Valuing target redemption notes by a stratified Longstaff Schwartz algorithm

Abstract

A great challenge using the traditional regression based Bermuda option valuation based on Longstaff and Schwartz (LS) (see Longstaff and Schwartz [10]) is the stability of solutions for different basis functions. In this paper we develop an alternative method in the spirit of LS which is less challenging with respect to proper choice of basis functions. The method also makes it possible to quantify the probability of exercise at future nodes in a Bermuda option when moving backward in time. We will apply the method to valuation of target redemption notes with early exercise features under stochastic interest rates based on a LIBOR market model.