Mixing it up

Abstract

We offer a brief review of the use of distributional mixture models with a finite number of components for the study of the distribution of income. In general, finite mixture models find a number of applications across fields, but they usually arise from theoretical considerations. Applications to the distribution of income present a joint inference about the number and types of components to include in a mixture, corresponding to how different income generating mechanisms’ statistical signatures are represented in the observed data. Many of the contributions in this area rest on an implicit (and sometimes explicit) information theoretic approach to this inference problem. Our review concludes with new illustrative findings from the US based on restricted-access Census data.