A multi-dimensional local average lattice method for multi-asset models

Abstract

We develop a multi-dimensional local average lattice method in order to compute efficiently and accurately the price of multivariate contingent claims. The proposed method improves the accuracy of the standard lattice method by considering the local averages of option prices around each node at the final time, rather than the prices at the nodes. The average value smooths the oscillatory behavior of the lattice method, which leads to fast convergence of the option values. Numerical computations show that the proposed local average lattice method is more efficient than other lattice methods for a given level of accuracy.