222 S. Drucker and M. Puri
capitalists as equity holders reduce IPO underpricing when they underwrite and, further,
gross spreads on IPOs decrease in the underwriters’ shareholdings of the firm (Li and
Masulis, 2004).
Why and when do banks choose to invest in equity, and what are the implications
for the firm? There is surprisingly little research on this issue.Hellmann, Lindsey, and
Puri (2006)explore this topic by focusing on the impact of bank venture capital rela-
tionships on the bank’s core lending division. Venture relationships may allow the bank
to foster an ongoing lending relationship. Also, the private information from the ven-
ture relationship may reduce the bank’s cost of lending due to informational economies
of scope, allowing firms to benefit from lower loan yield spreads. To examine these
issues, the authors collect detailed information on all venture capital investments for
the period 1980 through 2000 and gather lending data for the 10,583 venture backed
customers. To examine if venture relationships increase the likelihood that the bank in-
vestor will forge a lending relationship with the firm, the authors estimate a conditional
logit model in which each firm can choose among banks. The results reveal that the ven-
ture relationship does indeed increase the likelihood of being selected as lender, even
after controlling for the bank’s share of the lending market and the firm’s public status.
These results are confirmed through another test at the aggregate level in which the au-
thors find that banks are more likely, on average, to lend to companies with whom they
have a prior venture relationship.
To examine loan pricing,Hellmann, Lindsey, and Puri (2006)match loans where the
lender has a prior venture relationship with the firm (“relationship loans”) with similar
loans where no venture relationship exists (“non-relationship loans”) and compare the
yield spreads of the matched loans. Since it is difficult to match loans directly based
on multiple relevant characteristics, the authors use propensity score matching, which
reduces the multiple-dimension matching problem to a single-dimension, called the
propensity score.^34 These methods take into account the fact that the characteristics
of relationship loans may differ significantly from non-relationship loans and ensure
that such observed characteristics are not driving the results. Using various estimators,
the authors find that relationship loans have significantly lower yield spreads, by 18 to
26 basis points. In sum, the results suggest that as venture capitalists, banks tend to
be strategic investors in equity and use venture capital relationships to foster a lending
relationship that results in efficiencies that benefit both banks and firms. The results
control rights through board seats and equity holding is found in Germany byGorton and Schmid (2000),who
find that banks use their equity holding and board seats to improve firm performance. For Japan,Kaplan and
Minton (1994)find banks are more likely to get board seats following poor firm performance, andWe i n s t e i n
and Yafeh (1998)andMorck, Nakamura, and Shivdasani (2000)find that Japanese firms with a main bank
have lower growth and profitability than others.
(^34) To employ the methodology, the authors first run a probit model, where the dependent variable is one if
the loan is a relationship loan and zero otherwise, and the independent variables are the matching dimensions,
which include loan and borrower characteristics. Each loan is assigned a propensity score, which is the pre-
dicted probability from the probit model. SeeHeckman, Ichimura, and Todd (1997, 1998)for more details.
This method of matching loans is also used inDrucker and Puri (2005).