68 K. Li and N.R. Prabhala
their debt issues. This incentive is recognized and priced by the market, and the pricing
differential again feeds back into firms’ decisions about whether to include covenants.
In other words, the decision to include covenants influences and is influenced by the
expected pricing benefits from doing so. Goyal implements the structural self-selection
model of Section3.2to model the simultaneity.
Goyal estimates the structural model on a sample of 415 subordinated debt issues
made by firms between 1975 and 1994. He finds that yields are negatively related to
franchise value. This finding is consistent with the hypothesis that banks with greater
franchise value have less incentives to take risk, latent information that is recognized
and priced by financial markets. The inverse Mills ratio term is significant in the no-
covenant sub-sample but not in the sample with restrictive covenants. In the equation
explaining whether firms use covenants or not, the coefficient for the yield differential
with/without covenants is significant in explaining covenant choice, suggesting that an-
ticipated pricing benefits do influence whether firms select covenants or not in their debt
issues. Many of Goyal’s results are more prominent in the 1981–1988 sub-period, when
the risk-taking activity in the U.S. was more elevated.^26
8.5. Discussion
The public debt issue pricing area is interesting for the wide range of selection models
employed. One issue, however, is that it is a little difficult to place the literature in per-
spective because the sources of self-selection modeled vary across papers. An additional
issue is, of course, that there is probably self-selection on other dimensions as well, such
as maturity, collateral, or the callability of an issue, not speaking of the decision to issue
debt in the first place. This raises another thorny question, one that probably has no easy
answer. What dimensions of self-selection should one control for in a given empirical
application? Modeling every source of selection seems infeasible, while studying some
sources of bias while ignoring others also seems a little ad-hoc. Embarking on a purely
empirical search for sources of selection that matter is certainly undesirable, smacking
of data snooping. A happy middle way is likely to emerge as the literature matures.
- Other investment banking applications
9.1. Underwriter compensation in IPOs:Dunbar (1995)
Dunbar (1995)presents an interesting application of aRoy (1951)style self-selection
model to the study of underwriter compensation. Some IPO issuers offer warrants to
compensate their underwriters while other issuers do not. Dunbar examines the role
(^26) Reisel (2004)provides an interesting extension, a structural self-selection model applied to debt covenants
included in industrial bonds.