American/Bermudan options: Finite Difference vs Monte-Carlo

woodsdevil · October 28th, 2005, 10:49 am

Just a stupid question: do you expect the price of a Bermudan option to be much larger when pricing on a finite difference scheme than in Monte-Carlo.In particular for two or three-factor models, do you expect the upper bound returned by a Monte-Carlo dual-primal algorithm such as Andersen-Broadie to be less than that returned by a finite difference scheme ?My answer is: it should never be the case, unless the finite difference and/or the algorithm by Andersen-Broadie are not good.Does anyone have experience on this ?