Electronic Thesis and Dissertation Repository


Master of Science


Statistics and Actuarial Sciences


R. Mark Reesor


In this work we study the problem of pricing multiple exercise options, a class of early exercise options that are traded in the energy market, using a modified Longstaff and Schwartz approach. Recent work by Letourneau and Stentoft (2014) shows American option price estimator bias is reduced by imposing additional structure on the regressions used in Monte Carlo pricing algorithms. We extend their methodology to the Monte Carlo valuation of multiple exercise options by requiring additional structure on the regressions used to estimate continuation values. The resulting price estimators have reduced bias, particularly for small sample sizes, and results hold across a variety of option types, maturities and moneyness. A comparison of the original Longstaff and Schwartz approach to the modified Longstaff and Schwartz approach demonstrates the strengths of the developed numerical technique.