Title: Long-range correlations and nonstationarity in the Brazilian stock market

Abstract: We report an empirical study of the Ibovespa index of the Sao Paulo Stock Exchange in which we detect the existence of long-range correlations. To analyze our data we introduce a rescaled variant of the usual Detrended Fluctuation Analysis that allows us to obtain the Hurst exponent through a one-parameter fitting. We also compute a time-dependent Hurst exponent H(t) using three-year moving time windows. In particular, we find that before the launch of the Collor Plan in 1990 the curve H(t) remains, in general, well above 1/2, while afterwards it stays close to 1/2. We thus argue that the structural reforms set off by the Collor Plan has lead to a more efficient stock market in Brazil. We also suggest that the time dependence of the Ibovespa Hurst exponent could be described in terms of a multifractional Brownian motion.