This paper explores whether predictable autocorrelation structures exist in returns data on Australian stock indices. We explore the data using the power transformations of Ding, Granger and Engle (1993) and Hentschel (1995). We find that for a large number of different market indices there exists an autocorrelation structure in power transformations of returns that could be exploited in forecasting. To illustrate the forecasting potential we conduct a limited forecasting comparison using the All Ordinaries Accumulation Index. Our results show some gain in forecasting performance.