SEMI-PARAMETRIC ESTIMATION OF INEQUALITY MEASURES

摘要

Measures ot inequality are very popular in economics, especially in measuring the inequality in income or wealth within a population and between populations. In this paper we propose semi-parametric estimators of measures of inequality, building on previous proposals for such estimators. Our proposals make explicit use of results from extreme value theory and are applied to the Gini coefficient, the generalised entropy and the quintile share ratio measures. The properties of these estimators are studied through a simulation, including their mean squared errors, sensitivity to outliers and confidence intervals based on them. They are found to perform very well compared to the traditional nonparametric estimators. The results are applied to an income data set. Some areas of future research are indicated.