Endnotes
1 For example see Merton, (1974) Black and Cox (1976) and Longstaff and Schwartz (1995) and Ericsson
and Renault (forthcoming).
2 Duffee (1998) notes that there are tax effects arising from the different tax rates, which apply for U.S.
corporate and Treasury bonds. However, using U.S. bond data, he is unable to control for the impact of the
tax differential.
3 Batten, Hogan, and Jacoby (2005) use a sample of noncallable Australian Eurobonds and find evidence
supporting the predictions of the two-factor Longstaff and Schwartz model. Nonetheless the preference by
international investors for the with-holding tax exempt yields of Eurobonds may distort the pricing
relationship between high credit quality Eurobonds and Government bonds.
4 According to Duffee, following a given rise in Treasury yields, everything else being equal, the yield on
a corporate bond will have to increase by a higher rate, so that the after-tax yield spread will remain
unchanged. This implies that the pre-tax yield spread will widen following an increase in Treasury yields.
5 The reader may refer to Jacoby and Roberts (2003) for a detailed description of the SCM indices.
6 Mann and Powers (2003) refer to the make-whole call provision as “a recent innovation in the public
debt markets.”
7 Note that there is still a small probability that a BBB-rated bond will be upgraded in the future and that
its doomsday call will be in-the-money. However, given the small rate of upgrades, the expected value of
this state is trivial.
8 In a small number of cases, the cross-sectional standard deviation of the doomsday spread is zero, which
implies: μ = μ+σ = μ+2σ. This is due to a small number of bonds available during the given month,
usually issued by the same company, all sharing the same doomsday spread. In other cases, specifically in
the 06:1993-07:1994 period, the reported doomsday spread for BBB-rated bonds is zero.
9 Note that since our sample is not stratified into different sectors, there is no need in the current study to
use sector-specific stock indices as in Longstaff and Schwartz (1995).
10 Longstaff and Schwartz (1995) estimate their regression models using OLS. As previously noted, we
find that a combined autoregressive and GARCH (1,1) model fits our data best. However, the qualitative
results turn out to be similar irrespective of the estimation procedure used. Thus, for the sake of
comparison with Longstaff and Schwartz (1995), our focus in the discussion of regression models (1) and
(2) is on the results of the OLS estimation, which are presented in the tables below. The results of the
combined autoregressive and GARCH (1,1) regressions are reported in the appendix.
11 The estimated coefficient b for the BBB index under the Yule-Walker estimation is also statistically
insignificant. Thus, this result is robust.
12 A higher probability of a call for corporate bonds carrying the doomsday call provision reduces their
effective duration, or price sensitivity to changes in the riskless rate. This implies that following an
upward shift in the riskless rate, the corporate bond yield will rise by a lower rate, and yield spreads will
contract.
13 Note that we obtain the same result for the BBB index when the Yule-Walker procedure is applied.
14 In analyzing the results, note that one's conclusions are insensitive to whether one uses OLS or the
estimation method we apply here.
15 Our results are different from those in Collin-Dufresne et al. (2001). This may be due to the fact that
their data, devoid of callability, still suffer from the coupon bias which may generate the observed
negative relationship for noncallable bonds.
16 We adopt the same approach as Jorion and Goetzmann (2000) Table 1.
48