A number of studies using a variety of techniques over a range of different economies and time periods have found evidence of time instability in both individual common stock and porfolio betas. Finance theory would suggest that as individual stocks are aggregated into portfolios, a diversification effect should produce a beta for the portfolio that is relatively more stable, provided that beta instability in the individual stocks is driven, at least in part, by microeconomic factors. Despite this Collins et al. (1987) for US data, and Faff et al. (1992) and Brooks et al. (1992) for Australian data, found a higher degree of beta instability in portfolios formed from randomly chosen stocks, relative to the beta instability found in individual stocks. Collins et al. (1987) attributed this finding to a reduction in background noise through portfolio formation making beta instability easier to detect, and hence, dominating the diversification effect. It is also possible that the increased beta instability could be due to the instability in the individual stocks being driven by macroeconomic factors. This paper investigates the effect of portfolio formation on beta stability by considering a number of portfolios formed from individual stocks with particular beta stability characteristics. In particular, we wish to assess the empirical validity of the competing arguments associated with the different effects. The methodology used is similar to that of Brooks et al. (1992), in that, all testing is conducted using point optimal tests (see King (1987a)).