Measuring Inequality: On the Correlation of Indices

The preliminary step in assessing the extent of inequality is to decide how to measure it. Different indices exist, each responding to a built-in "aversion to inequality", and the choice of the index to be used affects conclusions. Whilst there is not a "preferable" index, the family of General Entropy Measures presents some beneficial properties. Yet, indices belonging to this family are not ordinally equivalent when Lorenz curves intersect. Therefore, conclusions about the extent and the change in inequality are driven by the assumptions behind the index chosen. Although the ranking correlation between inequality indices, empirically tested, is very high, a distinction should be drawn between cases involving non-intersecting and intersecting Lorenz curves. In the latter case, the choice of the index is fundamental to assess the distributional change. Therefore, a composite analysis of inequality through a series of indices, each sensitive to different parts of the distribution, is always preferable because it gives more details on the type of inequality and where it is more concentrated.