Full Replication, Stratefied Sampling, Optimization

Who’s ready to discuss the differences? (12 points)

Optimization is best, full replication is expensive

Full Replication - Lowest Tracking Risk of the 3 - Buy every bond in the index - Automatically rebalanced except for Dividends and changes in stocks on the index - Higher Costs since you have to buy each bond - Not good for markets where illiquidity is an issue or where there are many bonds Stratified Sampling - Breaks up the index into cells where each cell has a specific set of risk factors and a sample of bonds is bought within each cell - the more cells, the lower the tracking risk - has higher tracking error but good for indices with more than 1,000 securities or for illiquid markets Optimization - Based on a factor model that identifies specific risk factors that try to replicate that of the index. - Lower tracking risk than Stratified sampling, but if the inputs are improperly defined at the beginning, can lead to misleading results - Also have to rebalance on a regular basis which is costly

This must be for equity indexing Full Replication: ideal for less than 1000 stock, liquid. Costly, low tracking risk. Stratefied Sampling: Like cell matching, break stock into sector, finer the cell, better indexing, can help to avoid illquide stock. Once the cell is defined, choose the representative stocks. Doesn’t consider corrleation/covarian of stocks in the index. Optimizaito: a mutli-factor model, identified the risk factor of the index, then find stock to match the risk factor, does consider the correlation/covariance of the stocks. Lower tracking risk compare to stratefied method. May I get full credit for this??

Yes - if I could give out CFAs, I would give you one. ws Wrote: ------------------------------------------------------- > This must be for equity indexing > > Full Replication: ideal for less than 1000 stock, > liquid. Costly, low tracking risk. > > Stratefied Sampling: Like cell matching, break > stock into sector, finer the cell, better > indexing, can help to avoid illquide stock. Once > the cell is defined, choose the representative > stocks. Doesn’t consider corrleation/covarian of > stocks in the index. > > Optimizaito: a mutli-factor model, identified the > risk factor of the index, then find stock to match > the risk factor, does consider the > correlation/covariance of the stocks. Lower > tracking risk compare to stratefied method. > > May I get full credit for this??

ws Wrote: ------------------------------------------------------- > This must be for equity indexing > > Full Replication: ideal for less than 1000 stock, > liquid. Costly, low tracking risk. > > Stratefied Sampling: Like cell matching, break > stock into sector, finer the cell, better > indexing, can help to avoid illquide stock. Once > the cell is defined, choose the representative > stocks. Doesn’t consider corrleation/covarian of > stocks in the index. > > Optimizaito: a mutli-factor model, identified the > risk factor of the index, then find stock to match > the risk factor, does consider the > correlation/covariance of the stocks. Lower > tracking risk compare to stratefied method. > > May I get full credit for this?? " Lower tracking risk compare to stratefied method" Please point me to CFAI page where it says that. I’ve looked all over and couldnt find it

I am printing this out and taking this to my local CFA society and tell them that “Pimp said that I can get my charter.”

In addition, Optimization will lead to more rebalancing since the program will look for risk optimization and will require you to rebalance it to make it optimal.

Sorry, CSK, I read it a while back…remember it says that somewhere in the CFA text…I will look again.

ws Wrote: ------------------------------------------------------- > Sorry, CSK, I read it a while back…remember it > says that somewhere in the CFA text…I will look > again. actually it is from Schweser. I know it is i saw it there, but i couldnt find it in CFAI. On exam i will go with correlation stuff. I really hope they ask only for 1 reason

yes it is true that optimization will lead to lower trackign error than stratified sampling since optimization looks into var/cov mix

^Knowing CSK, he needs explicit wording from CFAI text.

ws Wrote: ------------------------------------------------------- > ^Knowing CSK, he needs explicit wording from CFAI > text. :slight_smile: it was just bothering me. At this point schweser will work too %)

krishna1 Wrote: ------------------------------------------------------- > yes it is true that optimization will lead to > lower trackign error than stratified sampling > since optimization looks into var/cov mix oh, I think it’s actually because of more frequent rebalancing with optimization?

CFAI in 2023 no longer puts out 1,000 as the cutoff between full replication and SS/Optimization , but a Wiley mock exam Q still ask this. What’d I do without AF.

This post made me smile. I remember when I was studying for this, I thought this would be the easiest of sections and then I got confused here as well. Nothing for granted at L3.