A Parallel Algorithm for Pricing American Options

Abstract

We propose a novel, fast, accurate parallel algorithm for pricing American options. We perform a thorough numerical analysis of existing methodologies and find that ours performs significantly better. The proposed method is stable, robust, and converges monotonically. We also show that the efficiency can be improved further through the use of Richardson extrapolation. The method is well-suited for parallel implementations that take advantage of multi-core processors and graphic processing units.