Decomposição da pobreza no Brasil usando um modelo de fatores dinâmicos.

Abstract

This paper investigates the common dynamic properties of poverty rates across Brazilian states during the period 1976-2012. Aiming to investigate whether movements in poverty levels were explained with greater emphasis by 'shocks' originated from influences at the national level (Macroeconomic Policies, e.g.), or by changes at the local/regional level (educational structure, health conditions, labor market, etc.). Therefore, it was utilized the Bayesian dynamic latent factor model approach, proposed by Kose, Otrok and Whiteman (2003, American Economic Review), allowing the decomposition of poverty into national, regional and state specific component factors. The results showed that, on average, the national factor was responsible for explaining approximately three quarters of the volatility in the poverty rate of the Brazilian states. Additionally, it was found that importance of the national factor grew, to the detriment of idiosyncratic factor, since 1995. This result emphasizes, in some way, the great importance of controlling hyperinflation and increased social spending by the federal government to change poverty rates in Brazil. However, the importance of regional and local components differed substantially among states. Variations in educational attainment and labor market dynamics may explain these regional differences.