A Complete Characterization of Efficient Liability Rules: Comment

In a recent article, Jain and Singh (2002) prove that a condition they call negligence liability is necessary and sufficient for any liability rule to be efficient. In this note I criticize their result on two accounts: First, their result crucially depends on implicit restrictions they impose on the liability assignment function. If we drop the assumption that the liability apportionment between two non-negligent parties is constant for all combinations of non-negligent care levels, the equivalence between the condition of negligence liability and the efficiency of liability rules breaks down. Second, their attempt to drop the assumption of uniqueness for the social optimum improves the generality of the model at a substantial cost, since it must be accompanied by a new assumption that is possibly even more unrealistic. The importance of the uniqueness assumption is shown in a simple discrete care model, in which comparative negligence may lead to an inefficient outcome when the existence of two social optima leads us to interpret due care as a varying standard based on the other partys actual choice.