
Abstract. We provide a general approach for deriving robust solutions to stochastic optimization problems based only on mean-covariance information about the distribution underlying the random cost vector. For a general class of objective functions, we show that the robust optimization problem is equivalent to solving a certain deterministic parametric quadratic program. Interesting results arise from comparing the robust solutions with those corresponding to entropy maximizing distributions for various criteria, such as target, fractile and option-type. We explore applications in robust portfolio management, multi-product pricing and insurance.