Suppose it is known that England supporters are more likely to behave violently at international football games than supporters of other national teams. If other countries impose a ban on English people watching football games, does this treat the law-abiding majority of England supporters unjustly? If statistics show that possession of illegal drugs is more common among young black males than in the general population, is it legitimate for the police to use a person's youth, blackness and maleness as reasons for stopping and searching him?
Frederick Schauer discusses a wide range of issues of this kind, and tries to find general principles to govern the ways we treat them. His thesis is that "decisions by categories and by generalisations" are defensible, and that people who are disadvantaged by the categories they fall into have no grounds for complaint.
In support of this, he marshalls some strong arguments, mostly pragmatic rather than principled. There is no way of avoiding generalisations in decision-making and legislation, since lines always have to be drawn somewhere. No evidence, even from eye-witnesses, is absolutely reliable; thus, all evidence derives its significance from statistical generalisations about reliability.
If each case is to be treated on its merits, someone has to be entrusted with judging those merits; since errors are unavoidable, the overall result of this may be more arbitrary than a procedure based on clearly defined categories.
Schauer writes clearly, with the slightly studied informality one might expect of an attorney's address to a jury or an introductory lecture from a law professor (which is what he is). Sometimes this legal viewpoint becomes a limitation. For example, in discussing the use of generalisations in stop-and-search procedures, he fails to consider a point that would occur immediately to a game theorist - that genuinely random selection is a defence against the strategic thinking of one's opponents. Thus, he recommends that when airport security staff choose which passengers to search, they should focus on categories that have been disproportionately represented among previously identified terrorists - young males of Middle Eastern appearance with one-way tickets and no check-in baggage. But it would be easy for terrorists to respond to this policy by using older, lighter-skinned female operatives and instructing them to buy return tickets and carry large bags.
Schauer also misses the significance of Bayes's theorem in interpreting statistical evidence. Suppose a defendant's DNA matches a sample taken at the crime scene, and it is known that the probability that a randomly selected person has the same DNA is one in a million. Does that establish beyond reasonable doubt the guilt of the defendant? Not if he was arrested only because he was the first person to be found in a search for someone with matching DNA.
Schauer presents himself as making a controversial stand against fashionable views, but his book might have been more interesting had he been braver. Whenever he seems on the point of reaching a particularly provocative conclusion, he draws back. While he endorses the use of statistical evidence in civil cases, he does not extend the argument to criminal cases, even though it seems equally applicable to them.
Curiously, one of Schauer's most controversial claims is orthogonal to the main argument of the book. He endorses Charles de Gaulle's remark about the impossibility of governing a country with 246 varieties of cheese as if it had been meant literally as a hypothesis in political science. Schauer takes the promotion of uniformity for its own sake to be a proper objective of national law. Is this a US constitutional lawyer's perspective on nationhood? Or just a sign that Schauer lacks a sense of irony?
Robert Sugden is professor of economics, University of East Anglia.
Profiles, Probabilities and Stereotypes
Author - Frederick Schauer
Publisher - Harvard University Press
Pages - 359
Price - £19.95
ISBN - 0 674 01186 4