Mark Geistfeld

Empirical study of the law is important, particularly for tort law. Fundamental components of the tort system are a “black box,” which largely explains why the field is riven by theoretical disagreement over the purpose of tort law. The claim that tort law efficiently reduces accident costs, for example, critically depends on the extent to which the threat of tort liability deters risky actors from behaving inefficiently. The available data on accidents, however, do not directly measure the relationship, no doubt because the injury rate is affected by a large number of other interrelated factors such as changes in wealth and technology that are extraordinarily hard to disentangle, making it extremely difficult to identify the impact that tort liability has had on actual accident rates. To isolate the influence of particular factors such as the threat of tort liability, empirical study must instead turn to the laboratory, where researchers can conduct experiments that are designed to tease out the role of the varied factors that plausibly explain the accident rate—an excellent example of which is provided by Theodore Eisenberg and Christoph Engel in their article, Unpacking Negligence Liability: Experimentally Testing the Governance Effect.

As persuasively argued by Frederic Schauer in The Force of Law (2015), important jurisprudential questions depend on the particular reasons why individuals comply with the law. In particular, individuals often have independent normative reasons for acting in the manner otherwise required by the law, in which case the law itself is not motivating the behavior. “Until we can understand the different ways in which law intersects with its subjects’ law-independent preferences, we cannot begin to understand the role of incentives and coercion in motivating legal compliance.” (P. 100.) The experiment conducted by Eisenberg and Engel was designed to address exactly this type of problem.

The experiment tries to disentangle four different motivations that could affect an individual’s decision of how much risk to impose on another. First, empirical studies have found that individuals may have an intrinsic moral inhibition to harm a passive bystander, in which case an individual could act safely for moral reasons that she has adopted independently of tort law. A related motivation involves the desire to avoid being blamed for another’s injury, which can again (though not necessarily) be independent of tort law. The remaining two motivations directly implicate tort law. Individuals may adopt a point of view that is internal to the law and simply desire to conform their behavior to the normative requirements of tort law, even if the law is never enforced. Finally, individuals may instead be motivated only by the concern for avoiding liability, acting like the Holmesian “bad man” posited by the economic analysis of tort law. In an attempt to isolate these varied motivations, the experiment involved four different stages that each added new information relevant to a decision by an active participant of how much to engage in a risky activity threatening injury to an otherwise passive participant in the experiment.

In the baseline treatment, announced as a one-shot experiment, the active participant chooses the amount of risky activity based on a specified profit function and is also informed about the corresponding loss function faced by the passive party. The instructions also tell the active participant about the amount of the risky activity that will maximize his or her profit. After a surprise restart, the two participants are told that the experiment will be repeated 50 times and are informed both of the (inefficiently high) level of risky activity that maximizes the risky actor’s payoff and the (efficient) lower level of risky activity that maximizes the expected joint income of the two parties. In the next stage, if the active participant has engaged in excessively high risky activity and the passive participant suffered injury as a result, the injurer faces a 50% probability of being “audited” and then blamed for having exceeded “the socially desirable level of investment.” (P. 150.) The final stage adds a liability component, requiring such an audited injurer to pay compensation to the injured party. The same experiment was then conducted with the passive victim being a computer, in which case the moral inhibition not to harm others presumably would be eliminated.

In an impressively nuanced discussion of the experimental findings, the authors conclude that the only individual motivation without any significance involves the intrinsic morality of not harming a passive bystander. Risky activity was instead reduced in each stage by the additive triggering of three different motivations. In the first or baseline stage only involving the profit and loss functions, 86% of the active participants invested in the profit-maximizing amount. In the second stage announcing the socially desirable level of activity, a significant number of participants were motivated to comply with the law without any threat of sanction (the fraction of participants investing in the profit-maximizing amount dropped to 65%). In the third stage that blamed but did not sanction those who were “audited” and found to be in violation of the announced standard, a fewer but still significant number of participants were motivated by the desire to avoid blame alone (the fraction of participants investing in the profit-maximizing amount was reduced to 58%). In the fourth stage that subjected these excessively risky decisions to liability for compensatory damages, a significant number of participants were motivated by the desire to avoid liability (the fraction of participants investing in the profit-maximizing amount dropped to 33%). The different levels of the risky activity at each stage provide a measure of the influence of each type of individual motivation, leading Eisenberg and Engel to conclude that negligence liability deters risky behavior through a “governance effect” involving the public announcement of a normative expectation coupled with blame and sanctions for violation of the legal norm.

And what should one make of these findings? Consider in this regard their empirical finding that the risky activity was significantly reduced simply because of a public announcement regarding the socially desirable level of activity, a normative expectation not backed up by sanctions (which only entered into the final stage of the experiment). This finding most readily lends itself to the interpretation that a significant proportion of the population has adopted the internal point of view that motivates them to comply with the law, regardless of sanctions. Tort law can deter merely by publicly announcing its normative expectations of safe behavior.

This finding, though important, becomes more puzzling when considered in relation to the actual practice of tort law. The standard case of negligence liability turns on the jury’s determination of whether the defendant exercised reasonable care. Jury instructions do not define the requirements of reasonable care, so the jury presumably makes some type of normative judgment about reasonably acceptable behavior that is not defined by conventional safety practices (like jaywalking in New York City). If so, then the publicly announced normative expectation of tort law is that risky actors should comply with the relevant social norms. The “governance effect” identified Eisenberg and Engel, therefore, must somehow depend on the intrinsic morality of social norms, yet they found that intrinsic morality has no independently significant effect on risky behavior. The experimental design has missed something.

The problem in my view stems from the baseline specification of the safety decision, which relied on mathematically complex profit and loss functions. The instructions were formulated to help the participants avoid making decisions based on mathematical errors, but the logic of the decision was not intuitive. (I found it easiest to perform some simple calculus to clarify matters.) This type of decision making could easily have misframed the normative problem for those who were actively participating in the study, particularly as they were only expressly told about the amount of risky activity that would maximize their own profit and were not given any other motivations for the decision. The decision was framed as one of rational prudence without any salient moral dimensions. The active participant could easily have assumed that he or she merely faced a complex investment decision involving profits that threatened some stranger with a probabilistic monetary loss, a decision with important attributes of ordinary market behavior that is normatively different from risky behavior threatening bodily injury or premature death to a passive bystander. The subsequent announcement of the socially desired behavior in the next stage of the experiment would then cause some individuals to change their decisions by inducing them to reframe the normative problem. Tort law, therefore, can deter by helping individuals to frame properly the normative problem, but this “governance effect” may be an artifact of the experimental design that would yield different results if it had properly framed the normative problem in the first instance.

Tort law in the laboratory does not necessarily generalize to the real world, but this article reveals the substantial potential of this empirical approach. For decades now, researchers from a wide range of disciplines have been conducting studies of this type to determine how social norms influence behavior. Important results have been replicated across different experimental settings, yielding insights about normative behavior that are not identifiable with armchair observation. The empirical study of tort law has not yet reached this point. As Eisenberg and Engel point out, “[t]here is no more than a handful of pertinent experimental papers” that have tried to isolate the different types of reasons that motivate risky behavior threatening injuries to others. (P. 120.) Their article has not fully identified the contents of this “black box” within tort law, but their experimental design provides valuable guidance for what one hopes will be a quickly growing body of scholarship.