Ch. 5: Banks in Capital Markets 199
for this time period. To compare default performance between ex ante similar bonds,
the authors use two methods: (i) matched-security tests, where bonds originated by
commercial bank affiliates are matched to similar investment bank-underwritten bonds
based on observable characteristics; and, (ii) logit analysis. For the matched-security
tests, the authors create a sample of ex ante similar commercial bank and investment
bank-underwritten securities, using the credit rating as the primary measure of bond
quality. In total,Kroszner and Rajan (1994)find 121 industrial bond matches, where
the bonds have the same initial credit rating, are issued within six months of each other,
have similar maturity and size, and have the same conversion provision.^10 Using this
sample, the authors find that at the end of each year after 1924, there are fewer cumula-
tive defaults among commercial bank-underwritten issues, and by the end of the sample
period in 1940, 32 percent of investment bank-underwritten bonds defaulted relative
to 23 percent of bonds that were underwritten by commercial banks. By dollar volume,
approximately 28 percent of investment bank-underwritten issues default by 1940, com-
pared with only 11 percent of commercial bank-underwritten issues. Further, not only
do investment bank-underwritten issues default more frequently, but they also default
earlier in their lives. All of these findings suggest that commercial bank-underwritten
issues performed better than similar, investment bank-underwritten issues, which is in-
consistent with commercial banks succumbing to conflicts of interest.
In addition,Kroszner and Rajan (1994)perform a log-rank test using the sample
of matched securities. This test takes into account both the number of defaults and
the timing of defaults by comparing the mortality rates of the two groups of bonds.
Consistent with their initial findings, the main result of this test is that the survival rate
of commercial bank-underwritten bonds is significantly higher than investment bank-
underwritten bonds. Importantly, these differences are strong in the non investment-
grade sample, but insignificant in the investment-grade sample. Since incentive conflicts
created by information asymmetries between underwriters and investors are larger in
low quality issues, this result supports the view that conflicts of interest were not large
during the pre-Glass–Steagall period.
Kroszner and Rajan (1994)confirm the lower default probability of commercial bank-
underwritten issues using logit analysis in which they estimate whether the type of
underwriter affects the probability of default, after controlling for security and firm
characteristics. The logit analysis complements the matched-security tests by allowing
the authors to use data on all of the investment bank-underwritten issues (instead of
just the smaller sample that is matched to commercial bank-underwritten issues) and
providing means to control for other factors that may be correlated with default. Ac-
cording to the estimates from logit models, underwriting by a commercial bank reduces
the probability of default by 11 percent, with large and significant reductions in default
probabilities seen among the lowest quality issues. In economic terms, an 11 percent
(^10) If there are multiple matches, the authors use other criteria, such as collateralization status, to select the
best match.