Can anyone please explain how this really works in reality and why it results in a higher degree of diversification then the mean-variance approach? Are you basically looking at the global index, determine what asset categories/securities it is made up of and the respective weightings. Then assuming efficient markets, you back out the respective market premiums for the various asset categories/securities, and as such, you obtain the implied return. Then, using this return, you setup your own portfolio using the implied return as your expected return and whatever weightings you determine to be appropriate? However, if you setup your own portofolio as described above, why would it be more diversified than a mean-variance approach? The Black-Litterman approach is described in Schweser Book 3, Study Session 8, page 54. I’m sorry but Im not in portfolio management, therefore, I have sometime difficulties understanding of how all the theory in the Level 3 curriculum actually is applied in real life. Thanks
I am not very clear either. By using a global portfolio though, you would include more assets so greater diversification I guess.
Try a search on this. I think maratikus had a good explanation awhile ago.
Crap, couldnt find it, maratikus has been quite active here… Searching for Litterman only generated 8 results 
Basically, the optimal portfolio that comes out of a Black-Litterman model is a linear combination (basically a weighted average) of two portfolios: 1) The Market Portfolio (M), and 2) A portfolio that is optimal, given your specific insights about performance in the next investment period, and your confidence in those insights. If you are a fundamental manager, you may only have a few securities or opinions, or those opinions might be in only one sector or type of investment. In addition, small differences in expected returns or correlations that come from the fact that the inputs are estimated rather than known can lead to vastly different portfolios for only slight differences in expected return or risk. These portfolios can also be highly concentrate. So the B-L portfolio is not as concentrated because it is basically mixed with the market portfolio in such a way that it maximizes: [market expected return + alpha from views] / sqrt[market variance + active variance] The market portfolio by definition diversified, which is easy to remember because the market portfolio has a little bit of everything in it. As a result of the mixture, the net B-L portfolio isn’t as concentrated. It turns out that the Treynor-Black portfolio model just boils down to a special case of the Black-Litterman portfolio process, so that’s why you hear about B-L much more than T-B in practice.
i think ^ and florinpop got it.
bchadwick, youre awesome! Thanks
In Black Litterman approach, global portfolio comprise of various sub indices, say their weights (w1,w2,w3…) (n x 1) vector, Variance-covariance matrix [Sigma] (n x n) and Expected mean excess returns [Mu] (n x 1) vector - accordingly model spits out ideal weights for the indices in Investor portfolio according to expected risk aversion and estimates of excess returns. Formula goes like this: [Weights] = (lambda)* [Mu] * [Sigma]^ -1 where Lambda is desired risk aversion. As these indices are not concentrated positions in a single or a bunch of stocks, resultant portfolio also is also free from such bias. In Markowitz world, one can possibly receive a concentrated allocation as a result of substantial allocation to a corner portfolio, this purely depends on the available investment opportunity and its Mean-Variance character. Hope this helps.
I thought the black-litterman approach, in a very boiled down view, is a way to create an efficient frontier without having to estimate Std dev, correlations, and expected returns since the information is “extracted” from the current market’s expectations of these variables, so it takes out the work and uncertainty from solving for these estimates for thousands of securities.
PhillyBanker, I think that’s how you’d answer it on the exam because you’ll chew up too much time on the more technical explanation. Plus it’s right. Also, CFAI is not looking for the level of understanding in the other explanations. If you’re a quant (and write really fast), go for it, but that’s not the point of level 3 from prior exams I’ve reviewed.
Methinks the LOS says that we know the advantage and disadvantage of the approach. thats all I read into the B-L approach
Thanks Chadwick. A question though – Was this thing written in CFAI/Schweser or is this what you’ve read somewhere else?
PhillyBanker Wrote: ------------------------------------------------------- > I thought the black-litterman approach, in a very > boiled down view, is a way to create an efficient > frontier without having to estimate Std dev, > correlations, and expected returns since the > information is “extracted” from the current > market’s expectations of these variables, so it > takes out the work and uncertainty from solving > for these estimates for thousands of securities. You are correct about expected returns “extracted” from the market weights (which is most important since optimization results are most sensitive to expected returns). However, covariance matrix is estimated using historical data.
Ok - thanks sterling & mara.