Efficient Trees for CIR and CEV Short Rate Models

Abstract

This article presents truncated-tree transforms for generating binomial and trinomial trees under the Cox, Ingersoll, and Ross (CIR) and constant-elasticity-of-variance (CEV) models of the short rate. The authors correct an error in the original square root transform of Nelson and Ramaswamy [1990], and modify their transform by truncating the tree exactly at the zero-boundary. This not only allows for the creation of more efficient trees for the CIR square-root process, but also for the entire class of CEV models of the short rate. The simulations in this article show fast convergence and significantly improved performance of the truncated-tree approach over the Nelson-Ramaswamy approach.

TOPICS: Statistical methods, fixed income and structured finance, performance measurement