Extending the Least Squares Monte Carlo Method for American Options with Discrete Dividends

This thesis develops an extension of the Least Squares Monte Carlo algorithm to price American-style options on assets with discrete dividends.Two extra exercise dates, immediately before and after each dividend, are inserted into the original time grid. Detailed mathematical derivations and a practical code solution are presented. Numerical experiments show pricing errors of approximately 0.2% for puts and 1 % for calls versus thechosen benchmarks. Optimization techniques are presented and implemented to reduce the algorithm’s runtime, but it still remains far slower than the competitive benchmark.