Pricing American Options by Canonical Least-Squares Monte Carlo

Abstract

Options pricing and hedging under canonical valuation have recently been demonstrated to be quite effective, but unfortunately are only applicable to European options. In this paper, a variation of canonical valuation called canonical least-squares Monte Carlo is proposed to price American options, which proceeds in three stages. First, given a set of historical gross returns (or price ratios) of the underlying for a chosen time interval, a discrete risk-neutral distribution is obtained via the canonical approach. Second, from this canonical distribution independent random samples of gross returns are taken to simulate future price paths for the underlying. Third, to those paths the least-squares Monte Carlo method is then applied to compute a price for an American option. Numerical results obtained from using simulated gross returns under geometric Brownian motion (GBM) show that this new approach yields reasonably accurate prices for American put options and can be utilized as an alternative to pricing American options, whether the underlying follows GBM or not.