70 K. Li and N.R. Prabhala
9.2. Analyst coverage:Ljungqvist, Marston and Wilhelm (2006)
Ljungqvist, Marston and Wilhelm (2006)examine the relation between the decision
to award an underwriting mandate to a bank and the coverage offered by the bank’s
analyst. The self-selection issue in Ljungqvist et al. is that banks self-select on whether
they cover a stock or not. If the bank covers a stock, we observe the nature of the stock
recommendation and we can tie it to the decision to award an underwriting mandate.
However, if a bank does not elect to cover a stock, we do not know what the nature of
its recommendation might have been had it chosen to cover the stock. Ljungqvist et al.
model this source of self-selection in testing whether a firm with more positive coverage
of a firm is more likely to win the firm’s underwriting mandates.
Ljungqvist et al. model the probability that bankjcovers firmi’s stock as a probit
model
yC=1ifyC∗=XCβC+uC> 0 ,
yC=0ifyC∗=XCβC+uC 0 , (63)where all subscripts are suppressed for notational convenience. If there is coverage, the
tie between coverage and the award of an underwriting mandate is established by the
equations
(64)yA=βAXA+uA
yL=IβLXL+δLyA+uL> 0}
ifyC∗> 0.If there is no coverage, we have
(65)
yA= 0
yL=IβLNCXL+uLNC> 0}
ifyC∗ 0 ,whereyAis the nature of an analyst’s recommendation,yLis a 1/0 dummy for whether
an underwriting mandate is awarded to a bank,Iis the 1/0 indicator function, andX’s
are explanatory variables. Equations(63)–(65)represent a switching regression system,
similar to the type analyzed in Section3.1. The difference here is that we have two re-
cursive equations observed in each regime instead of just one regression in Section3.1.
Ljungqvist et al. find that the decision to cover a stock is positively related to the type
of coverage offered by an analyst for debt underwriting transactions. Prior relationships
in the underwriting and loan markets are the other most significant explanatory vari-
ables. There is no evidence that the type of coverage influences the decision to award
equity underwriting mandates. Even when it is significant, the coefficient for analyst
recommendationβAin equation(64)is negative. Ljungqvist et al. interpret this finding
as evidence that even if analysts are overly biased, issuers refrain from using them for
underwriting.
The analysis of Ljungqvist et al. has appealing features. The choice of instruments
is carefully motivated, with both economic intuition and tests for instrument strength
suggested byStaiger and Stock (1997). Their analysis also suggests some natural ex-
tensions. One issue is that the very decision to cover a stock—rather than the type of