Sensitivities for Bermudan Options by Regression Methods

In this article we propose several pathwise and finite difference basedmethods for calculating sensitivities of Bermudan options using regressionmethods and Monte Carlo simulation. These methods rely on conditionalprobabilistic representations which allow, in combination with aregression approach, for efficient simultaneous computation of sensitivitiesat many initial positions. Assuming that the price of a Bermudanoption can be evaluated sufficiently accurate, we develop a method forconstructing deltas based on least squares. We finally propose a testingprocedure for assessing the performance of the developed methods.