Inequality measurement with grouped data: parametric and non- parametric methods.

Abstract

Grouped data in the form of income shares have conventionally been used to estimate income inequality due to the lack of individual records. We present a systematic evaluation of the performance of parametric distributions and non-parametric techniques to estimate economic inequality using more than 3300 data sets. We also provide guidance on the choice between these two approaches and their estimation, for which we develop the GB2 group R package. Our results indicate that even the simplest parametric models provide reliable estimates of inequality measures. The non-parametric approach, however, fails to represent income distributions accurately.