The model of precaution has become a central tool of law and economics, beginning with Judge Learned Hand’s brilliant opinion in United States v Carroll Towing Co. In it he argues that a defendant should be found liable for harm if and only if the expected cost of additional care is less than the expected benefit.

The model of precaution relies upon the economics of incentives, a subfield of game theory—the study of how individuals choose actions when these actions affect others. The landmark books of Professor William Landes and Judge Richard Posner, and Professor Steven Shavell illustrate how the precaution model illuminates a wide variety of legal rules. Professor Guido Calabresi and A. Douglas Melamed show how it can be used to integrate tort and property. Professor Robert Cooter uses the model to provide a unified analysis of tort and contract.

A central result of the model is that the standard for negligence provides incentives for individuals to take socially optimal actions. This perspective is controversial. Professor Richard Epstein argues that the model of precaution cannot explain observed law and does not provide an adequate model of causation as used in court. He suggests that the socially efficient rule should be one of strict liability. Posner responds, providing a number of examples in which Epstein’s approach would lead to undesirable results. He describes Epstein’s model as being based upon moral rather than economic considerations.

In this Article we show that the model of precaution is a special case of a more general economic model. We develop a simple technique for discussing this more general model that we dub a “Savage Table,” following Professor Leonard Savage. Our more general model based on general equilibrium theory and decision theory encompasses the views of both Posner and Epstein and sheds light on Epstein’s observation that the Hand Rule is not consistently used to determine liability. We show that rational choice does not imply the Hand Rule unless one imposes additional restrictions that are not often satisfied in practice.

We show that strict liability can be viewed as a special case of the negligence standard. In addition, the notion of a causal effect can be easily and naturally defined in our framework. The benefit of a clear definition is that it highlights the key evidentiary requirements to determine causation. Recent work in statistics has greatly clarified our understanding of causal inference. There is no way to prove the existence of a causal relationship. The best we can do is to use a credible model of the world and make causal statements within the context of the model.

Once we look at economic phenomena from our more general perspective, we are led to the following legal-impossibility theorem: For every legal rule that is potentially efficient there exists an environment with rational decision makers for which the rule is not efficient. A rational decision maker must make choices conditional upon the information she has. Given incomplete information, there is no guarantee that a rule that encourages good decision making in one context will work in another.

Posner observed that the best way to deal with this problem is to rely upon empirical evidence. Models are still useful. In fact Professor Paul Holland shows that model building is essential to the measurement of causal effects. Theory and evidence must work together to identify those worldviews that are more successful than the alternatives.