The main studies about convergence in Brazil have concentrated in the beta-convergence analysis. The present article reexamines this issue using a methodology developed by Quah (1997) to analyze the evolution over time of the conditioned distribution of income among Brazilian states and municipalities between 1970 and 2000. Three kinds of estimation are done: a) kernel smoothed densities methods; b) estimation of stochastic kernels to analyze the movements of the economies over the income space with the aim to identify convergence clubs in specific space intervals; c) estimation of stochastic kernels conditional on geographic localization, education, openness to international trade and income inequality, with the objective to identify possible explanations to the observed pattern of convergence. The use of the methodologies above allowed us to identify a tendency of club formation in Brazil, as already identified by Mossi et all (2003) and Andrade et all (2004). Variables concerning geographical localization and level of education appeared as the main conditioner for this process.
stochastic kernel; convergence clubs; Brazilian states and municipalities
