period. Thus, due to Duffee’s (1998) warning, we expect the impact of callability on the
results to be significant for that period.
***Insert Table 2 here***
The results for regression model (1) are in agreement with the results of Longstaff and
Schwartz (1995) for their U.S. bonds. The estimated coefficients for b are negative and
statistically significant for all indices during the 08:1976-07:2001 estimation period. As in
Longstaff and Schwartz (1995), the magnitude of the estimates of b is economically
significant for the evolution of yield spreads, and the coefficient b monotonically decreases
with credit quality. Also in agreement with Longstaff and Schwartz (1995), Panel A of Table
2 reports that the estimates of c are all negative and statistically significant. The coefficient c
is also found to decrease monotonically with credit quality, and is economically significant.
This supports the importance of Longstaff and Schwartz's (1995) asset factor, i.e., ceteris
paribus, higher firm values result in lower probability of default and consequently lower
yield spreads.
However, drawing the conclusion that default yield is negatively related to interest rate
simply based on the above result is premature. Following Duffee’s (1998) results, we believe
that this result is due to the negative impact of callability on the estimated relationship. To
test this hypothesis, we apply regression model (1) to all indices for the 01:1995-07:2001
period, in which our long indices are expected to be dominated by bonds carrying the
doomsday call.
This allows us to control for callability for the BBB index. Recall that for BBB-rated
bonds the doomsday call will always be out of the money, making them economically
noncallable. This leaves the default term as the only factor potentially affecting the sign of b
for the BBB index. Based on their findings, which are similar to our results for the entire 25-
year sample period, Longstaff and Schwartz (1995) conclude that the interest-rate factor is
17