Is the Pareto–Lévy law a good representation of income distributions?

Abstract

Mandelbrot (Int Econ Rev 1:79–106, 1960) proposed using the so-called Pareto–Levy class of distributions as a framework for representing income distributions. We argue in this article that the Pareto–Levy distribution is an interesting candidate for representing income distributions because its parameters are easy to interpret and it satisfies a specific invariance-under-aggregation property. We also demonstrate that the Gini coefficient can be expressed as a simple formula of the parameters of the Pareto–Levy distribution. We subsequently use income data for Norway and seven other OECD countries to fit the Pareto–Levy distribution as well as the Generalized Beta type II (GB2) distribution. The results show that the Pareto–Levy distribution fits the data better than the GB2 distribution for most countries, despite the fact that GB2 distribution has four parameters whereas the Pareto–Levy distribution has only three.