Moving Beta

[naos]

Hi all,

If I'm given a company having a CDS (say 5Y) and I want to see the dependency between the spread of the CDS and an index (say iTraxx Europe 5Y), what should I compute?

My idea is that I want to be able to say that, on average and during a certain period, when the index moved by 1 bp (or the spread moved by 1%), the CDS' spread also moved by x bp (y %), on average.

It seems that the beta is what I'm looking for. Since I'm not familiar with it, I don't know which beta should I compute : should I look for a beta in (\Delta ( Spread_{CDS}) = \beta \Delta ( Index ) ) or in ( \ln \left \( \frac{Spread_{CDS,t+1}}{Spread_{CDS,t}} \right \) = \beta \ln \left \( \frac{Index_{t+1}}{Index_{t}} \right \) )?
Which one is the most relevant? Reliable?

Lastly, I don't know how to compute moving betas. Indeed, If i'm looking for a daily beta over a certain period (say between 01/01/09 and now), I don't know whether I should take 10,20,30 or more days in order to get a relevant beta for a certain date. What would you choose?

Do not hesitate asking me details or re-explaining my problem since it's probably confuse on some points...

Thank you

Regards

Charles

 

[doug reich]

You choices are valid, just try them out and figure out a metric (e.g. performance of some strategy) to optimize them over.

To compute them, look up the Linest function in excel, if that's where you're working. Or, look up the explicit formula for linear regression coefficients and use that instead.

 

[naos]

Ok, thanks.

I did some computations by considering the % change, log or absolute change and it seems that log is the one which fits the best.

Thank you

Charles