Dos and Don'ts of Gini Decompositions

This paper critically assesses widely applied methods of Gini decomposition by income sources and population subgroups. We point to common pitfalls in the interpretation of decomposition results and show that marginal efects provide the only meaningful way to examine the relevance of income sources or population subgroups for total income inequality. Moreover, we show that existing methods are unsuitable to decompose the trend in the Gini coeficient, i.e. to examine the role of income sources or population subgroups for the change in the Gini over time. We provide a coherent method to decompose the Gini trend by income sources. Creation Date: 201311