Bargaining under incomplete information and the design of legal rules

This dissertation studies the game-theoretic relationship between legal rules and bargaining among privately-informed litigants. I employ mechanism design throughout, which facilitates a general analysis of bargaining without committing to a particular extensive-form protocol.; Chapter 1 examines the normative choice between property rules and liability rules within asymmetric-information bargaining mechanisms. Liability rules are shown to facilitate more efficient allocations through Coasean trade than do property rules, because the former induce "countervailing incentives" phenomena that reduce required information rents. Counter-intuitively, liability rules are sometimes more advantageous in the presence of judicial "inaccuracy" and/or litigation costs. I discuss various legal applications and limitations of the model, illustrating its compatibility with dynamic investment concerns.; Chapter 2 examines the "liquidated damages doctrine," under which courts invalidate stipulated contractual damages if such terms are deemed unreasonably punitive. While some argue that the doctrine is inefficient, others posit possible justifications for the doctrine. The most provocative justification argues that the contractual parties have an inefficient incentive to overliquidate damages, in order to extract rents from potential third-party entrants. This explanation, however, has not been thought robust to renegotiation. I defend against this apparent weakness in the context of asymmetric-information bargaining. I argue that because contractual penalties create effective property rules for the promisee, such terms are actually more resistant to renegotiation than a more moderate damages clause (which creates liability protection).; Chapter 3 analyzes pre-trial bargaining between privately-informed litigants in the shadow of liability-contingent fee shifting rules. I study a complete parameterization of such rules and find that the so-called "loser pays" or "British Rule" tends to induce lower settlement rates than any other type of fee shifting rule when (1) at least one agent has private information the probability of a finding of liability, and (2) it is common knowledge that the plaintiff's suit has positive net present value. In such circumstances, a rather unintuitive "winner pays" rule weakly outperforms to all others in inducing settlement. I briefly analyze variations on the model, including different forms of private information, deterrence, and the negative-present-value suits.