estimated parameters indicate a statistically significant highly negative infrequent jump process in the underlying security return distribution consistent with market crashes. When comparing to a stochastic volatility (SV) option pricing model, the SVJ is more robust but not always the superior model. The robustness of the models is further gauged by evaluating performance up to a year beyond the estimation data.