Measuring Vulnerability to Multidimensional Poverty in Latin America

GALLARDO, MAURICIO; SANTOS, MARÍA EMMA; VILLATORO, PABLO; PIZARRO, VICKY

Conferencia; Virtual International Conference of the Western Economic Association; 2021

Western Economic Association

Latin America is not the poorest region in the developing world. It is, however,a region with high inequality, precarious institutional frameworks andhigh exposition to covariate and idiosyncratic shocks. In this paper, witha sample of more than seven million observations, we perform estimates ofvulnerability to multidimensional poverty for 17 Latin American countriesat three points in time: 2005/6, 2012 and 2017. We use a multidimensionalBayesian network classier model to estimate the conditional probabilityof being multidimensionally poor. We then use these probabilities and thestandard downside semi-deviation as the risk parameter to identify the vulnerablehouseholds. Our ndings suggest that, despite signicant reductionsover the study period, in 2017, approximately 200 million people { aboutthe size of the population of Brazil { continued living at high risk of becomingpoor or remaining multidimensionally poor. We also observe thatvulnerability to poverty is reduced at a much slower rate than poverty itself,revealing that poverty reduction accomplishments can actually be quite fragile.Additionally, we perform a decomposition between poverty-induced andrisk-induced vulnerability and nd that as poverty decreases, risk-inducedvulnerability becomes relatively more important than poverty-induced vulnerability.However, it is the poor-vulnerable group that still constitutes thecore vulnerability group.