In this paper, the degree of persistence of the sulfur dioxide emissions in a group of 37 OECD countries is examined by looking at the order of integration of the series. However, instead of using integer degrees of differentiation (i.e., 1 in case of unit roots and 0 for stationarity), fractional values are also considered. The results indicate high degrees of persistence and very little evidence of mean reversion. This property only holds for the three Latin American countries examined, namely Chile, Colombia and Mexico if the error follows a white noise process. If autocorrelation is permitted, however, the confidence intervals are wider and mean reversion is not found in any single case. Thus, shocks in the series are expected to be permanent in the majority of the cases examined.