When a debt covenant is violated the lender has the right to demand immediate repayment of the loan. Using this right, the lender can extract certain concessions from the borrower (manager), which may be inefficient. I propose a theory that explains why, despite this potential inefficiency, tight and often violated debt covenants may be optimal. The key element of this theory is that ex-ante uncontractable effort is observable ex-post by all market participants. This provides an outside option for the manager in a renegotiation game with the bank.

First I consider an extreme case in which it is impossible to contract on important variables. I show that, in equilibrium, the parties will adopt strict covenants contingent on very noisy variables. This is consistent with several empirical facts, including the ex-ante strictness of debt covenants, a large number of violations, and the fact that a covenant violation by itself does not contain full information about actual conditions of the firm.

I apply this set-up to analyze changes in loan market conditions (part 2) and the changes in contractual incompleteness (part 3).

If market conditions are ex-ante very good, the manager expects to be rewarded for high effort and the optimal covenants will not be particularly strict. Once market conditions worsen, the punishment that strict covenants create for poor performance becomes important and equilibrium covenants become stricter.

By allowing partial contractability in part 3 of the dissertation, I suggest a set-up that is in the middle of extreme incomplete contracting and the standard complete contract model. In the latter case, no covenants are needed. I show that as the correlation between the contractable variable and the uncontractable variable increases, the probability of a covenant violation, as well as the probability of a renegotiation decreases. Surprisingly, conditional on a renegotiation taking place, the probability that this renegotiation is favorable for the manager generally goes up as correlation increases.