Is incoherence outrageous?

• Author: Sunstein, Cass R.

For their valuable comments, we are most grateful to Cary Coglianese, (1) Mark Kelman, (2) and the team of Theodore Eisenberg, Jeffrey Rachlinski, and Martin Wells (hereafter ER&W). (3) ER&W offer a great deal of information about the real world of punitive damages, presenting a mixed verdict on our basic claims. Coglianese provides an impressively detailed account of the problem of achieving coherence in the world of administrative penalties. Kelman questions the importance of coherence and also urges that in many cases, people's judgments may be better if they lack a global perspective. We are pleased to say that we have few significant disagreements with the three commentaries. In brief, we think that ER&W's evidence is fully consistent with our basic claims, despite a conservative test; that Coglianese has elaborated helpfully on the problem of achieving coherence in civil penalties; and that Kelman's basic arguments should not be taken to dispute our claims, which are less different from his own than a reader of his commentary might infer. Apart from these points, which we elaborate below, we offer a more general suggestion, one that is perhaps insufficiently emphasized in our main paper: Even if it is demonstrated, incoherence usually is not taken to be outrageous. The persistence of incoherence, in many domains of the law, is closely connected with this psychological fact.

1. EXPERIMENTS AND THE REAL WORLD

   ER&W provide an instructive look at the real-world of punitive damage awards. The most novel part of their essay offers some intriguing data analysis, suggesting that the incoherence that we describe actually exists, and hence that there is a "degree of support" (4) for our claims in real-world data. But thinking that few "complaints about a legal system resonate louder than charges of incoherence," (5) ER&W object that if our argument is right, we should "be observing a legal system on the verge of collapse," (6) because any "society would find a torrent of incoherent legal judgments to be unfair, unjust, unpredictable, and unmanageable." (7) And in a discussion of prior findings that appears to be in some tension with their new analysis, ER&W contend that the system of punitive damages is actually quite coherent, showing a high degree of predictability. Of the three comments, ER&W's raises the most complex issues, and we shall discuss those issues in some detail.

   1. A New Analysis

      ER&W's innovative empirical analysis is based on their (welcome) claim that our essay can be used to generate testable hypotheses. They offer three such hypotheses. The first involves juries. In their view, we would predict that juries, which make decisions in isolation, would produce, across diverse categories of cases, the same statistical relationship between punitive awards and compensatory awards. The second hypothesis involves judges, who observe cases from many categories. ER&W think that we would predict that judges will be engaging in cross-category comparisons and hence that the effects of case categories "should be greater in judge-tried cases than in jury-tried cases." (8) Perhaps most interestingly, ER&W urge that if we are right, judges will offer higher awards than juries in cases involving the more prominent category of harm. The reason is that in cross-category comparisons, the awards in the more prominent category typically increase; and judges, in ER&W's view, are engaged in cross-category comparisons. ER&W rightly say that the third hypothesis seems counterintuitive, even a bit odd. Could it be right to suggest that when the harm is prominent, judges will produce higher damage awards than juries?

      Comparing actual awards in cases involving bodily injury to actual awards in cases not involving bodily injury, ER&W find support for all three hypotheses. For juries, the ratio of compensatory to punitive awards is about the same between cases involving bodily injury and cases not involving bodily injury. For judges, punitive awards have a far higher ratio to compensatory awards in cases involving bodily injury. Most strikingly, the counterintuitive prediction is vindicated. The punitive award is likely to be much higher, compared to the compensatory award, when judges are making the decision than when juries are doing so.

      ER&W have done a very creative job of adapting our account of punitive damage assessment to produce testable hypotheses for their extensive data set. We are of course delighted that our model passed the tests that they set for it. Still, we should point out that these tests are overly severe and that our model might have failed them even if it is basically correct. Three points deserve mention. First, ER&W's analyses use compensatory damages as a proxy for the "outrage x harm" (9) component of our model. But in our model, as in the law, compensatory damages are linked only to the harm, not the outrageousness of the defendant's behavior. The outrageousness (often called egregiousness) of the defendant's behavior is not represented in the ER&W analyses, though it is central to our model of how punitive damages are set. Second, ER&W's analysis takes for granted that judges consider each case in a broader context, one that includes cases from other categories. What is certainly tree, and what we emphasized, is that judges could consider cases through a wider viewscreen because, unlike juries, they have encountered a varied set of cases. However, we are not aware of any direct evidence that judges do spontaneously consider particular cases in a broad context. Indeed, our original paper drew on work by Guthrie, Rachlinski, and Wistrich in raising doubts about precisely that point. (10)

      The Supreme Court apparently has similar doubts about the width of the screen judges use. In Leatherman the Court found it necessary to require even appellate judges to engage in an explicit comparison (i.e., to use a wide screen) in evaluating the constitutionality of damage awards, rather than trusting their experience. (11)...