On Measuring Income Polarization: An Approach Based on Regression Trees

Abstract

This article proposes the application of regression trees for analysing income polarization. Using an approach to polarization based on the analysis of variance, we show that regression trees can uncover groups of homogeneous income receivers in a data-driven way. The regression tree can deal with nonlinear relationships between income and the characteristics of income receivers, and it can detect which characteristics and their interactions actually play a role in explaining income polarization. For these features, the regression tree is a flexible statistical tool to explore whether income receivers concentrate around local poles. An application to Italian individual income data shows an interesting partition of income receivers.

On Measuring Polarization for Ordinal Data: An Approach Based on the Decomposition of the Leti Index

“…In this paper, we use classification and regression trees (CART) (Breiman et al, 1984) for uncovering polarization patterns when dealing with ordinal data. The use of regression trees to explore polarization in income distribution has been recently investigated by Mussini (2016). Classification trees for ordinal variables (Piccarreta, 2008) are used to handle ordinal data.…”

Analysis on Relationship between Pure Income Polarization and Economic Changes

“…The principal impact of the workforce position is greater essential than each of the other indicators in assessing dissimilarities in earning. High-educated labors have opportunities of growing their wage throughout their job impacts by age [4]. [5] displayed in assessing earning bipolarization, investigators regard two notions, or axioms.…”