For fixed sample size N, the distribution functions of the k-class estimators (k nonstochastic) are approximated up to terms whose order of magnitude is 1/@u,where @u^2 is what is referred to in the literature as the concentration parameter. Similar results are also given for the double k-class estimators.