Benchmark is Russell 2000 for small cap value index. ABC indicates that technique used to construct the new index portfolio assumes that the factors used to explain stock returns are uncorrelated. The most appropriate technique to construct portfolio is 1) Optimization 2) Full replication 3) Stratified Sampling

If tracking error not important, use Stratified Sampling. Otherwise, might need to go with Optimization.

Full replication, no?

since its a value index, which is usually longer-term focused and hence incur less trading costs to maintain this kind of portfolio, stratified sampling may be best.

full replication will cost you an arm and a leg having to deal with 2000 stocks. When its above 1000 stocks, you’re better off stratified sampling or optimizing (if tracking error means a lot to you). However, optimization is costlier.

i have no idea on this one. if whoever nails it can explain it to me… someone needs to review this section.

What is everyone’s understanding of what the question is asking? If the factors are uncorrelated, than you can’t use optimization, which is in a mean-variance framework.

The Russell 2000 small-cap value index contains a large proportion of illiquid stocks. ABC also indicates that they assume that the factors used to explain stock returns are uncorrelated. In such cases the best index construction method is stratified sampling. Mp is right. I was under the impression that if its factor model then it is Optimization but I was wrong.

Hmm, to be honest I overlooked that part. What really sticks out to me is that the index has 2000 stocks, which is too high for full replication. Stratified sampling allows you to mimic this index by splitting the stocks into cells in a matrix and having an portfolio similar to the R2K. It came down to stratified sampling and optimization for me, and according to Schweser, it mentions you decide based on how important tracking risk is for you. But with your input on uncorrelated factors, that should rule out optimization, so you’re left with stratified sampling.

Lol…the funny part is that I got it correct on last year’s mock (question #20 of the multiple choice). I forgot that the number one rule for this shit is “never use full replication”. Optimization (as per my previous post) impossible by removing the correlations, and the answer states that “The Russell 2000 contains a large proportion of illiquid stocks…” Combine that with them “disallowing” optimization by assuming no correlation and you’re left with stratified sampling. I took that fugging test yesterday, FWIW - pretty sweet retention…

It happens skill - don’t go hard on yourself