Kybernetika

Studies dealing with the quantification of inequality of a population with respect to a given quantitative attribute, provide us with a large class of measures. Among these, we can distinguish, because of their properties and operativeness, the ones coinciding with, or being ordinally equivalent to, the dimensionless ``additively decomposable inequality indices".

As indicated in previous papers, many populations, whose inequality in relation with an attribute is useful to quantify, are too large to be censused but large samples from them can be drawn and they arise naturally stratified. On the basis of these last two advantages, we will approach in this paper the optimum allocation in estimating inequality, and a comparison with the proportional allocation, and with the absence of strata, will be later established.