by Bernard E. Harcourt. Chicago: University of Chicago Press, 2006. 264pp. Cloth. $55.00. ISBN: 9780226316130. Paper. $25.00. ISBN: 9780226316147.

Reviewed by David S. Mann, Professor of Political Science, College of Charleston. Email: mannd [at]


AGAINST PREDICTION has its foundation in rational choice theory. Its reading audience should expand, given the topic and counterintuitive thesis and discussion. Having studied rudimentary rational choice theory in graduate school, but not having used it very much since, I took a moment to pause, especially since my primary scholarly focus is based on attitudinal models. First things first, however.

The book’s theme is simple: in pursuing what Bernard E. Harcourt refers to as actuarial methods, police and others are unwittingly assisting in increasing crime rates. If we really want to predict people’s behavior, we should use clinical methods rather than relying on statistical patterns based on probabilities. Taking four hypothetical examples in criminal law, Harcourt asserts that profiling – what he refers to as the actuarial method – in policing, sentencing, and parole decisions is a mistake. To behaviorists, this is totally opposite our training; to the curious, we look further, remembering always that it is with simplified assumptions that rational choice math works.

Harcourt, Director of the Center for the Study of Criminal Justice at University of Chicago, breaks down the text into three parts: the Rise of the Actuarial Paradigm, the Critique of Actuarial Methods, and Toward a General Theory of Punishing and Policing. I will summarize each section in as neutral a fashion as possible.

He begins his prologue with a definition of “actuarial methods in criminal law” as if it were to be found in Black’s LAW DICTIONARY: “the use of statistical rather than clinical methods on large datasets to determine different levels of criminal offending associated with one or more group traits, in order (1) to predict past, present or future criminal behavior and (2) to administer a criminal justice outcome” (p.1). Harcourt uses four not so hypothetical examples throughout the text to illustrate his counterintuitive notion. The first is the decision to release convicts early based on statistical prediction instruments. The second is the IRS decision to audit, which he reports to be based on a multiple regression equation. The third is taken from a sentencing guidelines commission report on likely recidivism, again, based on a risk assessment algorithm.
The last is the DEA attempt to identify “common characteristics of illegal drug couriers disembarking from planes at US airports” (p.15, footnote omitted). The author argues that risk assessment techniques, including profiling, dominant tools in the criminal justice system, ought to be discarded. [*535]

Before undertaking a literature review of the actuarial technique in chapters two and three, Harcourt offers three basic critiques which help the reader focus on his frame of reference. The first he characterizes as the “mathematics of criminal profiling” (p.22), which is really a microeconomic argument. He assumes that potential offenders are rational, respond to arrest probabilities, and know that law enforcement may be targeting them. Therefore the group’s offending rate falls toward that of the general population, at which time law enforcement should cease profiling (pp.22-23). He almost loses me when he addresses “relative elasticity of offending to policing” (p.23). Again, he assumes that potential offenders are rational, respond to punishment, and are deterred by the threat of more punishment. He asserts that “the use of actuarial methods . . . may increase overall crime in society” (p.26).

The second critique is what he calls “the overlooked social cost” (p.26), the incapacitation effect, where offenders are not rational and are unresponsive to policing. Here he uses a cost-benefit analysis. “[W]hen we profile, we are essentially sampling more from a higher-offending population[, i]nstead of sampling randomly . . . skewing our sample results” (p.29, italics original) So the criminal population (aka, sample) is disproportionate to the population as a whole and to those who manifest criminal behavior as a whole. This leads to the unfortunate social costs of disruption of families, work, and law-abiding life.

The third critique, which he calls “shaping the conception of justice” (p.31), addresses deterrence. I expected the conventional discussion and distinction between general deterrence and special or specific deterrence, commonly discussed in the most basic criminal justice course. That discussion did not take place, and as a consequence I am left a bit confused. The author discusses just punishment as that which deters, yet he then writes that utilitarian theories “fall aside because there is no good measure of deterrence” (p.33). We believe the higher offending groups should be targeted, they offend at higher rates, so we profile “the rich for IRS audits or minorities for drug searches” (p.33), and “we begin to feel morally righteous” (p.33). Just punishment, he argues, exists when “everyone who commits a crime (has) the same likelihood of being apprehended, regardless of . . . group trait” (p.38). Similarly, he argues to impose sentences based on an independent standard – several offered, for example, are “the harm principle, retribution, the seriousness of the offense” (p.38) and not to employ predictive devices.

Chapters Two and Three, as indicated above, review literature from the early 1900s that represents what I generically call the attitudinal or behavioral model. One of the more interesting arguments Harcourt makes is that individuals should be judged not on the basis of demographics, but rather using clinical methods directed at each individual offender, perpetrator, or accused. Curiously, that is the argument of Justice Powell for the 5-4 majority in MCCLESKEY v. KEMP. Recall that [*536] Justice Powell rules that McCleskey must show more than a statistical tendency that black defendants are more harshly judged than white. McCleskey must show that he himself was discriminated against. More curiously, it appears that Harcourt is unaware of that case – he does not mention it – nor does he mention in his literature review the famous Baldus study on which McCleskey’s argument stood.

Chapters Four and Five are where most of the economics and math reside, where Harcourt begins his critique of the actuarial method. These and Chapter Six elaborate on the critiques discussed earlier. I am guessing that the math is correct; I am guessing that the formula in Appendix B is correct as well. I personally quarrel with the author’s assumptions on which the math is based. For instance, he assumes that minorities offend at higher rates, that police are rational and efficient, and that potential offenders are rational and cognizant of deterrence models. He asserts that the status quo profiling techniques likely increase overall crime rates and have undesirable social costs, such as cultivated racial and other discrimination in society as a whole. I suppose there are data that show minorities to offend at higher rates than majorities (I question how we would address the state of New Mexico, where minorities are in the majority – but that may be trifling). There is insufficient time to discuss in depth why the other two assumptions may not be valid. As we all know, if the assumptions are correct, so should the math be correct. I simply question the veracity of the assumptions, thus putting the math, however it is shaped, into question.

Harcourt proceeds with Chapter Seven, “A Case Study of Racial Profiling” (pp195ff), where the first two critiques he mentioned very early in the tome are put to another test. Again using microeconomic modeling, some math, and lots of data, he argues that profiling “has allowed our war on drugs to turn into an offensive against black and Hispanic drivers” (p.214). A more fundamental notion of justice is that “all criminal offenders, regardless of race, ethnicity, gender, class or national origin, should face the same likelihood of being apprehended and punished for engaging in criminal behavior” (p.214). As a society, we presumably believe this, but do we “do” it as a matter of policy?

What about profiling in the name of counterterrorism? Harcourt reaches that question in Chapter Eight, after a discussion and mathematical critique of the Manhattan Bail Project, often a topic of consideration. The question of profiling Middle Easterners, or those who appear to be from the Middle East, also raises some eyebrows. He discusses offending differentials, comparative elasticities, group traits, and simply the question of whether “profiling young Muslim men in the New York subways will likely detect a terrorist attack” or instead “lead to recruitment of nonprofiled persons and the substitution of more deadly acts for subway attacks” (p.235). He admits we don’t know a lot more than we do know. I would ask the reader to remember when she first heard of the Oklahoma City bombing whether [*537] her first thought was “Middle Eastern terrorists”? Harcourt, strangely, does not mention Timothy McVey, which would illustrate his point, like his failure to offer MCCLESKEY as an example, mentioned earlier. As fictitiously addressed in the film “State of Siege,” we could round up a profiled class of persons and still fail to detect a terrorist act. Harcourt has a point here and could elaborate without the math.

The last chapter asserts one final time that, if we randomly search vehicles for drugs, randomly audit IRS returns, and base parole decisions using personal, clinical data rather than data from regression equations, we will have a better justice system where those who are interred actually reflect the US population as a whole, with much less devastating effect on social life. He could be right about parole decisions, provided that sufficient resources are devoted to the process. Is he right about drug couriers and IRS cheats?

Those whose focus is on behavioral and attitudinal models should read this book. Why? It is counterintuitive. It offers challenging assumptions. It raises questions about social discrimination. For instance, social discrimination may be enforced by law, as long as police concentrate their energies on those who, they think statistically, are the most likely offenders. Yet if we don the uniform of law enforcement, would we not concentrate on who we believe to be the likely offenders, using some profiling technique or other? Should we give this whole notion at least a sober second thought?

MCCLESKEY v. KEMP, 481 US 279 (1987).

© Copyright 2007 by the author, David S. Mann.