The issue of beta forecasting is explored using Australian stock retur
ns data. A simple market model is fitted to individual stock data over
the period 1983 to 1987 and the beta estimated from this sample is us
ed to forecast the market model beta over the period 1988 to 1992, It
is found that a simple transformation of the initial period beta produ
ces a forecast error which is remarkably close to the transformation w
hich produces the minimum least squares forecast error. This suggests
that the more complex forecasting techniques proposed in the literatur
e may not be worth the additional computational effort.