Ungrouping Income Distributions: Synthesising Samples for Inequality and Poverty Analysis

We describe a new method of facilitating inequality and poverty analysis of grouped distributional data by allowing individual income observations to be reconstructed from any feasible grouping pattern. In contrast to earlier methods, our procedure ensures that the characteristics of the synthetic sample exactly match the reported values. The performance of the algorithm is evaluated first by using household survey records to compare true income observations with their synthetic counterparts, then by comparing the true and generated values of the Gini coefficient and other inequality indices. The results indicate that the new technique is capable of reproducing individual data from grouped statistics with a high degree of accuracy.