International portfolio selection on European stock markets based on time-varying betas
José Soares da Fonseca
The research presented in this article estimated three asset pricing models, applied to sixteen European stock markets, to create efficient portfolios, based on the solution originally proposed by Elton, Gruber and Padberg (1976, 1978). This solution overcomes the need to solve a mathematical programming problem, as required by the mean-variance approach proposed by Markowitz (1952). The innovative approach proposed consists of extending the original method of Elton et al. to include the market timing and lower partial moments. In the empirical analysis of this article three asset pricing models were estimated and their parameters and risk measures were used to create beta-based efficient portfolios. The first asset pricing model is the standard CAPM, the second model adds an additional variable which represents market timing, and the third model includes two additional variables, one representing market timing, and the other representing lower partial moments. The asset pricing models were estimated by rolling regressions applied to moving samples. Three types of efficient portfolios were constructed and evaluated on a daily basis, each type being related to one of the three asset pricing models estimated. The efficient portfolios were recomposed daily as they benefited from the new parameters that were given by the most recent rolling estimation. The results showed that the beta-based efficient portfolios dominated the equally weighted portfolios and the investment in the European index, which served as benchmarks.
Keywords: Efficient portfolios, Lower partial moments, Market timing, Time-varying betas.
JEL classification: F36, G11, G15.