In this paper we show that it is invalid to use standard maximum likelihood procedures in estimating jump-diffusion models. The reason is that in jump-diffusion models the log-return is equivalent to a discrete mixture of N normally distributed variables, where N goes to infinity. Thus, from the...