An “American” Monte Carlo（345-351）.pdfVIP

13 An “American” Monte Carlo* Key words: American options, Snell envelope, Monte Carlo simulation We present a method to price American-style options using Monte Carlo simulation as proposed by Rogers (2002). This method is an alternative method to the famous simulation-based technique introduced by Longstaff and Schwartz (2001) (LS). Roger’s proposal shares the starting point of the analysis with LS. They both start with the traditional dynamic programming equation, which conveys the idea that pricing an American-style option is a problem of knowing, at each point of time, whether it is worth to exercise the option immediately or to continue holding the option. As opposed to LS who attempt to determine an optimal exercise policy, the method we adopt here transforms the dynamic programming equation in a dual problem, proposes a financial interpretation of this latter, and finally tries to solve it numerically. Several pricing methods for American-style options have been proposed in the specialized literature (see, e.g., Lamberton and Lapeyre (1996) for a quick overview). In the case of options written on a large basket of underlying assets, these techniques tend to become quite inefficient and computationally slow. In order to overcome these difficulties, Carrière (1996) proposed a simulation-based method, which suf- fers, however, from numerical instability problems. Bouchard, Ekeland and Touzi (2004) proposed an alternative based on the notion of Malliavin derivative. How- ever, it suffers from a computational drawback in terms of speed for large baskets. Broadie and Glasserman (1997) suggest a stochastic mesh method for overcoming these issues and their proposal has gained popularity among practitioners. The present chapter is organized as follows. Section 1 quickly summarizes the American-style option pricing theory (see Carr, Jarrow and Myneni (1992), Jamshid- ian (1992), and Lamberton and Lapeyre (19