Sample Sizes for Reliably Estimating Lower and Upper Income Shares in Income Distribution Analysis

This paper uses distribution-free formulas for the asymptotic variances of sample quantile income shares – as typically published by statistical agencies as measures of the distribution of income inequality – to calculate how large a survey sample must be in order to estimate a more refined quantile breakdown for a given level of confidence. The approach is applied to decile and quintile earnings data to calculate required increases in sample size to obtain tail 5 percent quantal share estimates and to test changes in income shares. Simple rules of thumb are offered for such a required increase.