Multidimensional Poverty Measures from an Information Theory Perspective

This paper proposes to use an information theory approach to the design of multidimensional poverty indices. Traditional monetary approaches to poverty rely on the strong assumption that all relevant attributes of well-being are perfectly substitutable. Based on the idea of the essentiality of some attributes, scholars have recently suggested multidimensional poverty indices where the existence of a trade-off between attributes is relevant only for individuals who are below a poverty threshold in all of them (Bourguignon and Chakravarty 2003, Tsui 2002). The present paper proposes a method which encompasses both approaches and, moreover, it opens the door to an intermediate position which allows, to a certain extent, for substitution of attributes even in the case in which one or more (but not all) dimensions are above the set threshold. An application using individual well-being data from Indonesian households in 2000 is presented in order to compare the results under the different approaches.