meaningless of beta for bond?

tigerbill · December 14th, 2005, 5:22 am

Beta is usually used to describe the systematic risk of stock or fund, but as to bond, the main risk is interest risk, so do you estimate beta for bond?or do you think it s rational to have such a beta for bond?thanks a lot.

mutley · December 14th, 2005, 6:50 am

I guess you could talk about the way the credit spread moves with the swap/gilt curve?i.e. if the spread is constant, a 10bp move in the swap curve will result in a 10bp move in the absolute credit curve --> beta = 1? if the spread moves in the same direction as the swap curve, beta > 1? if the spread moves against the swap curve, beta < 1?Don't know who'd use such a measure but that'd be my view on a bond's "beta"J

tigerbill · December 14th, 2005, 6:56 am

thanks, mutley, you r thinking the problem in a complecated way.just treat bond as stock, and regress the return of bond on that of benchmarkas stock does to get the alpha, beta of a bond, would this be nonsense?

mutley · December 14th, 2005, 7:43 am

Yeah I guess if you convert it from the bond's yield to its price and then regress that against a bond index it could work.But then again, how will you get around the pull-to-par effect as the bond nears maturity? The bond price volatility will shrink to zero making the beta very much non-stationary. Equity beta is also non-stationary but this reflects different market behaviour rather than a known function of time.Hmmm, not sure if I am confortable with the idea of bond beta but let me know if you come up with something interesting.J

tigerbill · December 15th, 2005, 12:13 am

"The bond price volatility will shrink to zero making the beta very much non-stationary"good point.i once saw a method to calculate beta of a bond, D(i)/D(m), here D is duration.thanks again.

mutley · December 15th, 2005, 7:58 am

Yeah, that method looks much more meaningful than regressing the bond price. Although I'm not a great fan of duration as a measure of risk, it does provide useful information on the sensitivity of the bond wrt. parallel shifts. So your definition of a bond's beta D_i / D_m makes sense (ignoring the fact that D_i is one bond and D_m is a diversified collection of bonds).J

anfieldred · December 16th, 2005, 7:36 pm

QuoteOriginally posted by: mutley Although I'm not a great fan of duration as a measure of riskWhat would you suggest as a more appropriate measure of risk for a bond?with regards beta of a bond relative to an index, you should regress chages in yields as opposed to changes in prices