This Comment attempts to explain two stylized facts: As the market interest rate rises, lenders demand either (a) more collateral, or (b) tighter covenants. In their Article, Conditions and Contract Design: Variations in Debt Covenants and Collateral, Choi and Triantis (“C&T”) use two models in their explanation of these facts: an adverse selection model and a moral hazard model. The adverse selection model formally analyzes only collateral contracts, but the authors claim that both the collateral contract and the covenant contract mitigate adverse selection. The moral hazard model also considers only collateral contracts; the claim here is that these best mitigate moral hazard.
The Article claims to derive three principal results:
(A) The likelihood that parties write collateral contracts is increasing in the market interest rate (both models). When the problem is adverse selection, this result applies to covenants as well.
(B) Good (i.e., less risky) types are more likely to offer collateral than bad types (both models);
(C) The difference in the contracts of good and bad types widens as the market interest rate increases (both models).
This Comment argues that under symmetric information, results (A) and (C) continue to hold but result (B) reverses: Bad types offer more collateral than good types. Symmetric information is the more plausible assumption in the Article’s adverse selection setup. I make three other points: (i) C&T’s moral hazard model assumes that borrowers are ex ante identical: Every borrower is equally likely to pursue a later project that disadvantages the initial lender. This setup cannot explain why some partie —lenders and borrowers—use covenants while other, apparently similar parties, use security; (ii) Letting borrowers differ ex ante may explain when parties prefer one or the other risk reduction device. In this framework, the better argument continues to be that bad types offer more collateral than good types; (iii) Turning to empirics, some data rejects C&T’s result (B): Bad types appear in fact to offer more collateral than good types. Because this data is not current, it cannot settle the issue. On the other hand, and to summarize, the data and the analysis raise the question whether C&T are using the right models for the problem.