Analysis of multilevel Monte Carlo path simulation using the Milstein discretisation

Abstract

The multilevel Monte Carlo path simulation method introduced by Giles ({\it Operations Research}, 56(3):607-617, 2008) exploits strong convergence properties to improve the computational complexity by combining simulations with different levels of resolution. In this paper we analyse its efficiency when using the Milstein discretisation; this has an improved order of strong convergence compared to the standard Euler-Maruyama method, and it is proved that this leads to an improved order of convergence of the variance of the multilevel estimator. Numerical results are also given for basket options to illustrate the relevance of the analysis.

33 pages, 4 figures. Corrections (marked in red) of some typos and omissions in the previously published versions; all results remain unchanged

Cited by in corpus (4)

• General multilevel Monte Carlo methods for pricing discretely monitored Asian options
• Mean-square convergence rates of implicit Milstein type methods for SDEs with non-Lipschitz coefficients
• Monte Carlo convergence rates for $k$th moments in Banach spaces
• Sub-sampling and other considerations for efficient risk estimation in large portfolios