An attempt is made to establish empirically the required number of lagged and leading terms to include in Dimson's (1979) Aggregated Coefficients Method (ACM) of beta estimation. It is shown that the technique used by Dimson (1979) for the same purpose is sample specific in a number of important respects. A simple evaluation of the significance of the obtained coefficients consistently indicates that at least one, but possibly two, lagged terms are required for ACM beta estimation using monthly return data for a sample of Australian securities.
