QuarcHome
Model catalogueBack to the model index

Variance Gamma (VG)

Uses gamma-subordinated Brownian motion to model skew and kurtosis in returns.

How it works

Variance Gamma evaluates a Brownian motion with drift at a gamma-distributed business time. The Brownian drift drives asymmetry and the random clock adds excess kurtosis. The resulting Lévy process has jumps and a tractable characteristic function for European option prices. Risk-neutral drift is set so the discounted asset price is a martingale.

Worked example

For an illustrative one-year clock with mean 1 and variance 0.2, two paths can have the same calendar length but very different accumulated business time. The larger clock gives a broader conditional distribution of the return. Sending clock variance toward zero recovers the corresponding Brownian return model.

Validation and limitations

Validate exponential moments and compare Fourier prices against simulation or numerical integration. Short-dated option data can give limited information about rare large moves. Use held-out strikes and stress the clock and drift parameters separately.

Research sources