I was trying to calculated implied excess returns for a portfolio consisting of four indices.The formula I was using was: Implied Returns = lambda (risk aversion coefficient) * covariance matrix of excess returns * asset allocation vector.The problem is that I was working with Bloomberg data and the indices were priced weekly not yearly.An obvious way around this is just to look at movements from year to year. The problem with that approach is that you don't fully use the information available. The method I ended up choosing was to create the matrix of excess weekly returns, calculate the covariance matrix and then simply use (52 * that weekly covariance matrix) in that formula.Is this correct, or is there a better way?Thank you,Paul Epstein

I'm not too sure if your suggested method is correct but I feel more comfortable with the first one that you suggested (compute for year-on-year movements).

We seem to agree 100%. I did solve the problem by computing year-on-year movements. However, my supervisor firmly ("firmly" is a euphemistic adverb here) rejected this approach because this method throws out a lot of the data (in his opinion). For example, if you have only 52 weeks of data (I had 780 weeks), the year-on-year method only yields a sample of one excess-return statistic so you can't really talk about a covariance matrix meaningfully with the year-on-year method. However, in my supervisor's opinion, you can get a meaningful covariance matrix for 52 weeks of data when investigating yearly returns.Anyway, it makes me feel much better about my work, knowing that a Senior Member agrees with me.Paul Epstein