Printer Friendly

Review essay / profiling: theory and practice.

Frederick Schauer, Profiles, Probabilities, and Stereotypes Cambridge, MA: Belknap/Harvard University Press, 2003. xiii + 359 pp.

Frederick Schauer has a case to make. He has paid careful attention to the legal and public discourse in our country over the last decade, and he sees a disturbing trend. In our discussions of vital public issues, particularly (but not only) racial profiling, Schauer has become more and more alarmed by a trend toward what he calls "the primacy of the particular" [ix]--the tendency to believe that the important and just approach to decision making must involve the particulars of any one case or event. (1) Today, he says, using generalizations is out; instead, most people tend to think that "making the right decision for this case or on this occasion is the primary building-block of just behavior." [Ibid] Schauer wants to challenge this thinking which he feels is not just ineffective and badly misguided, but flat wrong. And he makes no bones about where he is headed. He lays out his thesis on the very first page of his book:

In this book I defend the morality of decision by categories and by generalizations, even with its consequent disregard for the fact that decisionmaking by generalization often seems to produce an unjust result in particular cases. And because I defend the morality of decisions by categories and generalizations, I look more sympathetically than is fashionable nowadays at profiles and stereotypes, which are little more than generalizations in street clothes. [Ibid]

Schauer deserves credit for taking on what is often an incendiary subject. Too frequently, thoughtful public discussion ends when someone utters the word "profile." No matter how carefully it is used, many who hear the term perceive it not as a description of a law enforcement tactic--a way that race has been used by some police officers and their departments--but as an accusation that everyone who wears a police uniform is a racist. The phrase has become, in so many circumstances and to so many people, a conversation stopper. But Schauer, a well-known authority on the First Amendment and the Frank Stanton Professor at the Kennedy School of Government at Harvard, is not intimidated. He wants, he says, to restore an understanding of an important aspect of how we make decisions--nothing less than the proper, and seemingly inescapable, place of generalization in both public and private decision making.

Schauer pursues his task with grace, skill, and well-honed reasoning. I will not demean his work by saying that it is well-crafted for an academic book. In fact, it stands out as well-written nonfiction, especially among those books that depend heavily on logical reasoning and theoretical discussion to sustain a complex argument. Would that every author of complicated nonfiction could do as good a job as Schauer does. He also has an extraordinary eye for the perfectly illustrative example, a skill no doubt sharpened by years in the classroom. Indeed, I often felt that I could hear Schauer think as I read his book; more than a few times, I found myself wishing that I had had the privilege and the pleasure of being his student.

The one reservation I had about the book, however, was that theoretical discussion can take one only so far on questions of social policy. Had Schauer gone a bit further and confronted some of the existing evidence of what happens when racial profiling gets used in the real world--on the highway, on the sidewalk, or in the airport--we might see that, even when logic might justify it, using these tactics in law enforcement turns out to be a dead end. As tightly reasoned as this book is, and even though Schauer actually argues against racial profiling, the book ultimately does not recognize that, contrary to what so many people seem to think, using profiling in practice as a method for creating public safety leads us in the wrong direction.

Particularity versus Generality

Schauer's argument directly challenges the current thinking that only particular judgments can be just. It is not clear, Schauer says, that "particularism" is a "moral imperative," as many who favor it seem to think. [19] Instead, he argues, one can make at least a qualified defense of "the practice of painting with a broad brush"--sometimes called actuarial decision making, or generalization, or more pejoratively, stereotyping or profiling. [3] It can, Schauer says, be perfectly rational to attribute the characteristics or sins of the few to the many of the group, even though it may be unfair to do so. [ibid.] But it is important at the outset, he asserts, to distinguish good generalizations from bad ones. Good generalizations have a sound statistical and empirical foundation; generalizations without such a foundation are spurious. While some of the discussion in the book concerns spurious generalizations, the bulk of it concerns the statistically-based "good" generalizations. [6] These nonspurious, sound generalizations may be universal--all bachelors are unmarried, for example--but they need not be, as long as, comparatively speaking, the generalization is more true when applied to the particular population in question than it is to the population as a whole. For example, the statement that "bulldogs have bad hips" is sound as long as the probability of a dog having hip problems is greater for a bulldog than for any other dog "given no other information about the breed" of the dog. [11] Schauer correctly points out that the ambiguous way that some have used words like stereotype, prejudice, and profile has at times obscured the important intertwined questions of universality, statistical soundness, and morality that we must face if we are to handle these issues with real intelligence and sensitivity. [18] In order to remove these ambiguities and sharpen the debate, Schauer asserts that he wants to confront the central question directly: Is it, in fact, wrong to rely on statistically sound generalizations constructed around a characteristic that is relevant for the purpose at hand, even though the trait is not universal in the group under scrutiny? In other words, can it ever be wrong to use a nonuniversal but empirically proven generalization when it is relevant to our decision making? This is, in fact, the heart of the issue, and Schauer faces it head on. It is Schauer's position that using these kinds of generalities is not just defensible, but unavoidable. We actually cannot do otherwise--even when we think (mistakenly) that we are only making particular judgments about the cases before us.

Schauer builds his case by using a string of very sharp and provocative examples, sometimes in the form of questions, and other times as illustrative stories. For example, he offers, we now have strong medical and scientific evidence that the presence of a particular genetic mutation that is more common in women of eastern European Jewish origin than the female population as a whole creates a significantly greater chance that these women will become afflicted with breast or ovarian cancer during their lives. This means that we--or more to the point, an insurance company--can make a statistically sound, empirically based, but nonuniversal generalization that women of eastern European Jewish origin are especially susceptible to these cancers. Would a denial of insurance coverage or an increase in premiums paid by all members of this group necessarily be unjust? [34-35] Or suppose that there is solid evidence that it is twenty percent more likely that an English football (soccer, to Americans) fan will commit an act of violence at a game than would any other randomly selected European football fan. Would a football venue in Europe therefore be justified in deciding to ban all English spectators, as some have? [37] Among the more entertaining of these examples is Schauer's discussion of pit bulls--or rather, the proliferation of municipal ordinances in the 1980s banning, restricting, or regulating the ownership of pit bulls. According to Schauer, the 1980s saw a steep increase in injuries (and even occasional deaths) due to attacks by dogs, especially pit bulls. Towns and cities reacted by enacting laws designed to make ownership of pit bulls less common (sometimes even illegal), or at least to force their owners to exercise much more care, purchase insurance, or the like. [55-56] Though these new laws were based on a statistically valid, nonspurious generalization about pit bulls and clearly meant to increase public safety from a known danger, the reaction to them was often quite negative. Pit bull owners, Schauer says, denounced these restrictions in terms that foreshadowed later debates about racial profiling: they described these anti-pit bull laws as "canine racism." [56] (To borrow a phrase from the humorist Dave Barry, Schauer is not making this up.) A somewhat more nuanced position came from the American Kennel Club, the national arbiter of dog breeds. The group's spokesperson pointed out that there were both many--perhaps an overwhelming number--of docile pit bulls, as well as a significant number of vicious dogs of other breeds. The pit bull ordinances were thus the wrong approach. Laws should not target any particular breed of dog, but only dogs whose past behavior proved they were dangerous--in essence, a particularistic versus a general approach.

The Kennel Club even came up with a catchy slogan to oppose the law, saying that laws should deal with "deeds, not breeds." [57] With these and many other examples--laws that do not allow even very careful and responsible fifteen-year-olds to obtain a driver's license, or that prohibit a thoughtful and politically aware seventeen-year-old from voting, or that require all pilots to retire at the age of sixty, no matter how good their reflexes and eyesight are--Schauer makes a strong case that it has simply become too easy to condemn a judgment by declaring it a stereotype or a profile or a prejudice. Making decisions by using generalizations, "popular rhetoric to the contrary notwithstanding, appears to be an unavoidable feature of our decision making existence." [76-77] In the final analysis, Schauer asserts, "all evidence is ... probabilistic" and even individualized decisions "turn out to rely more on generalizations than many people suppose." [101, 103] In fact, "many of the things we perceive as particular objects or particular observations turn out to depend on the kinds of generalizations that ... are much the stuff or ordinary reasoning." This, he argues, should make us "properly skeptical of a widespread but mistaken view that the particular has some sort of natural epistemological or moral primacy over the general." [106]

So far, so good, and a very persuasive case indeed. But Schauer is properly careful to concede that there may be certain types of generalizations--generalizations that are nonspurious, statistically valid, and relevant--that we should hesitate to use as the basis for public-policy decisions. These are generalizations based on gender and--perhaps surprising to those who anticipated from the book's title and its first few chapters that Schauer would emerge as a champion of racial profiling--race. Schauer argues that members of these categories may share characteristics that might support nonspurious generalizations that are actually relevant to the decision at hand, but it would be a mistake to use them in this way. Making these kinds of generalizations can amount to true and invidious discrimination of a type that our society should not allow. This is because the empirical basis for generalizations based on race may in fact come from the cultural context. For example, the relative inability of black Americans to pass a police department's written test may come from pre-existing social forces like poor educational systems that blacks find themselves in, and not biological factors, and relying on generalization ("blacks can't pass this test at the same rate that whites do") only reflects and reinforces the preexisting biases in the society. As Schauer says, "the generalization ... probably [has] a nonspurious empirical basis, but the cultural contingency of that empirical basis, at least in the context of race or national origin, makes it wrong to translate the empirical generalization into public policy." [141]

This is particularly dangerous with race-based generalizations, Schauer says. Years of well-accepted research has shown that race, as a highly visible, salient characteristic, will almost inevitably be overvalued by observers as a characteristic with the ability to predict something important, especially something negative. The upshot is that most people "are consequently likely to amplify the extent to which members of a race other than their own are represented in a larger population with negative attributes." [179] This helps explain why police officers are more likely to think that African Americans, the "out group," are involved in criminal activities than they actually are, and to take enforcement action as a result. [Ibid.] This is a key insight, and it may go some considerable distance in explaining the staying power of racial profiling as an accepted law enforcement tactic despite considerable evidence in the last few years which fundamentally undermines the conventional wisdom that racial profiling helps police catch crooks. Schauer shows how the historic, psychological, and cultural context of race (and of gender) mean that we should not base our policy decisions on them.

Racial Profiling

All of this is very well presented, precisely argued, and useful to anyone interested in how we make our personal and national decisions. But a few seams in the argument begin to show when Schauer comes to specific discussion of the subject that most directly provoked his exploration of the generality vs. particularity debate: racial profiling in the criminal justice system. And this, unfortunately, makes the book less useful to practitioners and policy makers than it might otherwise be.

Schauer begins his discussion of racial profiling by making a distinction between what he might have called real profiling and profiling in name only. Real profiling involves study of all of the relevant data in a mix of possible predictive factors that might have a tendency to point toward some particular kind of criminal behavior. The best known example of this type of profiling is the so-called serial killer profile, put together by the FBI's behavioral scientists. (2) Those who constructed these profiles studied every piece of data they could find on serial killings, whether the cases were open or closed. They also interviewed at length every serial killer in custody who would talk to them, and gathered information about the crimes, the perpetrators, and the surviving victims from every conceivable source. The result was the systematic collection of many millions of pieces of data. These data were then rigorously analyzed so that correlations and clues might be sifted out. The painstaking efforts to handle the profiling process correctly did not end with dredging the important serial-killer indicators out of the mountains of data collected. The FBI personnel were also careful with the way they used the indicators. They did not, for example, simply look for persons who shared these characteristics among thousands of people passing through areas where suspected serial killings had happened. Rather, they made use of the profile only after they had narrowed the field of suspects to just two or three or half a dozen, in order to help them decide who among this handful of people might be the most likely suspects. All in all, a real profile takes an immense amount of work to prepare correctly. It must be constructed through gathering of data from all possible cases, those data must then be thoroughly analyzed, and the profile must then be used carefully.

Schauer contrasts such real profiles with profiles in name only. These are not the kind of models of criminal behavior rigorously constructed from all of the available data; rather, they are much more likely to come from stationhouse folklore or individual police officers' recollections of the commonalities of their biggest arrests. (In fact, one of the first drug courier profiles--one studied extensively by the U.S. Drug Enforcement Administration when it decided to try to transfer its efforts to interdict drugs in airports onto the nation's highways--was constructed by a Florida state trooper in exactly this way. Bob Vogel, a Florida Highway Patrol officer who later became nationally famous for his highway drug interdiction work, put together his own list of "cumulative similarities" by studying the commonalities among his thirty biggest highway drug arrests. (3) He did not think it important to consider his (far more common) stops and searches that did not turn up drugs, or the enforcement efforts of other troopers. This meant, of course, that the data he analyzed could be called incomplete, highly selective, biased, or worse, but this did not stop the DEA from relying on it.) These types of profiles feature neither thorough collection nor rigorous analysis of all pertinent data, and could include any kind or number of "factors," without any meaningful attempt at quantification. Moreover, they have been used as little or as often, as haphazardly, and for as many situations as officers like; no consideration has been given to what limits on their use might be appropriate. The most egregious example, and surely the best known of such profiles, Schauer points out, was the use of profiling by state troopers in New Jersey. (4) Correctly noting that New Jersey was hardly alone among states or police agencies that used such racial profiles, Schauer writes, "there is no indication ... that New Jersey had performed the statistical analysis necessary to produce such a formal profile. Indeed, there is no evidence that the New Jersey state police had produced any kind of formal profile, whether statistically grounded or not. In an important sense, therefore, the so-called profiling in New Jersey was worlds apart [from real profiling].... Indeed, the original New Jersey procedure ought not to be glorified by referring to it as a 'profile,' for it would be more accurate to call it a 'guess.'" [192]

This is both true and well said, and Schauer therefore limits his consideration of the question of profiles to real profiles--most importantly, the profile used by the U.S. Customs Service in the 1990s and the CAPPS II profiling system proposed for airports to detect possible terrorist suspects in the wake of the September 11, 2001, terrorist attacks. He calls these "profiles that [include] race" versus the New Jersey-type "racial profiles." [195] For Schauer, the New Jersey-type racial profile is hardly worth thinking about; we cannot even tell if there might be some empirical evidence to sustain the claim by proponents of the practice that there is a correlation between minority race and drug courier or other criminal behavior sought out by troopers, because "the empirical evidence is spotty." [194] Schauer therefore limits the meat of his profiling discussion to "real profiles": profiles that are correctly prepared and used, and that might include race or ethnic origin in a mix of other factors. Such is the case with U.S. Customs profiling and CAPPS II.

While theoretically correct and logically defensible, this approach threatens to define away the problem as the public understands it. In the public discourse that has led to Schauer's book, it is the weak form of profiling--New Jersey-type racial profiling, without any real statistical basis--that people mean when they say racial profiling. Failing to deal with this issue in a thorough enough fashion risks generating misunderstanding, through no fault of the author. Many may hear Schauer to say that he supports New Jersey-type racial profiling, despite the fact that he actually says nothing of the kind--he is even ready to condemn most race- and gender-based profiling even when it has a statistical basis and is relevant. This would be an unfortunate misunderstanding, but not a surprising one. Schauer's book has already been written about--unintentionally--in a way that may lead to the inaccurate impression that it is a "pro racial profiling" book. In an article in 2003 describing his book, The New York Times said that one of Schauer's goals in the book "is to fine tune 'profiling' and rescue it from being pejorative." (5) A reader could be forgiven for thinking that Schauer favors the sloppy, non-statistically based, New Jersey-style of racial profiling. Of course, he does not--a fact not reported until a number of paragraphs further into the article. (6) This is surely not Schauer's fault, and the reporter who wrote the piece did articulate the whole of Schauer's more nuanced view, even if important parts of it came at disparate points in the article. But if such an incorrect impression can come from the careful pages of the august Times, publications less sensitive to getting Schauer's whole thesis before the reader seem likely indeed to characterize it incorrectly. This, I would argue, is because the argument in the book fails to meet squarely the central failing of racial profiling as the public understands the phrase.

Looking at the studies of New Jersey-type racial profiles shows a central fact: when police use race as one factor--not the only factor, but as one factor among others--in deciding who to stop, question, and search, they do not improve or boost their crime-fighting results, as so many people believe. In fact, when they use race as one factor in the mix, police officers find and arrest fewer bad guys, get fewer guns off the street, and confiscate drugs less often than they do when they do not use race as a factor (that is, when they stop, question and search whites also). (7) In study after study, similar results have upended the conventional wisdom behind racial profiling--that because blacks, or Hispanics, or other minorities show a higher degree of involvement in crime than do whites, it simply makes sense for police to stop and search greater numbers of people sharing these racial or ethnic characteristics because this will allow police to find more bad guys. Not true; police actually find fewer bad guys and less contraband, by a statistically significant, measurable amount, than they do when race is not used. (The use of race as described above does not include physical descriptions of observed suspects, which has nothing to do with any kind of profiling; using race to describe a known suspect is good police work. Rather, racial profiling is when race is used among other factors to attempt to predict which people in a large population might be involved in unknown crimes--drug smugglers driving down an interstate highway, for example.) Given the number and consistency of these studies, it is somewhat surprising that Schauer chose not to comment on them at all; likely as not, he thought it unnecessary, since he regards these kinds of profiling efforts as nothing more than a "guess."

Importantly, these studies tell us something quite important that undermines the practical significance and bottom-line acceptability, not just of New Jersey-style profiling, but also of Schauer's "real profiles"--those based on valid statistical basis and relevant to the question at hand. He is willing, he says, to accept these kinds of profiling systems, using as his examples the U.S. Customs Service drug courier profiles and the CAPPS II airline anti-terrorist profiling systems, because these kinds of formal, written, and statistically validated systems would improve law enforcement's effectiveness while guiding the use of discretion by the front-line law enforcement officials who would use them. But the studies of law enforcement effectiveness using race suggest a far more troubling outcome, even when the efforts are based on statistically valid profiles. Schauer actually discusses one important reason that the use of race would not help, and would likely undermine, law enforcement efforts: race is far too likely to be overvalued as a predictive indicator. This is because the majority population which does not share the racial characteristic tends to view it negatively, and to do so quite strongly. Schauer concedes that this may be reason enough to consider banning or at least regulating the use of race even in statistically valid profiles. But there is more to the point than this. The predictive value of race is not only almost inevitably overstated; it also distracts law enforcement from engaging in what really counts in ferreting out otherwise unknown criminal activity: close observation of behavior. It is behavior that counts in detecting such criminals, whether drug couriers or terrorists, and looking at race as a proxy for this behavior will produce not only many false positives--stopping and searching many people who have nothing to do with the particular behavior--but also false negatives, as drug mules or terrorists who do not "look like" our mental images of those who we "know" perpetrate these crimes slip through the net, because we are looking for someone who looks different.

It turns out that these very problems have, in fact, been among the biggest issues in the implementation of even the "real" profiling systems to which Schauer gives his limited approval. During the course of the writing of this essay, the CAPPS II profiling system, designed to pinpoint those airline passengers at highest risk of terrorist involvement, was scrapped by homeland security authorities after months of development work and millions upon millions of federal dollars, at least in part because of concern about how its use of race or ethnicity would effect both passenger privacy and the system's effectiveness. (8) Schauer's example of the Customs Service in the late 1990s is even more troubling. Schauer states that a few customs agents had apparently begun to use race or ethnicity to decide which passengers to subject to so-called "intrusive searches." The problem was not the use of race in profiling, he argues, but the lack of a written, well-thought-out, statistically valid profile for agents to follow, since the Customs Service did not tell its agents to use race. Schauer says that if the agency had done so, after finding a valid statistical predictive value to race in proper studies, this would have prevented the problem. Unfortunately, this is incorrect. Schauer is insufficiently skeptical of the Customs Service's denial that its policy allowed profiling using race. In fact, the evidence indicates that race and ethnic appearance were indeed among the forty-three "indicators" that the agency told its agents to use before these practices were exposed. (9) At the very least, even if using race this way was not official policy, allowing such a set of attitudes to fester when the Customs Service's own data showed what was happening indicated a management failure at the highest level. Schauer is correct when he points out that, with newly appointed Customs Commissioner (now the Commissioner of the New York City Police Department) Raymond Kelly at the helm, the Customs Service did a much better job of catching drug smugglers after its use of profiling was exposed. But this was not because the Service suddenly turned to a better, more formal, and statistically valid use of race. It was, in fact, because Kelly forced the Service to give up its sloppy use of race and concentrate fully on analysis of behavior. In other words, Kelly's solution to the Customs Service's profiling problem was not better use of race, but no use of it, and the concentration of law enforcement attention on what really counted.

All of this teaches us important lessons. The use of race or ethnic appearance in street- (or airport-) level profiles is part of a set of tactical decisions concerning how to meet a particular threat. Using race in such profiles now to intercept terrorists, no matter how acceptably obvious this seems to the public in general, is only an adoption of the same tactic that has failed against drug couriers and others. It cannot be rescued from ineffectiveness either by more rigorous crafting or by more careful use when race is made part of the mix of factors used. When we include race as one of the proxy indicators of criminality, we overvalue it, and it distracts us from watching behavior. It is the failure to completely come to grips with this issue that makes this otherwise excellent book a bit less useful than it otherwise might have been for practitioners and policy makers.

Still, Schauer has made a significant contribution. This is a well-done book that poses an important challenge to conventional--and somewhat misguided--ways of thinking. In that respect, it falls within the finest traditions of scholarship. For that alone, all of us concerned with how we make decisions, and how we debate issues of public policy, owe Schauer a great deal.


(1) Frederick Schauer, Profiles, Probabilities, and Stereotypes (Cambridge, MA: The Belknap Press of Harvard University Press, 2003). [All bracketed page numbers in the text refer to this volume.]

(2) For more on the construction of the serial killer profile, see David A. Harris, Profiles in Injustice: Why Racial Profiling Cannot Work (New Press, 2002), 18-19, and the sources cited therein.

(3) Harris, Profiles in Injustice, 21-23.

(4) See, e.g., Peter Verniero and Paul Zoubek, Interim Report of the State Police Review Team Regarding Allegations of Racial Profiling, 29 April 1999 < intm_419.pdf>

(5) Felicia R. Lee, "Discriminating? Yes. Discriminatory? No," New York Times (December 13, 2003): B7.

(6) Lee, "Discriminating?"

(7) See, generally, Harris, Profiles in Injustice, 71-84.

(8) Matthew L. Wald, "A Plan for Screening At Airports Is Dropped," New York Times (July 25, 2004): 3, sec. 5.

(9) See generally Harris, Profiles in Injustice, Ch. 8, "A Case Study: How One Police Agency Changed for the Better." For example, in the course of my research, I interviewed a reporter at an NBC television affiliate, who told me of having seen Customs Service training materials that taught recruits to use race.

David A. Harris, author of Profiles in Injustice: Why Racial Profiling Cannot Work (2002) and the forthcoming Good Cops: The Case for Preventive Policing, is Balk of Law and Values, University of Toledo College of Law, and Soros Senior Justice Fellow at the Open Society Institute of New York.
COPYRIGHT 2004 Institute for Criminal Justice Ethics
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2004 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:IN THE LITERATURE; Profiles, Probabilities, and Stereotypes
Author:Harris, David A.
Publication:Criminal Justice Ethics
Article Type:Book Review
Date:Jun 22, 2004
Previous Article:Communicative punishment and the role of the victim.
Next Article:Publications received.

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |