## The Normal Inverse Gaussian Lévy Process
The Normal Inverse Gaussian distribution and associated stochastic processes was introduced by Barndorff-Nielsen in 1) and 2). The name derives from its representation as the distribution of Brownian motion with drift time changed by the inverse Gaussian Lévy process. Since the distribution is infinitely divisible it gives rise to a corresponding Lévy process, which has been shown capable of accurately modelling the returns on a number of assets on German, Danish and US exchanges, see 4). The Normal Inverse Gaussian Lévy process is in many ways similar to the Variance Gamma process due to Madan and Seneta. Both belong to the family of Lévy processes of the Generalized Hyperbolic type, however they posses unique properties that make them particularly tractable and convenient for option pricing. In particular alone among hyperbolic family they have the property of being closed under convolution (i.e. the sum of independently distributed variables of the given type has the same type). Both processes can be represented as a time change of a Brownian motion with drift by a Lévy process with increasing increments. In the case of the Variance Gamma process the time change process is the Gamma process, in the case of a NIG process it is an Inverse Gamma Process. Both processes are pure jump Lévy processes (they have no continuous Brownian component) but they differ in the nature of jumps: the Variance Gamma process has jumps of finite variation while the variation of the NIG process is infinite. There are several common parametrization of the NIG process. Most of them use 4 parameters. Here, however, we use a 3-parameter representation given in 4). The parameters are the drift and volatility of the time changed Brownian process and the variance of the time change process, which is assumed to have expectation 1. (This does not loose generality due to scaling properties for the NIG process). This representation brings out clearly the similarity between the NIG process and the Variance Gamma process; in particular when the variance of the time change tends to 0 both processes approximate the Brownian motion process. 1.) O. E. Barndorﬀ-Nielsen, "Normal inverse Gaussian distributions and stochastic volatility modelling". Scandinavian Journal of Statistics 24,1997 pp. 1–13. "The Normal Inverse Gaussian Lévy Process" from The Wolfram Demonstrations Project http://demonstrations.wolfram.com/TheNormalInverseGaussianLevyProcess/ Contributed by: Piotr Teklinski, Piotr Wroblewski, Andrzej Kozlowski |