LOS 26: Black-Litterman vs. Resampled Efficient Frontier

Hi everyone, I’m currently reading LOS 26 but I’m having some trouble differentiating Black-Litterman and Resampled Efficient Frontier. Would someone kindly offer their insight on each of them and the differences between them please? Thank you. GreenTomato

In mean-variance optimization we try to figure out the mean returns and risks of each assets by using a sample return-risk model. Now the drawback is that the returns from this model may not be correct at all. (as it is from a sample) We are trying to solve for how much money to allocate to different assets by using return and risks of each assets. So we do this using the below two methods: Black-Litterman: In black-litterman, we already assume that the initial expected returns are whatever is required so that the equilibrium asset allocation is equal to what we observe in the markets. We just state our different assumptions about the returns and how confident we are of those assumptions. Then we invest money in those assets. Resampled efficient frontier: To account for uncertainty about returns from one sample, we create different efficient frontiers based on different samples. We then aggregate them into a re-sampled efficient frontier which is an average of all the different samples. Thus the average mean and variance of different assets is used to invest money in.

They are both similar in their objective , to get stable representative weights for different asset classes in the face of uncertainties in the returns estimation. Resampled eff. frontier tries to do this by running simulations or trials using small variations in expected returns and getting more robust estimates at each point along the frontier. Process involves averaging weights across trials.( I think this is a mechanical process , doesn’t involve your views as an investor) B-L does this by taking current asset weights as given and using covariances of returns ( historical ) , estimating the expected returns, then imposing views of either absolute return estimates or relative expected returns , then re-estimate the asset weights ( what should be the weights for small divergences from current weights) Resampled is a patented , proprietary method , but it is methodological , not based on much research , at the same time gives stable , realistic asset weights . b-l is based on theory and there is research to back it. Nevertheless there is some criticism because covariances can be finicky and hard to estimate as well.

Resampled method is based on bootstraping idea from statistics, BL is based on combination of equilibrium approach and shrinkage estimators.

Check this out http://datalab.morningstar.com/knowledgebase/aspx/Article.aspx?ID=217&Country=us

janarski knocked it out the park