We examine the repercussions of a methodological approach to selecting portfolio risk premias, based on the genelarized portfolio models of Markowitz [1952] an Sharpe [1963], in wich we suggest a tiem-varying vairance-covariance matrix of the individual stock returns and modeling then as ARCH (Autoregressive Conditional Heteroscedasticity) models [Engle (1982)]. Moreover, we carry out an empirical application using this methodology. To do this we use the pricing data of the 71 individual scurities with the highest trading volume and frecuency trading in the Spanish Stock Market. The full sample goes form April 10, 1991 to December 31, 1993, inclusive.
JEL classification: C22; G11; G12.