A switching regression approach to the stationarity of systematic and non-systematic risks: the Hong Kong experience

The switching regression method of Goldfeld and Quandt (Technique for estimating switching regressions, in Studies in Nonlinear Regression, ed. S. M. Goldfeld and R. E. Quandt, Ballinger, Cambridge MA, 1976) is used to examine the stationarity of systematic and non-systematic risks of Hong Kong's common stocks via the stability of the market model parameters for six industry portfolios and four family portfolios. Empirical evidence for the industry portfolios suggests that the systematic risk component is fairly stable throughout the sample period. However, non-systematic risk tends to decline over the 13-year horizon from February 1980 to December 1992. This may also imply a reduction in the industry's unique risk proportion relative to its total risk level. Hence, security analysts may as well direct relatively more efforts and resources towards analysing the overall market performance rather than focusing extensively on individual industry. Similar findings emerge concerning the non-systematic component of the family portfolios. The evidence of a reduction in industry as well as family-specific risk may further suggest that the benefits of diversifying across different industry sectors or across different family groups may have been diminishing over the past decade. However, the shifts in the structure of systematic and nonsystematic risks of the family portfolios do not appear to have drastically affected the pay-off from analysing and monitoring stocks of individual families.