Forums  > Basics  > Black-Litterman and Benchmarks  
Page 1 of 1
Display using:  

Its Grisha

Total Posts: 48
Joined: Nov 2019
Posted: 2019-11-21 16:37
Hi all, long time lurker first time poster. Have a conceptual question on BL:

The market equilibrium return vector is derived from market cap weights. To me, the elegance of the model comes from the fact that in the absence of views, you revert to the prior and hold the market cap weights after MVO. This reconciles with tracking error to a benchmark IF the benchmark weights are the market weights used for the prior. Then, in the absence of views, we have tracking error 0, which works out quite nicely.

How does this work if we want to hold securities not in the benchmark? Then we have tracking error by default, even when we don't have any views.

Specifically, I am looking to understand the case of a long only strategy benchmarked to R1000 Value, where I want to hold names not in R1000 Value

Do I simply impose a constraint on TE to benchmark in final optimization after deriving the Black-Litterman return vector using a broader universe? And then this TE to benchmark constraint is completely separate from the lambda risk penalty (calculated based on broader universe)?

EDIT: I think the solution is to use a market weight vector with zero's for the names not included in benchmark. They still receive a reasonable equilibrium return. Would appreciate if someone can confirm this is the way to go.
Previous Thread :: Next Thread 
Page 1 of 1