Computational Complexity and Tort Deterrence

Standard formulations of the economic model of tort deterrence constitute the injurer as the unboundedly rational bad man. Unbounded rationality implies that the injurer can always compute the solution to his caretaking problem. This in turn implies that optimal liability rules can provide robust deterrence, for they can always induce the injurer to take socially optimal care. In this paper I examine the computational complexity of the injurer’s caretaking problem. I show that the injurer’s problem is computationally tractable when the precaution set is unidimensional or convex but that it is computationally intractable when the precaution set is multidimensional and discrete. One implication is that the standard assumptions of unidimensional and convex care, though seemingly innocuous, are pivotal to ensuring that tort law can provide robust deterrence. It is therefore important to recognize situations with multidimensional discrete care, where robust tort deterrence may not be possible.