At least through the beginning of the twentieth century, knowledgeable public intellectuals discussed deterring crime in a conceptually wide-ranging and popularly accessibly way. Consider the example of Oliver Wendall Holmes, Jr. By 1897, Holmes was the well-regarded author of the legal text, The Common Law, and a justice on the Supreme Judicial Court of Massachusetts, the highest court in Massachusetts. In a public address in 1897 at the dedication of a new Boston University Law School building, Holmes declared:

For the rational study of the law the black-letter man may be the man of the present, but the man of the future is the man of statistics and the master of economics. ^

Like Beccaria and Bentham, Holmes questioned traditional ideas and practices and encouraged wide-ranging, unsentimental discussion of penal policy. Holmes asked:

What have we better than a blind guess to show that the criminal law in its present form does more good than harm? I do not stop to refer to the effect which it has had in degrading prisoners and in plunging them further into crime, or to the question whether fine and imprisonment do not fall more heavily on a criminal’s wife and children than on himself. I have in mind more far-reaching questions. Does punishment deter? Do we deal with criminals on proper principles? ^

Holmes here gender-stereotyped the criminal and invoked the popular belief that prisons, by permitting association and communication among prisoners of all classes, serve as schools of crime. Holmes then passed on to consider deterrence in relation to personal types, imitation, and fashion:

A modern school of Continental criminalists plumes itself on the formula, first suggested, it is said, by Gall, that we must consider the criminal rather than the crime. … If the typical criminal is a degenerate, bound to swindle or to murder by as deep seated an organic necessity as that which makes the rattlesnake bite, it is idle to talk of deterring him by the classical method of imprisonment. He must be got rid of; he cannot be improved, or frightened out of his structural reaction. If, on the other hand, crime, like normal human conduct, is mainly a matter of imitation, punishment fairly may be expected to help to keep it out of fashion.

Holmes’ simile “as that which makes the rattlesnake bit,” followed by the direct, forceful, “He must be got rid of” surely is effective popular communication. Holmes combined such language with display of his knowledge of the current scholarly literature:

The study of criminals has been thought by some well known men of science to sustain the former hypothesis. The statistics of the relative increase of crime in crowded places like large cities, where example has the greatest chance to work, and in less populated parts, where the contagion spreads more slowly, have been used with great force in favor of the latter view. But there is weighty authority for the belief that, however this may be, “not the nature of the crime, but the dangerousness of the criminal, constitutes the only reasonable legal criterion to guide the inevitable social reaction against the criminal.” (Holmes’ footnote: Havelock Ellis, “The Criminal,” 41, citing Garofalo. See also Ferri, “Sociologie Criminelle,” passim. Compare Tarde, “La Philosophie Pénale.”) ^

Consider Holmes’ authorities. Franz Josef Gall, a German physician, founded phrenology, which he initially called craniology. Phenology replaced analysis of rational choices with biological description. Jean Gabriel Tardé emphasized social imitation as a cause of crime. Enrico Ferri focused on social groups in analyzing crime. Both harshly criticized Beccaria and Bentham’s individualistic analysis of criminal rationality. With respect to considering the criminal rather than the crime, Holmes concluded with a quotation from Henry Havelock Ellis. Ellis was paraphrasing Raffaele Garofalo. Garofalo’s work emphasized the biological characteristics of criminals and provided intellectual foundations for indeterminate sentencing and individualized parole decisions. Garofalo strongly criticized the assumption of free will at the foundations of economic analysis of crime.

Holmes considered crime in terms of biological types, imitation, and fashion. That’s much different from penal thinking of the early masters of proto-economics. Holmes cited no statistics. More generally, his address served a broad audience with appreciation for display of learning and without strong claims to narrow, technical expertise. Imagining the conduct of the “bad man” — the person like everyone except you and me and all women — is an intellectual exercise accessible and appealing to everyone.

Academics trained in economic analysis, as if controlled by an invisible moderator of disciplined discussion, have vigorously competed to produce the best estimate of the elasticity of crime with respect to imprisonment. On the supply side, regressing is a predominate practice among empirical economists. An elasticity is a single coefficient produced cheaply in a log-linear regression. Since an increase in crime might naturally be associated with a subsequent increase in imprisonment, identifying the effect of imprisonment on crime raises the sort of causality questions that have been central to expertise within the discipline of economics in the second half of the twentieth century. Estimating elasticities thus provides opportunities for high-interest investments in expertise. On the demand side, the elasticity of crime with respect to imprisonment maps directly into summary statements that support a strong claim to policy relevance, e.g. “a 1% increase in the prison population will produce a 0.3% reduction in crime.” Such claims help to stimulate demand for expert quotations in newspaper articles and interviews on television. Scholars seeking value in economic analysis have rationally allocated considerable research effort to estimating elasticities.

The value of estimating elasticities does not lie within the foundations of economic analysis. Paul Samuelson, winner of 1970 Nobel Prize in Economic Sciences and perhaps the most academically influential twentieth-century economist, called elasticities “essentially arbitrary” and “more or less useless.” In a more nuanced evaluation, he stated, “On the whole, it appears that their importance is not very great except possibly as mental exercises for beginning students.”^ Consequences of changes in a control variable could be estimated, predicted, and described in a wide variety of ways. An elasiticity is merely one, particular, highly specific way of describing consequences.

Apart from disciplinary credit, the utility from estimating the elasticity of crime with respect to imprisonment appears astonishingly small. Prior to 1968, scholars had done little work empirically evaluating the effect of deterrence on crime. In the subsequent decade, at least twenty-four empirical studies addressed this question. A literature review, sponsored by the U.S. National Research Council, reported in 1978:

Despite the intensity of the research effort, the empirical evidence is still not sufficient for providing a rigorous confirmation of the existence of a deterrent effect. Perhaps more important, the evidence is woefully inadequate for providing a good estimate of the magnitude of whatever effect may exist.^

The associated report, the U.S. National Research Council’s Report of the Panel on Research on Deterrent and Incapacitive Effects, offered a carefully worded declaration of non-knowledge:

we cannot yet assert that the evidence warrants an affirmative conclusion regarding deterrence. …Our reluctance to draw stronger conclusions does not imply support for a position that deterrence does not exist, since the evidence certainly favors a proposition supporting deterrence more than it favors one asserting that deterrence is absent.^

A book-length study of criminal incapacitation published in 1994 indicated that the state of knowledge had changed little:

Unfortunately, virtually nothing is known about the size of even the aggregate elasticity. Literally dozens of {estimates of the elasticity} have been calculated on the basis of econometric analysis, but the prevailing opinion is that none of these estimates are persuasive^

By 2000, more than 100 scholarly studies had attempted to test for deterrence.^ Through more than two decades of estimating the elasticity of crime, many technical improvements were declared in the scholarly literature. These improvements led to an important conclusion:

The ultimate futility of our analysis is at last revealed. We can be nearly certain the elasticity lies somewhere between -0.1 and -0.5, and we can be fairly sure it lies between -0.2 and –0.4, but we have no idea at all whether it is greater than or less than -0.265.^

The figure of -0.265 comes from the author’s back-of-the-envelope calculation of the elasticity threshold for the cost effectiveness of prison construction. Its three-digit precision, which is not convincingly supported, seems merely illustrative. A literature review in 2010 found relatively little evidence of severity of punishment having a deterrence effect, but found relatively strong evidence that the probability of punishment deters. The researchers concluded, “we see much need for new research.”^

Estimating the elasticity of crime with respect to deterrence isn’t a socially cost-effective contribution to public discussion of criminal justice policy. Almost everyone has almost nothing to say about deterrence when it is described as an elasticity. Public discussion of criminal punishment also commonly includes justifications that are unrelated to deterrence.^

In 1996, economist Steven Levitt published in a leading economics journal an analysis of elasticities of crime. Levitt’s abstract began by declaring his technical innovation for treating intricate causal relations:

Simultaneity between prisoner populations and crime rates makes it difficult to isolate the causal effect of changes in prison populations on crime. To break that simultaneity, this paper uses prison overcrowding litigation in a state as an instrument for changes in the prison population.

Then his paper reported his elasticity estimates:

The resulting elasticities are two to three times greater than those of previous studies. A one-prisoner reduction is associated with an increase of fifteen Index I crimes per year.^

A one-prisoner change in imprisonment is a situation that a prosecutor might ponder in considering whether to charge someone with a wide array of crimes and then to seek a guilty plea on at least one of them. Is Levitt’s analysis relevant to that one-prisoner prosecutorial decision? More generally, in what social, criminal, and prison circumstances, at what magnitude of imprisonment, and at what scale change in imprisonment is Levitt’s analysis empirically relevant? An elasticity of crime abstracts from those questions. Abstracting from such questions promotes the value of economic analysis by enhancing its apparent generality.

Consider some circumstances concerning imprisonment in the U.S. In the U.S. from 1971 to 1996, the number of prisoners sentenced to more than one year of imprisonment grew an unprecedented 474%. The share of prisoners with drug offenses as their most serious offense rose from 9% to 24%. At year-end 1996, the U.S. had in prison 611 persons per 100,000 residents. U.S. incarceration prevalence was about the same as incarceration prevalence in Russia, which had the highest incarceration prevalence about 1996 in a dataset covering nearly all countries. Moreover, 65% of countries had an incarceration prevalence less than or equal to 150 persons per 100,000 residents. Thus, in 1996, the extent of U.S. incarceration was extremely high relative to U.S. history and to contemporary international comparisons. Nonetheless, the U.S. has continued to incarcerate a larger share of its population. In 2010, U.S. incarceration prevalence reached 733 prisoners per 100,000 residents.

An elasticity of crime obscures the specific historical circumstances. In a paper published in an economics journal in 2004, Levitt used his 1996 estimates of the elasticity of crime to explain part of the crime reduction in the U.S. from 1991 to 2001. Within that same analysis, those elasticities do not contribute to a satisfactory explanation for U.S. crime trends from 1973 to 1991. However, in his 1996 paper, Levitt estimated his elasticities of crime using U.S. state-level data from 1971 to 1993. Levitt’s elasticities of crime apparently work better in explaining crime across years for which they are less directly relevant.^

The value of an elasticity of crime seems to depend on idiosyncratic deliberative circumstances. In his 1996 scholarly paper estimating elasticities of crime, Levitt offered the following conclusions:

While calculations of the costs of crime are inherently uncertain, it appears that the social benefits associated with crime reduction equal or exceed the social costs of incarceration for the marginal prisoner. …

If anything, the results of this paper suggest that increasing the amount of time served by the current pool of prisoners would be socially beneficial.^

Levitt co-authored a paper, published in 2004, that presented more elasticities of crime. This paper focused on incarcerating drug offenders. Its results affirmed that incarcerating drug offenders had a similar effect to incarcerating other offenders. Nonetheless, this paper suggested that existing levels of incarceration were excessive:

On the margin, locking up drug-offenders has roughly the same impact on violent and property crime as incarcerating other types of criminals. …

The question of foremost public policy interest related to our work is whether or not the investment in drug-offender incarceration has been cost effective. A serious cost-benefit analysis is beyond the scope of this paper, but a few points are worth making. First, for typical values of the costs of crime, even the most generous estimates of the crime reduction attributable to prison (Levitt, 1996) suggest that current levels of incarceration are excessive. Thus, it is not easy to justify drug imprisonment based on associated declines in violent and property crime alone.^

The reason for the difference between Levitt’s 1996 conclusions and these conclusions cannot be found easily in the papers’ technical analysis of causation. The U.S. level of incarceration in 2004, 729 inmates per 100,000 residents, like the corresponding figure of 611 in 1996, is extraordinarily high from historical and cross-country perspectives. Deliberative circumstances offer an alternative perspective on causation. Levitt in 2004 had greater disciplinary status and hence probably benefited less from making provocative claims. In addition, most persons, especially academics, tend to be more sympathetic to drug offenders than to other types of offenders. These deliberative circumstances seem to explain better than any obvious reasons within his papers’ technical analyses the differences in Levitt’s conclusions.