210 S. Drucker and M. Puri
could underwrite its securities. They allow the choice of underwriter to depend upon the
size of the issue, existing lending and underwriting relationships, and the reputation of
the underwriters in both the underwriting and lending market.Bharath et al. (2004)find
that prior lending relationships significantly increase the likelihood of winning debt un-
derwriting mandates and being selected as lead manager on IPOs.
Ljungqvist, Marston, and Wilhelm (2006)also provide evidence related to the influ-
ence of bank lending relationships on underwriter selection. Using a sample of 16,625
debt and equity deals over the period December 1993 to December 2002, the authors
estimate a probit model, providing each firm with the potential to choose any of the
16 most active underwriters. In their model, the choice of underwriter depends upon
bank-firm underwriting and lending relationships, as well as bank reputation and ana-
lyst characteristics and behavior.Ljungqvist, Marston, and Wilhelm (2006)find that the
probability of winning both equity and debt underwriting business is increasing in the
bank’s share of the issuer’s prior loans.
Yasuda (2005)provides an examination of the impact of existing lending relation-
ships on the choice of debt underwriter during the period 1993 to 1997. As previously
explained,Yasuda (2005)estimates a joint model of the gross spread and the firm’s se-
lection of underwriter, allowing the firm to choose between sixteen underwriters. The
joint framework allows the author to include in the underwriter selection equation the
estimated fee that each underwriter would have charged the issuer to underwrite the
offering. Therefore,Yasuda (2005)can examine if the lending relationship influences
underwriter selection above and beyond any effect that charging lower gross spreads
has on underwriter selection. The author finds that prior lending relationships signifi-
cantly increase the likelihood that the lending bank wins the bond underwriting business
over and above the effect of the gross spread discount. Further, lending relationships
have a stronger impact on underwriter selection among junk rated issuers and new is-
suers, where a bank’s private information is likely to be most valuable. These results
are consistent withKanatas and Qi’s (1998, 2003)theoretical models, which indicate
that lenders will select their bank as underwriter when there are likely to be large in-
formational economies of scope. The estimates ofYasuda’s (2005)model indicate that
firms are willing to pay a higher underwriter fee to banks with which they have a prior
relationship. One of the major benefits of this framework is that the author can explicitly
calculate how much more an issuer is willing to pay. For the sample mean issue size of
$180 million, an issuer is willing to pay $2.23 million more to use a relationship com-
mercial bank and $2.62 million to have a relationship investment bank as underwriter.
Junk rated issuers and first time issuers, where the value of a bank’s private information
is likely to be largest, are willing to pay even more. These results are consistent with a
certification effect for relationship banks.
4.3. Can investment banks survive?
Overall, the empirical evidence shows that using relationship banks as underwriters im-
proves the pricing of issues and lowers fees, and both prior lending relationships and