Handbook of Corporate Finance Empirical Corporate Finance Volume 1

66 K. Li and N.R. Prabhala

if a pure investment bank syndicate is chosen. Thus, a hybrid syndicate is observed
(regime 1) whenAi=1 andBi=1; a pure investment bank syndicate (regime 2) is
observed whenAi=0 andCi=0, while a commercial bank led syndicate (regime 3)
is observed whenBi=0 andCi=1.^25 Song assumes that the latent errorsηare i.i.d.
normal, correlated with yieldsYwith regression coefficientsσωjwhereω∈{A, B, C}
andj∈{ 1 , 2 , 3 }. The yields in each regime can be expressed in regression form as

(54)

E(y 1 i|Ai= 1 ,Bi= 1 )=X 1 iβ 1 +σA 1

φ(ZAiγA)

1 −Φ(ZAiγA)

+σB 1

φ(ZBiγB)

1 −Φ(ZBiγB)

,

E(y 2 i|Ai= 0 ,Ci= 0 )=X 2 iβ 2 −σA 2 (55)

φ(ZAiγA)

Φ(ZAiγA)

−σC 2

φ(ZCiγC)

Φ(ZCiγC)

,

E(y 3 i|Bi= 0 ,Ci= 1 )=X 3 iβ 3 −σB 3 (56)

φ(ZBiγB)

Φ(ZBiγB)

+σC 3

φ(ZCiγC)

1 −Φ(ZCiγC)

.

Song’s sample comprises 2,345 bond issues offered between January 1991 and De-
cember 1996. In the first step probit estimates, Song reports that compared to pure
investment bank syndicates, hybrid syndicates underwrite small firms that have made
smaller debt issues in the past, have low S&P stock rankings, invest less, and use more
bank debt. These findings are reminiscent of those inGande et al. (1997)andGande,
Puri and Saunders (1999)that commercial banks underwrite informationally sensitive
companies. Compared to commercial bank led syndicates, hybrid syndicates underwrite
smaller firms with lower stock rankings that issue to refinance debt and lower ranked
firms, consistent with the claim that these underwritings potentially alleviate conflicts of
interest with commercial banks. Only two out of six private information coefficients in
equations(54)–(56)are significant. Pricing benefits are seen in pure investment banking
syndicates (equation(55)) where excluding a commercial bank leads to higher yields,
consistent with a certification hypothesis. On the other hand, picking an investment
bank to run the syndicate increases yields, because the coefficientσC 2 in the same equa-
tion(55)is positive. Thus, the ex-ante effect of awarding a syndicate to an investment
bank cannot be a priori signed.
Relative to prior work,Song (2004)has very different sample, sample period, and
explanatory variables, not to mention the changes in underwriter classification, which
is based on syndicate structure rather than on classification into commercial/investment
bank or on bank reputation. Thus, it is hard to pinpoint the specific value added by her
elaborate selection model. In addition, absent additional diagnostics, it is also difficult
to interpret whether the general insignificance of most selection terms reflects coeffi-
cients that are truly zero, the lack of power, perhaps due to collinearity, or perhaps an
unmodeled correlation between errors in equations(48)–(50)that are assumed to be

(^25) Song does not explicitly write out the extensive form of the model she estimates. It is unclear whether pure
investment bank syndicates should also include the node at which an investment bank is awarded the mandate
and chooses to invite a commercial bank but the bank declines to join.