Valuing American Derivatives by Least Squares Methods

Least Squares estimators are notoriously known to generate sub-optimal exercise decisions when determining the optimal stopping time. The consequence is that the price of the option will be underestimated. We show how to use variance reduction techniques to extend some recent Monte Carlo estimators for option pricing and assess their performance in finite samples. Finally, we extend the Longstaff and Schwartz (2001) method to price American options under stochastic volatility. This is the first study to implement and apply the Glasserman and Yu (2004b) methodology to price Asian options and basket options.