A REGRESSION-BASED SHAPLEY DECOMPOSITION FOR INEQUALITY MEASURES

Kévin Fourrey (1, 2, 3)

1 CEET - Centre d'études de l'emploi et du travail
2 ERUDITE - Equipe de Recherche sur l’Utilisation des Données Individuelles en lien avec la Théorie Economique
3 TEPP - Théorie et évaluation des politiques publiques

This article proposes an innovative tool to decompose the inequality of an outcome distribution between a set of attributes contributing to that inequality, based on econometric models. We use the recent developments of the Shapley decomposition by Chantreuil et al. (2019) that we apply more broadly to a case where the outcome distribution by sources is not predefined by a natural additive structure. We show that this development is highly relevant in applied studies and that it has a number of advantages compared to the existing regression-based decompositions of inequality measures. An example of application to wage inequalities in France is given, with a focus on the attribute of gender.

Keywords: Decomposition of Inequality Measures; Distributive Interactions; Shapley Value; Gender Inequality

Domains: Humanities and Social Sciences; Economics and Finance; Sociology

Kévin Fourrey. A REGRESSION-BASED SHAPLEY DECOMPOSITION FOR INEQUALITY MEASURES. Annals of Economics and Statistics, 2023, 149, pp.39-62. ⟨10.2307/48718079⟩. ⟨hal-04083219⟩