October 30, 2019
CORE C 035
Inference for monotone single-index conditional means: a Lorenz regression approach
Alexander Jacquemain, UCLouvain ISBA
The tools available to assess how the observed economic inequality is affected by different explanatory variables are quite limited. They are mainly based on decomposition ideas, which happen to be quite restrictive. For example, we can decompose the Gini coefficient of a variable of interest (income) when it can exactly be written as a linear combination of the explanatory variables (income sources). We aim to propose a broader econometric framework to address such issue. We define the explained Gini coefficient as the Gini coefficient we would observe if individuals are ranked by a given linear combination of the explanatory variables rather than income. On that basis, we develop a regression procedure which attributes to each explanatory variable a weight with the aim of maximizing the explained inequality. We show that the underlying econometric model is a special case of the well-known single-index model. Interpretations as well as avenues for applications are presented.
Looking forward to seeing you all in the seminar!