This strategy looks for anomalies between an index and its constituent names. For this example, I will be concentrating on the ITRAXX indices

http://www.indexco.com/data/IndexMap.asp?menu=data&s=0&opa=505&index=menu8 (registration required)

There are several reasons why the index may not be trading in line with the underlying entities, one being whether the market outlook is bullish or bearish. If the outlook is bullish the skew (difference between actual index spread and average of constituent entities) will shrink (and maybe become negative) as investors look to increase their credit exposure by selling protection.

If the index price appears too high compared to the underlying, an investor can sell protection and sell on the underlying. This will only make sense on a small index (e.g. autos - 10 constituents) to keep transaction costs to a minimum.

Simply taking an average of the spreads is not the most accurate way of arriving at the theoretical index spread for the following reasons

• Dispersion bias

If you take a simple index comprising 2 entities, one trading at 200bps and the other at 300bps. The chances of the 300bps entity defaulting is higher, so the hazard rate must be taken into account. This gives a spread of 249.6 bps.
Note that this isn’t too much of  a problem when the spreads are tight, however if one name were to gap out, the difference would be significant

• Maturity mismatch

CDS contracts roll over quarterly (20th of March, June, September and December) whereas ITRAXX rolls over every six months (20th March and September). A solution to this is to use interpolation using the 3 and 5 year points.

e.g. When a 5 year ITRAXX contract is 3 months old, we would be looking for a CDS contract with 4.75 years. We can either use linear or polynomial interpolation.

• Quotation bias

ITRAXX indices trade at a fixed spread. If someone wishes to buy sell protection on an index, they will have to pay an upfront premium. The upfront value must be converted into a conventional spread.

Some information about upfront calculations can be found here

http://www.noelwatson.com/blog/PermaLink,guid,49d6809c-0c1b-4dc4-8e4e-d15ad9acb209.aspx

These values can add up to as much as 20bp difference from the simple average shown above for the 10Y iTRAXX Crossover index

References:

DJITraxx: Credit at its best

http://www.nuclearphynance.com/User%20Files/464/DJ%20iTraxx-%20Credits%20at%20its%20best.pdf

Attachment