Description: This paper considers the problem of parametric estimation of loss distributions in a very general context. It uses the minimum distance method with the Cramér-von Mises statistic as the particular choice of distance. The authors show how to compute the influence function for the estimator. They also use the influence function to obtain more information about the variance and the robustness of the estimator. They demonstrate, with an example, how much more resistant to contamination this estimator is compared with the more classical maximum likelihood estimator.