How do you select which methods (replication, optimisation or stratified sampling) to track the performance of a certain index? 1) Full replication - less than 1,000 stocks in the index - stocks in the index are liquid - manager has more funds to invest 2) Optimisation - increase in the number of stocks in the index - decrease in liquidity - uses a factor model to match the factor exposures of the fund to those of the index (accounts for covariances between risk factors - lower tracking risk than stratified sampling (note that optimisation requires frequent balancing since the optimisation must updated to reflect changes in risk sensitivities) 30 Stratified sampling (aka representative sampling) - increase in the number of stocks in the index - decrease in liquidity - separates the stocks in an index using 2 or more dimensions (i.e. industry, size or PE ratio) - does not require or depend on the use of a model - higher tracking risk than optimisation (hence stratified sampling can be used if tracking risk is not highly important, which can minimise trading costs which could be very high is some emerging countries) Anyone has any other points to add for the creation of an indexed portfolio?
Full Replication leads to lowest tracking error?
joker Wrote: ------------------------------------------------------- > Full Replication leads to lowest tracking error? Yes, I think so !
Yes, correct, full replication should result in minimal tracking risk. Also another advantage is that a full replication portfolio based on a value-weighted (or float-weighted) index is self-rebalancing* (i.e. stock weights in the portfolio will mirror changes in the index weights due to changing market prices). *Self-rebalancing implies that trading is only necessary for: 1) reinvestment of dividends 2) changes in the index composition