I am looking to take a cross sector subset of a larger stock index universe. What steps to I take to assure that sector representation is as equal as possible to help smooth out my variance(while admittedly giving away some return). Thanks.
You first need to understand how the index values are computed. Is the index market cap weighted, equally weighted,...depending on that you pick a sub set that replicates the properties of the index by weighting the following properties (not an exhaustive list but I hope a starting point):
- large vs small market cap names
- high vs low beta names
- high vs low return volatility names
- sector/group subset representation
- market/country subset representation (in case you are dealing with global indexes)