Multivariate binomial approximations for asset prices with nonstationary variance and covariance characteristics

In this article, we suggest an efficient method of approximatinga general, multivariate log-normal distribution by a multivariatebinomial process. There are two important features of such multivariatedistributions. First, the state variables may have volatilitiesthat change over time. Second, the two or more relevant statevariables involve may covary with each other in a specifiedmanner, with a time-varying covariance structure. We discussthe asymptotic properties of the resulting processes and showhow the methodology can be used to value a complex, multipleexerciseable option whose payoff depends on the prices of twoassets.