This paper points to flaws in Gini decompositions by income sources and population subgroups and to common pitfalls in the interpretation of decomposition results, focusing on methods within the framework of Rao (1969). We argue that within this framework Gini elasticities may 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 coefficient and provide a coherent method to decompose the Gini trend by income sources. We add to the recent trend of multi-decompositions by deriving Gini elasticities from a simultaneous decomposition by income sources and population subgroups.
