Volatility Bursts: A Discrete-Time Option Model with Multiple Volatility Components

Abstract

Abstract I propose an affine discrete-time model, called Vector Autoregressive Gamma with volatility Bursts (VARG-B) in which volatility experiences, in addition to frequent and small changes, periods of sudden and extreme movements generated by a latent factor which evolves according to the Autoregressive Gamma Zero process. A key advantage of the discrete-time specification is the possibility of estimating the model using Kalman Filter techniques. Moreover, the VARG-B model leads to a fully analytic conditional Laplace transform which boils down to a closed-form option pricing formula. When estimated on S&P500 index options and returns, the new model provides more accurate option pricing and modeling of the IV surface compared with some alternative models.