Stochastic Convergence Across Brazilian States

Abstract

We analyze the dynamic properties of the relative income of Brazilian states for the 1985-2008 period. Unit root tests suggest that shocks to relative income have permanent effects, contradicting the stochastic convergence hypothesis. On the other hand, interval estimates of the largest autoregressive root produce wide confidence intervals that include many alternatives consistent with stationarity. Additionally, the confidence interval estimate of the half-life suggests that relative income shocks die out relatively quick, within an average of 0 to 5 years. These results suggest that relative income shocks have a temporary effect, thus supporting the stochastic convergence hypothesis. Furthermore, we build a relative income series for the Brazilian states for the 1947-2008 period and

redo the exercises above. Finally, all of our results remain the same when we extend the sample to the 1947-2008 period.