In the Longstaff and Schwartz (1995) model an Ordinary Least Squares (OLS)
regression analysis is applied to both absolute yield spreads (the difference between the risky
and riskless yields) and relative yield spreads (the ratio of the risky to riskless bond). Using
Moody’s corporate bond yield indices, they find a significant negative yield spread - riskless
rate relationship for absolute spreads and a stronger negative relationship for relative spreads.
Longstaff and Schwartz conclude that this empirical result supports their two-factor model
corporate bond-pricing model. 3 However, Duffee (1998) points out the failings of this
approach owing to the construction of Moody’s index, which uses both callable and
noncallable bonds. For callable bonds, higher interest rates imply a lower chance that the
issuer will exercise the call option. Thus bondholders will demand a lower yield for this call
provision, which will result in an overall decrease in the bond yield spread. To accommodate
the call features present in most U.S. corporate bonds he constructs a noncallable bond index
and regresses spread changes on changes both in the short yield and in a term structure slope
variable, for both callable and noncallable bonds. Interestingly, he finds that the negative
relationship between credit spread and interest rate is much weaker once the call option
effects are removed from the data. Specifically, changes in yield spread on callable bonds are
found to be strongly negatively related to changes in Treasury yields, while on the other hand,
for noncallable bonds, he finds a weak, although still negative relation. Therefore the
negative relationship found by Longstaff and Schwartz (1995) can be attributed to either the
default premium or call premium.
In a related study using Australian Eurobond, Batten, Hogan and Jacoby (2005) find
results consistent with the implications of the Longstaff and Schwartz (1995) theoretical
model, with actual and relative credit spreads both negatively related to changes in the All
Ordinaries Index (a proxy for the asset factor in the Longstaff and Schwartz model), and
with changes in Australian Government bond yields as well (a proxy for the interest rate
factor in the Longstaff and Schwartz model). Their contribution lies in providing an insight
into why the relative measure tends to be statistically more significant than the alternate