Mathematical Finance Seminar

May 1, 2003, 5:30 PM to 7:00 PM

Sergei Levendorskii, UT Austin

Optimal stopping in non-Gaussian models, with applications to pricing of American options and real options

Abstract: A general approach to optimal stopping problems based on the Wiener-Hopf factorization technique is developed. Jump-diffusion models and wide classes of exponential L\'evy models in continuous time, and discrete time models are considered. The failure of the smooth pasting condition in discrete time models and certain L\'evy models is shown, and its generalization is suggested. In the infinite time horizon case, explicit analytical formulas are derived, and in the finite time horizon case, semi-explicit pricing algorithms are constructed. In continuous time models, the algorithm is an approximate method: an analog of Carr's randomization procedure. For wide classes of random walks and jump-diffusion processes in continuous time, the pricing algorithm reduces to a series of simple algebraic problems, which makes the method efficient for numerical implementation. As applications of the method, pricing of American options and Bermudan options, and optimal timing of investment under uncertainty are considered, and the non-standard behavior of the early exercise boundary in the non-Gaussian case is analyzed.