19. A stratified sampling approach is “blind” to the relationships among cells. This can be remedied by complementing a stratified sampling approach with the use of an optimizer that accounts for the correlations among cells.

20. The sector breakdown is: Banking, Finance, Basic Industry, Consumer Cyclical, Consumer NonCyclical, Communications, Energy, and Utility.

21. The number of bonds selected from each cell is set to 10 since it strikes a good balance between having a realistic size for the replicating portfolio (240 bonds) and reducing idiosyncratic risk. If fewer bonds are used to represent each cell, the variation in tracking errors may reflect not only the difference between the two systematic risk measures (DTS and spread duration), but also the idiosyncratic performance of the set of bonds selected.

22. The scheme used to weight the 10 bonds within each cell is fairly complex, involving a further subdivision of each cell into four quadrants, to ensure that we can match both market weight and the desired additional characteristic by a rule-based algorithm that always results in positive weights to all selected bonds. Details may be found in Ambastha and Ben Dor, “DTS (Duration Times Spread) in the Credit Crunch: Did It Live Up to Expectations?”