Estimating the Lorenz curve and Gini index with right censored data: a Polya tree approach

Abstract

In this paper we estimate income distributions, Lorenz curves and the related Gini index using a Bayesian nonparametric approach based on Polya tree priors. In particular, we propose an alternative approach for dealing with contaminated observations and extreme income values: avoiding the common practise that removes these critical data, we instead treat them as censored observations and apply a Polya tree model for incomplete data. The proposed method is illustrated through an empirical application based on the European Survey on Income Living Conditions data.