The dissertation consists of three essays in law and economics. The first chapter compares the efficiency of negligence and strict liability in unilateral accident cases when the injurer faces ambiguity about accident risk. It generalizes the standard accident model to allow for ambiguity by assuming the injurer is a Choquet expected utility maximizer and representing the injurer's beliefs about accident s risk with a neo-additive capacity. The central result is that neither strict liability nor negligence is generally efficient in the presence of ambiguity. A key implication of the results is that negligence is more robust to ambiguity, which may help explain why negligence is the general basis for accident liability under modern Anglo-American tort law. The second chapter examines how different allocation rules influence the risk that putative class members will opt out of a mass tort class action. It analyzes a two-stage model of class action formation. The main result is that the class will be asymptotically stable if the net recovery will be allocated pro rata by expected claim values, but may not be asymptotically stable if the net recovery will be shared equally or allocated pro rata by damage claims. Other results explore how the shape of the distribution of the plaintiffs' damage claims, the scale benefits of the class action, and the plaintiffs' probability of prevailing and bargaining power in settlement negotiations influence the stability of the class. The third chapter offers a model of analogical legal reasoning. Under the model, the outcome in the case at hand is a weighted average of the outcomes of prior cases, where the weights are a function of fact similarity and precedential authority. The main theoretical result is an axiomatization of similarity-weighted averaging with an exponential similarity function based on a quasimetric. The chapter also investigates whether the analogical model provides a better fit than a rule-based model (represented by a fractional polynomial) to the reported decisions by federal judges in U.S. maritime salvage cases from 1880 to 2007. The principal conclusion of the empirical analysis is that the rule-based model fits the data better than the analogical model.