QuarcHome
Model catalogueBack to the model index

Normal inverse Gaussian (NIG)

Represents asymmetric, heavy-tailed return distributions using a normal variance-mean mixture.

How it works

The normal inverse Gaussian distribution is a normal mean–variance mixture with an inverse Gaussian mixing variable. Varying the mixing time produces heavier tails than a fixed-speed normal process, and a drift in the subordinated Brownian motion can make the distribution asymmetric. A pricing specification needs moment and martingale checks.

Worked example

Imagine two conditional normal returns, one at a short business time and the other at a much longer business time. The second has a larger conditional variance. Mixing the two produces more far-tail observations than a single normal with similar central dispersion. The NIG model uses a continuous distribution of such business times.

Validation and limitations

Check parameter restrictions, tail sensitivity and integration accuracy. A fit to physical returns needs separate risk-neutral calibration before option pricing. The CGMY family provides another way to vary asymmetry and jump tails.

Research sources