I don’t really understand this concept. And I especially don’t understand the difference between the regular BL model and the unconstrained model. Anyone care to take a crack at explaining it to me?
The model starts with the equilibrium assumption that the asset allocation of a representative agent should be PROPORTIONAL TO THE MARKET VALUES of the available assets, and then modifies that to take into account the ‘views’ (i.e. the specific opinions about asset returns) of the investor in question to arrive at a bespoke asset allocation.
Everytime I look at this model I dont really get it either. It seems like there should be no way Black and Litterman should have got in a textbook for simply saying, start with the weights of each asset class, and then adjust slightly based on your own views. I am reading wondering how they come up with this. Unconstrained allows short selling of some assets to buy more of another asset class, just like when you are doing inconcstrained corner potfolios. So i think you are probably overanalyzing this model. Here are what i took away are the key points from the book: -BL is a top down approach -uses returns implied by the value weighted global market index -alter it slightly based on analyst opinions (like tactical asset allocation) -It will result in a well diversified portfolio, and will avoid the input bias from E® -Disadvantage is you must use historical volatility. -If you want to make your portfolio less risky but have no views, combine world portfolio with risk free asset.
Yeah, if I recall correctly one of the Schweser instructors said that they love to test Black Litterman, but I’m not really sure what makes it so special. I guess we can just memorize what CFADreams wrote but does anyone know how this has been tested before?
Take a look at the 2007 exam–there is a question about BL on there…
Aimee Wrote: ------------------------------------------------------- > Yeah, if I recall correctly one of the Schweser > instructors said that they love to test Black > Litterman, but I’m not really sure what makes it > so special. I guess we can just memorize what > CFADreams wrote but does anyone know how this has > been tested before? It will be tested just because memorization is required.
So is the fact that the unconstrained model allows short selling the only difference between the two?
I wouldnt just memorize what i wrote by any means… its really just a 5-7 min read if you commit the time to it to understand the concept… Vol 3 pg 274-282 (lots of pictures)… There is also like a 5 step process on running the model, but i doubt they wil lack you to run a mean-variance optimization on the test. They really only focused on the BL model, not unconstrained BL. But as far as I can tell, short selling seems to be the only difference in the actual model.
What about re-samplified efficient frontier ? Can anyone summarize the key points ? TKVM !
AMC Wrote: ------------------------------------------------------- > What about re-samplified efficient frontier ? Can > anyone summarize the key points ? TKVM ! The only thing I remember about re-sampling is that it uses computer to run mean variance optimization to generate many efficient portfolios (as opposed to just one under in mean variance analysis) and given many runs it somehow produces more accurate (?) efficient portfolio. It also suffers from input bias.
^ it produces a more diversified portfolio since you keep re-running the optimization, but the disadvantages are that it suffers from input bias and there is no real rationale for doing this method.
So both Black Litterman & re-samplified efficient frontier are more diversified and suffers from input bias ?
I agree that the text says that re-sampled efficient frontier has no theoretical base. However, bootstrapping does have nice asymptotic properties. http://en.wikipedia.org/wiki/Bootstrapping_(statistics)
Sorry guys don’t have my books with me, and the definition of input bias is…?
creperie Wrote: ------------------------------------------------------- > Sorry guys don’t have my books with me, and the > definition of input bias is…? I assume people are just talking about sensitivity to inputs. Under certain conditions (don’t want to get into detail unless people are interested) even a small change of expected excess returns can substatially change optimal weights.
AMC Wrote: ------------------------------------------------------- > So both Black Litterman & re-samplified efficient > frontier are more diversified and suffers from > input bias ? I don’t think BL model suffers from Input bias?
no input bias. You only suppy it with weights and covariances and then it spits out an expected return. Then you use this return along with your own forecasts and run the traditional mean variance. Perhaps your own forecasts can be considered input bias? I’m not sure.
no input bias. You only supply it with weights and covariances and then it spits out an expected return. Then you use this return along with your own forecasts and run the traditional mean variance. Perhaps your own forecasts can be considered input bias? I’m not sure.
Read this part in CFAI3, p274 again. It says “We call this unconstrained Black-Litterman model, or UBL model, becasue the procedure does not allow non-negative constraints on the asset class weights.” Doesn’t it sound like “short” is not allowed in UBL? Can anyone explain it? Thanks.
double negative… does not allow non-negative constraints… i think constraints is the key word. It does not allow for any non-negative “constraint”. A non-negative constraint would be a constraint against short selling. so it does not allow (does not have) a constraint against shorting. their wording sucked here, i had to read it like 19 times…