callability drives the negative sign of both level and slope coefficients.
5.3. Collin-Dufresne et al.’s (2001) Comprehensive Regression
Next we establish if the previous results are affected by model specification by
applying Collin-Dufresne et al.’s (2001) comprehensive regression (4). Since our data
consists of bond indices, whereas they use firm level data, we omit some firm-specific
explanatory variables used in their regression analysis. As in the previous sections, we apply
a maximum-likelihood estimation procedure for a combined autoregressive model and a
GARCH (1,1) model. Table 5 outlines the estimates for regression model (4). Again, in some
cases, we find that our SCM data set is characterized by the autoregressive nature of the OLS
residuals of regression model (4).
***Insert Table 5 here***
In general, we obtain results consistent with our previous regressions. Panel A of Table
5 outlines the results for our entire sample, covering the 08:1976-07:2001 25-year period,
during which corporate bonds carrying a standard call provision dominate the data. We find
that yield spreads are negatively related to Government of Canada yields for all bond indices,
and also to the return on the Toronto Stock Exchange 300 index.
Panel B of Table 5 reports our results for the 01:1995-07:2001 period. For AA- and A-
rated bonds, the yield spread is still significantly negatively related to Government of
Canada yields. For BBB-rated bonds, which are economically noncallable during this period,
the yield spread - Government yield relation is still insignificant. These results provide
additional support for our conclusion that callability dominates the observed negative
relationship between credit spread and riskless rate.15 In the absence of an economically
viable call option, the correlation between credit spread and riskless interest rate is
insignificant.
22