This paper investigates the degree of association in the identification of the poor between the standard monetary FGT measure and the Alkire-Foster Multidimensional Poverty Index. For this purpose, we use a measure of redundancy between the two poverty measures (R0). In Chile, over the past 25 years, R0 has declined at a rate of 1.5% per year. The decline is unimportant during the 1990s, a decade of rapid economic growth, while it is notable thereafter, in a period characterized by modest economic growth and the progressive introduction and deepening of social policies. The conditional correlation between socioeconomic and demographic characteristics with R0 is examined at the province and household levels. After controlling for household non-eligibility across some of the indicators of the Multidimensional Poverty Index, we find that the divergence in the identification of the poor can be explained by improvements in education, increasing urbanization, and a reduction in the household size. Consequently, the divergent identification of the poor seems to be a real process, which is not randomly distributed across the population. On the basis of our results, we hypothesize that this divergence is a general phenomenon that tends to occur in countries undergoing demographic transition, urbanization, and progress in education. If so, and given the fact that poverty alleviation strategies are adopted partly on the basis of poverty statistics, the diverging identification of the poor might have distributive consequences for the poor themselves.
Citation: Klasen, S. and Villalobos, C. (2019). ‘Diverging identification of the poor: A non-random process. Chile 1992–2017’, OPHI Working Paper 128, University of Oxford.