Statistics, Not Experts
Meadow, William, Sunstein, Cass R., Duke Law Journal
Legal proceedings, both judicial and administrative, frequently raise questions about what people, including experts, ordinarily do. (1) If a doctor is accused of negligence, for example, it is necessary to know about the customary practice of doctors. (2) Of course, negligence judgments depend, at least in part, on an assessment of ordinary practice. (3) But how is ordinary practice by professionals or others to be assessed?
In trials and on appeal, the basic answer is that the assessment comes via statements from expert witnesses who describe the ordinary practice. (4) There can be no doubt that experts know a great deal about topics on which ordinary people lack information. But experts, no less than other people, are subject to predictable biases. (5) Their judgments about risk are affected by the same heuristics and biases to which most people are subject, even if (and this is a disputed question) expertise tends to reduce the most serious errors. (6)
To correct the resulting problems, we offer a simple proposal: The legal system should rely, whenever it can and far more than it now does, on statistical data about doctors' performance rather than on the opinions of experts about doctors' performance. For the first time, it is becoming possible for law to rely on this evidence, for the simple reason that it is becoming increasingly available. If our argument is convincing in the medical context, it should apply in many other settings in which experts are asked to testify about negligence or deviations from ordinary practices. In many settings, the fallible opinions of isolated experts should be supplemented or replaced by statistical data. Those opinions should be seen as a kind of crude second-best, far inferior to the data that it approximates. A step toward reliance on such data would dramatically increase the sense and rationality of tort law, and perhaps other areas of law as well.
We are hopeful that the use of statistical data will simplify some of the central issues in litigation and reduce the role of strategic behavior in the litigation process. Battles between isolated experts will be easier to mediate when there is a large pool of evidence on which to draw. But our larger claim is that by using statistical data, the legal system will reach far more accurate results. If the law is seeking to determine the standard of care, it should not depend on fallible memories and recollections of local experiences. It should draw instead on more global evidence of the kind that is increasingly available for use in court and elsewhere. (7)
I. EXCESSIVE OPTIMISM ABOUT RISKS
A. In General
It is now well known that most normal people tend to be "risk optimists," in the particular sense that they believe themselves to be relatively insulated from risks that are faced by others who are similarly situated. (8) This is one of the most robust findings in cognitive psychology. For example, ninety percent of drivers believe that they are safer than most drivers and less likely to be involved in a serious accident. (9) Most people believe that they are distinctly unlikely to be subject to various risks, such as cancer, heart disease, and divorce. (10) In some important domains, people appear not only to believe themselves less likely than others to fall prey to certain risks, but also to underestimate the actual risk to which they are subject. (11) With respect to automobile accidents, people believe that the danger they face is less than it is as a matter of statistical fact. (12) Smokers appear to know the statistical risks of smoking, (13) but they believe that they are less likely than most smokers to fall victim to the various risks. (14) In one study, less than half of smokers surveyed believed that they had a higher-than-average risk of cancer or cardiovascular disease; indeed, most heavy smokers (at least forty cigarettes per day) believed that they were not at any increased risk. …