Abstract

This paper investigates the computational performance of direct and inverse finite element methods for pricing American options. The underlying concept of the direct approach is similar to that of conventional finite element method. But the inverse approach is a relatively new development that involves trading the roles of financial variables. Based on the same constitutive model and linear elements, a performance analysis of the two approaches is carried out against the benchmark solution. Furthermore, we present experimental results on their accuracy-efficiency trade-off. Results indicate that although both approaches possess good convergence to the benchmark result, the inverse method is more efficient in term of the acceptable computing time and accuracy.              

Key words:  Direct finite elements, Inverse finite elements, American-style options, Black-Scholes  model.