Estimation of income inequality from grouped data

Grouped data in the form of income shares have conventionally been used to estimate income inequality due to the lack of individual records. We provide guidance on the choice between parametric and nonparametric methods and its estimation, for which we develop the GB2group R package. We present a systematic evaluation of the performance of parametric distributions to estimate economic inequality. The accuracy of these estimates is compared with those obtained by nonparametric techniques in more than 5000 datasets. Our results indicate that even the simplest parametric models provide reliable estimates of inequality measures. The nonparametric approach, however, fails to represent income distributions accurately.