They both seem very similar. What’s the difference between them?
Aim is the same but optimization uses computer models.
Optimization uses a factor model where factor exposures are matched to the index. Stratified sampling you basically separate stocks by a number of characteristics or ‘strata’ and then choose random stocks to represent each characteristic. The more characteristocs there are the less tracking error.
With optimization, the model accounts for covariances between factors; stratified sampling does not, and so theoretically, optimization could have lower tracking risk than stratified sampling. But optimization can require more frequent rebalancing when factors get out of whack, so could be more expensive.
I recall that according to one of the sources I studied, both tend to yield similar results, despite everything I just said.
Thank you!!!
The way I understand it, the question seems to point you to a certain direction.
I think it was in the 2015 pm mock exam, they wanted to replicate the Russel 2000, but cost was an issue.
Full replication is ruled out because its only feasable for approximately 1000 stocks or under, and Optimization was out because it required more frequent rebalancing, hence also affecting the cost.
My general rule of thumb is if cost is a factor, or the individual in question doesn’t want to use computer modeling and can’t fully replicate, then stratified is the answer.