This article discusses a controversy between tests of convergence of per-capita income based on time series analyses and cross-section regressions. The cross-section analyses are based on the hypothesis of beta convergence, while the second one is equivalent to sigma-convergence analysis. Here, it is argued that this conflict is due to a possible non-linearity in the convergence process associated with formation ofconvergence clubs. Tests of the convergence hypothesis based on nonlinear models of Hansen (2000) and Phillips e Sul (2007) are conducted for the Brazilian case. The first test is based on a cross-section regression, while the second in time series analysis. Both tests point to the formation of two convergence clubs alike. For each club is noted evidence of both beta and sigma convergence, and the controversy of results previously reported is resolved.
