Purpose This paper aims to investigate the production of sugar cane ethanol in Brazil for the time period 1983-2016, separating the data by geographical location. Design/methodology/approach For this purpose, the authors use techniques based on the concept of fractional integration. Findings The authors show that the data corresponding to the total production is highly persistent, with an integration order smaller than 1 but close to it. In fact, the unit root hypothesis cannot be rejected implying that shocks have a permanent nature, and thus requiring policy measures to recover the level from exogenous shocks. Separating the data into two sub-regions, namely, North-Northeast and Central-South, higher levels of persistence are detected in the latter, while the former presents some evidence of mean reverting behavior, implying that shocks will disappear by themselves in the long run in the former regions. These results are obtained from all the different methods used. Originality/value The originality is based on the time series techniques used in the paper that departs from the classical methods based on unit roots and integer degrees of differentiation.