On the Normal Inverse Gaussian Stochastic Volatility Model

In this article, the normal inverse Gaussian stochastic volatility model of Barndorf-Nielsen is extended. The resulting model has a more flexible lag structure than the original one. In addition, the second- and fourth-order moments, important properties of a volatility model, are derived. The model can be considered either as a generalized autoregressive conditional heteroscedasticity model with nonnormal errors or as a stochastic volatility model with an inverse Gaussian distributed conditional variance. A simulation study is made to investigate the performance of the maximum likelihood estimator of the model. Finally, the model is applied to stock returns and exchange-rate movements. Its fit to two stylized facts and its forecasting performance is compared with two other volatility models.