full replication, stratified sampling, optimization ranking

my thoughts requires confirmation:

===if ranked by tracking error:

Full rep (lowest tracking error)

optimization (low tracking error)

stratified sampling (higher tracking errror)

===if ranked by rebalancing frequency:

optimization (need to frequently rebalance)

stratified sampling ( less rebalance)

full rep (atomatically self rebalance, therefore least rebalance frequency by manager)

===if ranked by cost:

full rep (most costly)

optimization and stratified sampling are similar, both are cheap

is that pretty much it?

If you’re ranking worst to best , first one i.e. tracking error is an odd answer. Most institutions and quite a few private clients too want low tracking error. So full rep would be worst . Rest are all worst to best

Full replication would have NO tracking error. But would have highest fees. Typically since After mgmt fees performance would be of concern, this is typically not the route taken, even though tracking risk does not exist.

Optimization / Stratified Sampling - introduces low tracking error, Costs are however much more in control.

Not sure why Optimization would involve frequent rebalancing, and why that would be any different from stratified sampling. Anybody care to elaborate.

in the book it says:

optimization, due to frequent adjustment to risk exposure (to match that of the benchmark) requires frequent rebalance.

got to differentiate costs between rebalancing costs and investment costs…

full replication has lowest rebalancing costs but highest invst costs

if ranking soley on investment costs, i think optimzation is second and strat sampling is third (cheapest)

does full replication always have the least rebalancing costs? if we are indexing to a value/float weighted index, a fully replicated portfolio will be self-rebalancing. But if we are indexed against an equally weighted index, wouldn’t rebalcing costs be quite high?

Also, it seems to be wrong to say that a fully replicated portfolio has ZERO tracking risk, it just has very low tracking risk…


it doesn’t because you also have to re-invest the dividends and there is some slight tracking error created there