Identificación de tipologı́a de pobreza multidimensional a través del enfoque de cluster probabilı́stico (Identification of typology of multidimensional poverty through the probabilistic cluster approach)

The measurement and characterization of poverty is a topic of social interest that is statistically it can be approached from different perspectives. By using using Factorial Analysis and/or Cluster Analysis techniques, partitions were defined households with different degrees of vulnerability. The objective of the research is the multidimensional measurement of poverty and construction tion of typologies of households (poor, vulnerable and non-poor) from data surveyed by the Encuesta Continua de Hogares -ECH (Continuous Household Survey) of the Instituto Nacional de Estadı́stica de Uruguay INE (National Institute of Statistics of Uruguay). between 2008 and 2012. This document presents the results corresponding to a clustering process, based on a probabilistic mixture of distributions Bernoulli butions. It assumes the presence of an unobservable categorical variable that dictates the membership of individuals to groups. The parameters are estimated by means of the EM algorithm and the individuals are classified using probabilities "to later" The variables considered in the analysis refer both to characteristics of the dwelling (ceiling, wall and floor materials), as well as characteristics of the home and its great. Regarding the home, the following are taken into account: its composition, the educational climate, the number of children under 6 years of age, possession and access to information technologies and communication (TICS), as well as the possession of comfort goods. In regards to people, the employment status of the boss, income and ethnicity are considered. The results indicate the existence of 3 or 4 groups that determine a gradient of vulnerability, from the most critical households (poor most vulnerable), to those are in better conditions (less vulnerable).