Multilevel Monte Carlo simulation for the Heston stochastic volatility model (Q6144993)

This article is devoted to the investigation of a nonstandard approach to apply the multilevel Monte Carlo (MLMC) method. More precisely, the author combine the (MLMC) method with a numerical scheme for the Heston model that simulates the variance process exactly or almost exactly and applies the stochastic trapezoidal rule to approximate the time-integrated variance process within the SDE of the logarithmic asset process. Novel MLMC estimators for path-independent options and for path-dependent options are defined. It is shown that, under some Lipschitz assumptions on the payoff, the convergence rate of the MLMC variance is 2 for path-independent options and \(1 - \epsilon \) for path-dependent options, for any \(\epsilon > 0\). These results apply for all parameter regimes. The author present numerical results to demonstrate the efficiency of the new MLMC estimators.