Manhattan Institute for Policy Research.
Subscribe   Subscribe   MI on Facebook Find us on Twitter Find us on Instagram      

Civic Report
No. 8  September 1999

Right-Sizing Justice: A Cost-Benefit Analysis of Imprisonment in Three States

Anne Morrison Piehl,
Associate Professor of Public Policy,
John F. Kennedy School of Government, Harvard University
Bert Useem,
Director of the Institute for Social Research
and Professor of Sociology, University of New Mexico
John J. DiIulio, Jr.,
Fox Leadership Professor of Politics, Religion and Civil Society
and Professor of Political Science at the University of Pennsylvania;
Senior Fellow, Manhattan Institute

“Americans face at least three distinct but related crime challenges. First is the challenge of preventing at-risk children from becoming juveniles or young adults who criminally violate the life, liberty, or property of others. . . . Second is the challenge of protecting ourselves from victimization at the hands of juvenile and adult criminals. Third is the challenge of restraining convicted but community-based juvenile and adult criminals so that they cannot commit additional crimes against persons or property.”

Council on Crime in America,
The State of Violent Crime in America, p. 1
(The New Citizenship Project, 1996).

The Criminal Justice Sorting Machine

Violent crime in America has fallen by about 21 percent since 1993, juvenile crime has descended from its 1994 peak, and property crime is at a post-1973 low. While no one really knows what factors account for these welcome trends, most experts believe that both punishment and prevention efforts are part of the story. But this good news has not dramatically lessened Americans’ concerns about crime; though most rightly feel safer today than they did five years ago, most continue to identify crime as one of the main problems facing the country today. Indeed, more Americans were “personally most concerned about” crime in 1998 (46 percent) than in 1974 (30 percent).1

No experts predicted the post-1993 crime drop, but most now contend that the good news about crime will continue. A few dissenting voices, however—James Q. Wilson of UCLA, James Alan Fox of Northeastern University, and a co-author of this report, John DiIulio of the University of Pennsylvania—continue to warn that certain demographic trends are likely to exert upward pressure on crime rates; for example, by the year 2006, the number of teens aged 14 to 17 will be a fifth greater than it was in 1996, and the nation’s total teen cohort will be its largest since 1975.

With an eye to this trend, the Council on Crime in America, an independent bipartisan group led by former U.S. Secretary of Education Dr. William Bennett and former U.S. Attorney General Griffin Bell, argued in its January 1996 report that the justice system could do a better job of “restraining known, convicted, violent, and repeat criminals.”2  In its February 1997 report, the Council argued that the justice system could do a better job of promoting crime prevention programs, especially those in which responsible, caring non-parental adults “monitor, mentor, and minister” to at-risk youth, juvenile delinquents, and young adults on probation.3 

Generally speaking, liberals liked the Council’s second report far more than its first, and conservatives liked its first report far more than its second. Taken together, however, the Council’s two reports reflect what we would identify as three emerging points of consensus about crime, prevention, and punishment in America today.

First, to the extent that public policies, independent of other factors, can effect and sustain reductions in crime, both punishment and prevention—both “prisons and programs”—are necessary. Second, though it is true that spending on all aspects of criminal justice remains small compared to many other categories of public spending (for example, a single federal-state health program, Medicaid, consumes more tax money than all federal, state, and local spending on prisons and jails combined), any increases in justice-system spending still must be cost-effective—in other words, lead to proportional increases in public safety. Third, even at its best, the justice system cannot cost-effectively detect, arrest, convict, sanction, and supervise more than a small fraction of all criminals, adult and juvenile, suspected and adjudicated.

Metaphorically speaking, the justice system functions as a “sorting machine” with loose-fitting parts (courts, cops, and corrections), competing legal purposes (punish, deter, incapacitate, rehabilitate), multiple enforcement methods (ranging from police investigation to plea-bargaining, community corrections to maximum-security prisons), and diverse leaders and constituencies (federal, state, and local policy makers, voters, advocates, and others).

How much sorting does the system do? Consider a few relevant facts. In 1994, Americans experienced some 4.2 million serious violent crimes (murders, rapes, robberies, and aggravated assaults). In the same year, the justice system registered about 146,000 convictions for those serious violent crimes, and sent some 95,000 adjudicated felons to prison for them. On any given day in 1996, nearly 60 percent of offenders convicted of rape or sexual assault were on probation or parole rather than incarcerated. Similarly, on any given day in 1997, some 3.9 million persons were on probation or parole, including hundreds of thousands of persons convicted of a violent crime.

Suppose that, on average, every individual criminal was responsible for four serious violent crimes in 1994, and that the system caught, convicted, and imprisoned all of them. That would have added 1,000,000 serious violent felons—not 95,000—to prison in 1994. Or suppose that all of the convicted sex offenders under the custody of corrections officials (i.e., on probation, on parole, or incarcerated) in the United States had been incarcerated in 1994. That would have increased the number of such offenders behind bars from about 99,000 to 234,000. Or imagine that all persons on probation for a violent crime in 1994 were incarcerated instead. That would have landed another 400,000 or so persons in prison.

If any such policy changes were to be made, the incarcerated population would increase substantially, as would the costs of inmate care and custody. Furthermore, these changes would certainly increase the percentage of Americans who come under the control of state correctional agencies some time in their lives. A recent study from the Bureau of Justice Statistics4  estimated that the present estimated lifetime chance of imprisonment is 16.2 percent for blacks, 9.4 percent for Hispanics, and 2.5 percent for whites. (Males make up the majority of prisoners, so the corresponding estimates for males only in these groups are 1.5 to two times as high.) Because certain assumptions are built into this calculation in order to arrive at these estimates,5  it is important to interpret these figures cautiously. Nonetheless, one might want to consider that increases in the use of imprisonment are likely to increase Americans’ (already high) lifetime incidence of incarceration. Any additional increases that failed to promote public safety in a cost-effective way would be difficult, if not impossible, to justify.

As the nation continued its post-1980 prison expansion—a “prison binge” that followed a post-1960 bout of “prison bulimia,” during which the ratio of prison commitments to arrests for serious crimes declined—the system generally “sorted” so as to imprison only those adjudicated felons who posed acute or chronic threats to persons or property. Clearly, however, even after this expansion, the system does not—and cannot—cost-effectively incarcerate more than a small fraction of all adjudicated felons, let alone of all serious criminals.

Presently, the justice system is rapidly approaching an average daily total of 2 million persons in federal and state prisons and local jails. Is this number too high, too low, or just about right? Our study seeks a preliminary answer to this question by comparing the costs of keeping a prisoner in jail for a year to the costs that prisoner imposed on society through his or her illegal activity.

Our empirical research estimates the criminality of prisoners entering the New York, New Mexico, and Arizona prison systems in 1997, and our cost-benefit analysis estimates the returns to public safety realized by incarcerating various categories of these offenders. These analyses lead us to conclude that policy makers in these and other states need to revisit mandatory-minimum drug laws that are increasing prison populations without demonstrably and cost-effectively increasing public safety.

Benefit/Cost Analysis: Estimating the Criminality of Incoming State Prisoners

Benefit/cost analysis allows us to compare things that seem very different from each other. That is, we can compare the same outcome at different times (building a bridge today versus building a bridge in 10 years) or different outcomes (a bridge today versus a school today). In order to do this, however, one must define a criteria for measuring the value of these outcomes. Generally, the outcomes are translated into dollar terms, which then allow easy comparisons. We should not overlook, however, the importance of the method by which these outcomes are translated into dollars. A full discussion of all the different ways this can be done is beyond the scope of this report. Here we describe the conceptual issues involved and our attempts to resolve them.6 

The purpose of a benefit/cost analysis of incarceration is to calculate the social costs and the social benefits of prison and compare them. It is generally thought that calculating the costs of incarceration is relatively simple: just add up the costs of building and operating a cell. The range of estimates for these costs is about $20,000–$50,000 per year.7  But there may be many social costs left out of these numbers. By “social costs” we mean any burdens on society in addition to the resources it takes to run a prison system. They include the lost labor-market productivity of inmates, the loss to families of having a member away from home, and the loss to communities of having a resident removed.

On the benefit side, there likewise may be a variety of effects. First, incarceration will incapacitate the offender so that he or she will not victimize other (nonincarcerated) citizens during the period of confinement. Second, the incarceration of one person may serve as a deterrent to others. Moderating against these influences are the possibility that the criminal activities of inmates are picked up by other inmates (prisons as “schools of crime”), the possibility that criminals are simply “replaced” by other individuals in the community, and the likelihood that at some point in time an offender naturally reduces his criminal activity regardless of government sanction.

The first serious effort to apply benefit/cost analysis to corrections was conducted over a decade ago. In 1987, Edwin Zedlewski, a staff economist for the National Institute of Justice, tried to calculate the costs and benefits of incarceration. He did this by dividing the yearly cost to keep one inmate in prison ($25,000 per annum, he estimated) by the product of the number of crimes the “typical” inmate would commit if on streets (187) and the average cost of a crime to society ($2,300).8  Simple arithmetic showed that the benefits greatly outweighed the costs, on the order of 17 to 1. The results, as interpreted by Zedlewski, strongly supported the idea of increased use of incarceration.

Numerous researchers challenged the 17 to 1 ratio. Zedlewski had estimated the number of crimes the “typical” inmate would commit if on the streets based upon a survey of prisoners and jail inmates in three states (Michigan, Texas, and California). The inmates were asked about the number of crimes they committed in the period immediately prior to their incarceration. Critics argued that Zedlewski should have used the median, rather than the mean, to calculate the typical number of crimes committed. A few inmates claimed an extraordinarily high rate of crimes; some of these claims may have been boasts or even deliberate jests to “put on” the researchers. The use of the median would have reduced the impact on the estimate of these possibly outlandish claims.

But even if one was to accept Zedlewski’s calculations as 100 percent correct, the benefit/cost ratio might depend crucially on the size of the prison system. That is, if most highrate offenders are already in prison, then prison growth would result in the imprisonment of less and less dangerous offenders (assuming the offender population is not growing larger or more dangerous). Actually, Zedlewski thought about this possibility. In a rebuttal to two lawprofessor critics, Zedlewski argued that the number of offenders behind bars was then low enough that expanding the prison population would not affect the inmate profile. There is, he stated, “no basis for believing that the average commission rates should decrease in the 300,000 to 600,000 inmate range under discussion.”9 

In fact, the country exceeded the upperboundary of 600,000 in 1988, the year after Zedlewski published his report. Is increased use of incarceration still cost-beneficial, with a much larger inmate population?

In the following, we use benefit/cost analysis not so much to argue for a particular number of prison cells, but to compare the policies of several states and to consider variations of criminal justice policy. We calculate the benefit of incarceration by multiplying the number of specific crimes committed by inmates by the social cost of each specific crime, thereby deriving the cost of the crimes particular inmates would have committed in the community had they not been incarcerated. The data regarding the number of crimes each prisoner committed are from a 1997 survey of prisoners in three states—New York, New Mexico, and Arizona—as those inmates were passing through intake facilities. The data regarding the social cost of each crime are taken from a recent report published by the National Institute of Justice, with one exception.

Our analyses are quite similar to those reported by two of us in the Brookings Review10  utilizing surveys of inmates in Wisconsin and in New Jersey. These other articles emphasized that the answer to whether prison expansion is warranted depends on how the additional cells would be used. The answer also depends upon how well criminal justice agencies sort offenders for incarceration from among the greater population of individuals who have committed criminal acts—in other words, which law-breakers are selected for incarceration. In what follows, we present profiles of the prison populations in the three states, find that many people currently imprisoned do not pass a benefit/cost test, and conclude that states seeking to maximize the benefits of current prison space should reexamine the policy of imprisoning drug-only offenders.

The Cost of Crime

Table 1 displays the estimates of the value of crimes surveyed from a recent report from the National Institute of Justice. These numbers are based on the compensation awarded by juries to injury victims. The $5 cost per drug sale is our own estimate,11  based on two competing considerations. On the one hand, many of the social costs of drug crime comes from the violence and theft associated with the trade. These costs will be largely accounted for on their own. On the other, drug usage has its own inherent costs—they are the reason that drug use is criminalized. Thus, unless society is sorely mistaken about the costs of drug usage—and we do not believe it is—then it would be unrealistic to assign no costs to drug offenses. We believe that $5 reflects a reasonable middle ground. This $5 cost per drug offense should represent the social savings from incarcerating a drug offender and therefore preventing him or her from making a sale. Note that if the offender is “replaced” by another drug dealer satisfying the same customers, this estimate will be “too high.”

Table 1.

While researchers disagree over the precise values that should be used, in practice it turns out that their various estimates are close enough that choosing one over another does not make much difference to the policy conclusions.12  These numbers simply provide a way of ranking the crimes people commit.

We have omitted murderers from the analysis for several reasons. First, in most cases it is difficult to argue that the major benefit from incarcerating murderers is incapacitation, and it is incapacitation that this study design is best able to evaluate. Second, the number of murderers is relatively small13  and will not affect our medians in any substantial way. In what follows, we do not consider changes in criminal justice policy regarding the sentencing of murderers.

Inmate Surveys

During the first half of 1997, we surveyed inmates entering prison in New York, New Mexico, and Arizona. (See the appendices for a discussion of the methodology.) Table 2 describes some of the features of our surveys. The first row shows the number of inmates, male and female, surveyed in each state.

Table 2.

The inmate surveys are divided into three groups, as represented by Rows 2–4. Row 2 reports the number of inmates who said that they committed a crime to which we can attribute an identifiable social cost, that is, one of the crimes listed in Table 1. The third row tabulates inmates who reported that they committed drug offenses, but no others. For purposes of this study, “drug offenses” is defined as the sale of illegal narcotics. Depending on the state, the proportion of the total number of offenders with “only” drug offenses is substantial, between 15 and 30 percent. This group is a subset of Row 2. Row 4 is comprised of people who report having committed no crimes listed in Table 1 during the period we asked about (the four months prior to arrest for the current term).14 These are prisoners who either (A) committed no crime, (B) refused to admit to the crimes that they did commit, or (C) committed crimes we did not ask about.

How should prisoners in Row 4 be treated in our cost-benefit analysis? We, of course, are unable to distinguish between those inmates who were falsely imprisoned (Group A) and those who refused to admit to us the crimes that they did commit (Group B) and cannot include them in our analysis. Nor can we include those in Group C. With regard to this group, the purpose in conducting the survey was not merely to classify inmates by the crimes they committed, but to measure the costs of their crimes. Economists have calculated the costs of many, but not all, crimes. Thus, inevitably, there would be inmates in our sample who committed crimes to which we cannot impute any monetary value.

Table 3.

Some sense of what these “unvalued” crimes are can be gained by examining the type of crimes these particular inmates are serving time for, that is, their official sentencing offense. Table 3 presents the crimes that these inmates reported they were serving time for. Note that the survey specifically asked about some of these crimes, but not others.

Two items stand out from Table 3. First, many inmates who refused to admit to committing crimes were convicted for crimes listed in Table 1. Second, the largest single type of crime committed by these people would be classified as “drug-only” offenders. Indeed, if those who report serving time on drug offenses (possession or sales and who report serving time on no other types of offenses) are considered “drugonly” offenders, the proportion of all prisoners who are offenders “drug only” in New York rises to 49 percent for males and 68 percent for females; in New Mexico, to 26 percent and 41 percent; and in Arizona, to 32 percent and 49 percent respectively. These are substantial jumps.

In the following analysis we only examine the group of offenders with “relevant information,” that is, those who self-report the number of crimes they committed listed in Table 1. We exclude those in Row 4 from our benefit/cost analysis because even if we can establish a sentencing offense for which we have estimated a social cost, we cannot establish how many such crimes these individuals committed and hence are certain to underestimate the social benefits of incarcerating them.

Table 4 reports the social costs of crimes committed by the inmates in the category “offenders with relevant information.” It does this for each offender by multiplying the number of the particular crime type committed by the average social cost of that crime. For offenders with a range of offending, the costs of each type of crime are then added together.16  The table includes values only for male inmates; the results for females are discussed in endnote 17.

Table 4.

Table 4 improves on Zedlewski’s calculation in two key respects. Whereas Zedlewski was forced to assume that all crimes have the same costs, we can relax that assumption because we can draw on the work of economists who have identified the different costs of different crimes. Second, Zedlewski calculated a single cost for the “average” inmate. Table 4 reports the costs at various cutting points (percentiles). It is useful to know not only the benefit/cost ratio of imprisoning the “typical” inmate, but also that ratio for the most and least serious offenders. Thus, we ascertain various cutting points to see the proportions of inmates for whom incarceration clearly is or is not cost-beneficial.

What can be said about the social costs of male inmates in the three states? First, it should be noted that these are inmates who had just been admitted to prison. The social costs of all inmates behind bars would, on average, be higher (assuming that judges give longer sentences to offenders who inflict higher social costs—which is no doubt true).

Second, at all cut points the social cost inflicted by New York inmates is higher than inmates in the other two states. The median social cost in New York is about $32,000, whereas the median cost in New Mexico and Arizona is $26,000. Put differently, in New York, half of the incoming inmates inflicted more than $32,000 in damages, and half less than that amount. Likewise, in New Mexico and Arizona, half of the incoming inmates inflicted $26,000 in damages and half less than that.

Another way to consider the criminality of the median offender is to look at his reported criminal acts. In New York, the $31,866 figure represents the damages associated with the commission of 3 assaults per year. In New Mexico, the median offender committed robberies at the rate of 3 per year (and no other offenses). In Arizona, the median offender committed burglaries (rate of 3 per year), thefts (rate of 6 per year), and car thefts (rate of 6 per year).

Third, as is always found in inmate surveys, the social costs of the most costly offenders greatly exceed those at the lower end of the spectrum. While it clearly pays to incarcerate those at the 80th percentile in all three states, on incapacitation grounds alone, it does not appear to “pay” to incarcerate those below the median. The social costs associated with the offender at the 40th percentile are all below $15,000.17

The offenders going to prison in our three states exhibit two different patterns. Many high rate offenders and many drug offenders take up New York’s prison space. In Arizona and New Mexico, the nondrug-only offenders have committed, on average, fewer crimes than their New York counterparts. (The median number of nondrug crimes committed by non-drug-only offenders is 16 in New Mexico and 15 in Arizona, compared to 24 in New York.) But, compared to New York, these two states imprison fewer drugonly offenders.

What explains these interstate differences? One possible explanation could be derived from the three states’ crime rates and incarceration rates which are listed below in Table 5. However, that table does not seem to help much. First, we would anticipate that high incarceration states would produce low social costs per inmate, because their prison systems would, all else being equal, tend to incarcerate relatively lessserious offenders. Yet New York’s incarceration rate per unit of population is roughly halfway between Arizona and New Mexico; thus, that factor cannot explain New York’s high offender costs.

Table 5.

Second, the crime rates reported in Table 5 are also not very helpful as explanatory factors. New Mexico has a high violent crime rate, compared to New York and Arizona. However, New York has recently experienced dramatic declines in violent crime. The traditionally high level of violent crime in New York may explain the higher per inmate costs. At the same time, however, Arizona and New Mexico have property crime rates nearly double that of New York. No clear pattern emerges here.

Still another explanation comes to mind when one thinks about the prison-system demographics in the three states. Comparing the three states, New York has a higher percentage of AfricanAmerican inmates, and New Mexico and Arizona have higher percentages of Hispanic inmates. Arguably, these differences might be linked to the differences in costs in the three states. Yet when we looked at the cost by race and ethnicity, we found costs were very similar across nonHispanic whites and blacks in New York and in Arizona.18 Within the inmate population, “drugonly” offenders were more likely to be AfricanAmerican and Hispanic (except among women in New York, who were no more likely to be Hispanic).

Policy Alternatives: Parole Violators, Repeat Offenders, and DrugOnly Offenders

Without knowing the precise causes of the patterns observed in Table 4, we can use these data to consider some policy options. In doing so, we note that implementation of policy with consequences for incarceration takes place on the street, in courtrooms, and in probation and parole offices. The policies themselves are primarily set by the legislature and by elected and appointed officials heading law enforcement agencies. Those seeking to improve the corrections system’s “sorting” capability could, quite naturally, use data on aggregate outcomes of the sorting process. This would permit us to determine which types of persons should or should not be sent to prison based on a benefit/cost analysis. For this reason, it is revealing to consider the benefit/cost differences among three types of individual-specific data that policy makers may consider relevant: parole violators, those with criminal histories, and those incarcerated for drug offenses. This information is provided on the following page in Table 6.

Parole Violators

Use of parole revocation as a “sorting” factor raises an interesting question. Some offenders are returned to prison for violating parole without having been convicted of new crimes. Since (presumably) they are under closer surveillance, it is possible that, on average, these offenders have been doing “less harm” just prior to their admission then other potential inmates. On the other hand, some argue that this is a population that should be punished even more severely because of past behavior. Table 7 shows that parole violators are a substantial proportion of the new admissions to New York’s and New Mexico’s prison systems, but far less so in Arizona. Also, it appears that New Mexico has a high proportion of “technical” violations—less than a third (29%) were returned to prison for having committed a new crime.

The top panel of Table 6 reports the social cost estimates (as in Table 4) for the population of male inmates separated according to those on parole at the time of arrest and those not on parole. No clear pattern emerges from this comparison. In fact, it appears that parole violators have a similar level of social cost to those who were not on parole at arrest. This analysis does not suggest any particular changes to the treatment of those who have violated parole.

Table 6.

Criminal Histories

One may also want to consider whether it appears beneficial to concentrate additional prison resources on offenders who already have criminal histories, regardless of their current parole status. Across the three states, from 39-48 percent of the new inmates had served prior terms in adult prisons. The second panel in Table 6 reports the social cost estimates for inmates by whether or not they had prior prison terms. Again, the numbers bounce around somewhat and no clear pattern emerges. In no state is it obvious that there is a higher benefit from incapacitating those with criminal histories.

Drug Offenders

The final policy alternative we consider has to do with drug offenders. We have already shown that a substantial fraction of offenders are drug offenders—a very substantial portion if the data cited at pp. 7-8 is included. In this analysis, we compare those who report their current prison term is due to a violation of drug laws to those serving time for other types of offenses. This is not the same comparison as above, because here we divide the sample based on “sentencing offense” rather than reported criminal activity. (Note that the “sentencing offense” is self-reported, as is everything else on the survey).

One reason to divide the sample this way is that it is quite likely that some of those serving time for drug offenses have committed other types of violations as well. Because social cost numbers incorporate the whole range of offending, this “experiment” will show whether any drug laws lead to the incarceration of violent offenders.

The bottom panel of Table 6 reports the social cost estimates for those sentenced for drug offenses to those who were sentenced for other charges. To simplify the comparisons, we omitted those whose only reported crimes were drug sales but were sentenced for something else. (The concern here is the same as that expressed earlier at pp. 7-8). No estimates are reported for drug offenders in New Mexico due to the small number in this category.

In contrast to the other two policies considered, a clear picture emerges in this panel. The social costs associated with those sentenced on drug offenses are substantially lower than for other inmates, thereby providing policymakers interested in rationing prison space according to benefit/cost analysis with a clear reference point with which to do so. We now turn to our conclusions based on these analyses.

Table 7.


U.S. Supreme Court Justice Robert Jackson once remarked that the Constitution is not a suicide pact. The same must be said for our commitment to punish criminal offenders. This commitment deserves respect, but it can also be too unyielding—taken to extremes it can amount to a suicide pact. Benefit/cost analysis allows us to see if prison is being overused, beyond the normal criteria of fairness and due process of law. The purpose of our study was to conduct such a benefit/cost analysis for the incoming prison population in three states. Several conclusions can be drawn.

Offenders with Social Costs. Arguing against further prison expansion is the principle of diminishing returns. That is, as noted above, if the most serious offenders are already in prison, then prison growth requires the criminal justice system to reach deeper into the pool of prisoneligible offenders, such that increases in incarceration are less and less cost-effective. One of the most surprising, and significant, findings of this study is that this has not been the case with regard to the prison systems we surveyed when drugonly offenders are excluded. When drugonly offenders are included, however, it appears that the value of incarcerating the least “costly” half of inmates (least costly in terms of the social-costs of their offenses) is quite low.

To establish some comparison points, Zedlewski relied on data collected in three states in 1978 and 1979. Anne Piehl and John DiIulio conducted similar surveys of inmates in Wisconsin in 1990 and in New Jersey in 1993. Strikingly, the median number of crimes committed seems uncorrelated with time. Of course, we are comparing different states at different times, whereas ideally we would be comparing the same states at different times. (Actually, the ideal comparison would be data collected before and after a dramatic change in sentencing policy, implemented quickly enough that one could assume the offender population would not have changed.)

Still, again with the exception of drugonly offenders, the comparisons suggest that prison expansion in the three states under study has not come at the expense of incarcerating larger numbers of offenders who, if left on the streets, would commit few crimes.

Imprisonment of DrugOnly Offenders. All three states imprison large numbers of drugonly offenders. The main effect of imprisoning drug sellers, we believe, is merely to open the market for another seller. Numerous students of drug policy attest to the existence of this “replacement process.” Still, it is difficult to observe and measure directly. One indirect measure is the price of drugs. Presumably, if the incarceration of drug offenders does make a dent in the drug market, we would expect to see an association between the number of drug offenders behind bars and the street price of drugs.

The evidence is not encouraging. It is probably not enough to merely point out, as Alfred Blumstein and Allen Beck19  have, that the street price of cocaine has decreased since 1980 while the rates of incarceration have increased. First, one needs to know, as well, what the price of cocaine would have been in the absence of that enforcement level—and we do not. Second, as Mark Moore has pointed out, the “effective” price of a drug includes, not only the drug’s cash price, but also the risk of imprisonment and other inconveniences and danger associated with its purchase.20 Still, the data suggest that the market for the illicit drugs has not been disrupted by increased incarceration.

In sum, it seems to us that the imprisonment of large numbers of drug offenders is not a costeffective use of public resources. At least some prison beds currently occupied by drug offenders would be better reserved for highrate property and violent offenders.

Penal Harm. In the benefit/cost calculations, we did not take into consideration the harm that high incarceration rates can have on communities. A striking finding of our survey is the extent to which the incoming inmates appeared to be contributing members of their communities. As noted below, many of them were working in the month of their arrest, and the vast majority had held a job in the past for more than three months. In all three states, more than 75 percent say they would take a fulltime job paying the minimum wage upon release. Also, the vast majority of the inmates were optimistic that they would be able to find a job upon release, and not end up back in prison.

In terms of the broader debates that routinely swirl around prisons, these findings can be interpreted in two ways. On the one hand, they suggest that the vast majority of prisoners are not driven to crime by dire economic need. Crime appears to be an addon to economically viable lives. On the other hand, the findings further highlight the costs to the community of imprisonment. These are large numbers of people who would otherwise be working and paying taxes.

It is likely that the community costs associated with additional incarceration would increase as the incarceration rate increases. One reason for this is that prison may lose its value as a penalty if it is seen as commonplace. A second reason is that high incarceration rates may undermine the legitimacy of the government if citizens come to see the government as too intrusive and coercive. Of course, low incarceration rates may have that effect too, if citizens come to feel that the government is not “doing enough” about crime. Thus, it is important to get the level of incarceration “right” for reasons of justice as well as to ensure the prudent use of tax dollars. Given the dramatic increases in the proportion of the population under correctional supervision in recent years, these costs must be considered.


I. Highlights from Surveys

 The surveys asked a large number of questions. This section outlines the highlights of the surveys in several sections: preincarceration experiences; the meaning of imprisonment; understanding of sentencing policy; and expectations about life after release. For readers interested in more detail, the values for each of the questions in the surveys are included in Table 8.

A. Lives of Inmates Prior to Incarceration

The survey asked inmates about certain features of their life prior to being admitted to prison. From the responses, we can get a better idea of their life circumstances.

    Prison inmates are at high risk of victimization. Overall, 40 percent of offenders have been the victim of a robbery; robbery victimization is higher in New York than the other two states. Forty-seven percent of men and 69 percent of women have been shot at sometime in their lives. For women (but not men) these rates are higher in New York.

    Most prisoners have legitimate employment histories. Many were working in the month of arrest (40–60 percent), and 85 percent had held a job in the past for more than three months. In all states, more than 75 percent said they would take a fulltime job paying $5.25 per hour, if offered one, upon release. (The percentage is higher for women than for men—88 percent versus 78 percent.) “Drugonly” offenders were neither more, nor less, likely to have been employed than the other inmates in the sample.

    For men, the median offender received less than 10 percent of income from crime (this figure is held down by Arizona inmates, who had somewhat lower earnings than men did in the other states.) For women, the median offender received 10–25 percent of income from crime. “Drugonly” offenders received appreciably higher proportions of their income from crime than did other offenders.

    Two points can be made about these findings. One is that the image of offenders all being driven to crime by poverty and unemployment is, at best, exaggerated. The majority of the offenders in our study had jobs, and few derived a high proportion of their incomes from crime. At the same time, however, these findings highlight one of the costs of imprisonment. Many of these offenders, if not in prison, would be contributing to the economy through work.

    Most prisoners are responsible for one or more dependents. Most inmates were parents, with women somewhat more likely than men to report having children. Around 65 percent of inmates (across the board) were helping to support at least one dependent at the time of arrest. These could be children, parents, or others. Many of the inmates were unwilling to answer a question about who was providing support for those dependents now that they were in prison. Still, a general pattern emerged: the “other parent” primarily supported the dependents of male inmates while grandparents primarily supported the dependents of female inmates.

B. Meaning of Imprisonment

Thomas Bonczar and Allen Beck21  estimate that 5.1 percent of all persons in the United States will serve time in a federal or state prison; among AfricanAmerican males, more than a quarter (28.5 percent) will do so. Richard Freeman22  calculates that, in 1993, one man was incarcerated for every 50 in the workforce and one African-American man was behind bars for every 11 AfricanAmerican men in the labor force. Over a third (34 percent) of the 25 to 34-yearold African Americans who dropped out of high school were behind bars in 1993.

 Our survey included several questions to tap the extent to which inmates perceive their imprisonment as a “normal” course of events. The respondents in two states (New Mexico and Arizona) were asked if they agreed or disagreed with three statements:

    • “Among my friends on the outside, one gains respect for having done time in a  state prison.”
    • “Doing time in a state prison may be hard, but it is ‘no big deal’ for me;” and
    • “For someone like myself, going to prison is just another part of life.”

Thirtytwo percent of the respondents stated they either strongly agreed or agreed that they gained respect from their friends for going to prison; 17 percent said that they strongly agreed or agreed that going to prison is “no big deal” for them; and 21 percent said that they strongly agreed or agreed with the statement that going to prison is “just another part of life.”

Because such questions have never been asked of inmates before, we have no baseline to compare the responses. It may be, for example, that the 17 percent who said that going to prison was “no big deal” were merely expressing a general defiance toward authority. This level of defiance may have been equally prevalent among earlier generations of inmates. Perhaps more worrisome is that onethird of the respondents felt they gained respect for going to prison, and a fifth said that going to prison was a “part of life.”

C. Expectations of Post-Release Life

We also asked the inmates what they thought the chances were that, upon release, they will: (a) “try to make it on the outside, that is, without doing crime”; (b) “find a keep a job at least for a year”; and (c) “end up back in prison or jail.” Most inmates expressed a positive outlook on these questions. Seventyone percent said that they probably will or are certain to try to “make it on the outside”; 77 percent reported a high chance or certainty that they would be able to find and keep a job upon release; and 81 percent thought there was no chance or a low chance that they would return to prison or jail after their release.

Such hope, however, may have little predictive power. Mark Fleisher23  observes that many criminal offenders are masters of “verbal camouflage.” Through repeated encounters with the criminal justice system, they learn to say what will advance their interests in passing through the system (e.g., to “accept” responsibility for their criminal acts), not what they truly believe. At the start our survey, inmates were told that their responses would be kept confidential and could not affect their future circumstances. Still, some may not have believed this assurance or, in any case, may have resorted to verbal camouflage simply out of habit.

D. Inmates’ Understanding of Sentencing Policy

Mandatory minimums are important in the sentencing systems in each of the states studied. Nevertheless, many inmates do not seem to understand the penalty structure. Depending on the state, 40–54 percent of inmates believe they are currently serving mandatory minimum sentences (New York is at the high end; this knowledge does not vary by “drugonly” status). But not everyone serving a mandatory minimum knew about the mandatory sentence at the time of the crime. The proportion who was not aware of the sentence ranged from 20–35 percent of those with mandatory minimum sentences. 

Furthermore, only 20–30 percent of inmates thought that the penalty for the crime for which they are now serving time would include prison at all. “Drug-only” offenders are somewhat more likely to think they might face prison time.

Table 8.

II.  Methodology

The surveys from the three states used almost identical instruments and very similar protocols. In all cases, the survey was anonymous and voluntary. Correctional officers brought inmates to the classrooms in which the surveys were administered by outside researchers. Spanish versions of the survey were available. Inmates received no compensation for their effort.

The survey instrument was very similar to that used by DiIulio and Piehl in New Jersey and Wisconsin, which in turn were similar to the RAND surveys conducted in the 1970s. In those other administrations, however, inmates received compensation for their participation (in the form of stipends to commissary accounts).

New York:  A survey of 634 new admissions to the New York prison population conducted in late 1996 and early 1997. 541 of the surveys were useable for analysis. The sample included inmates who entered the system through each of the four intake centers for the state Department of Corrections. The sample is a census of one week’s incoming admissions. As such, the sample represents the flow of inmates into the system, not the current prison population.

New Mexico: A total of 429 inmates were surveyed; 401 were useable. Female inmates were sampled at a higher rate than were male inmates. New Mexico has one intake facility for men and one for women, where the surveys were administered. The first interviews were conducted on May 2, 1997 and the last on August 22, 1997. During this period, an interviewer visited the men’s facility every week (except over the Fourth of July holiday) and the women’s facility every two or three weeks.

Arizona: A total of 554 surveys were administered; 534 were useable. Like New Mexico, Arizona has one intake facility for men and one for women, where the surveys were administered. The first interviews were conducted on July 3, 1997 and the last on September 6, 1997.


  • Blumstein, Alfred and Allen Beck, Factors Contributing to the Growth in U.S. Prison Populations, (Crime and Justice conference, Washington, 1998).
  • Bonczar, Thomas P. and Allen J. Beck, Lifetime Likelihood of Going to State or Federal Prison, (Bureau of Justice Statistics, No. NCJ160092, 1997).
  • Boyum, David and Mark A. R. Kleiman, “Alcohol and Other Drugs,” in Crime (James Q. Wilson & Joan Petersilia, eds., San Francisco: Institute for Contemporary Studies, 1995).
  • Criminal Justice Institute, The Corrections Yearbook, (Criminal Justice Institute, 1997).
  • DiIulio, John J., Jr. and Anne Morrison Piehl, Does Prison Pay? The Stormy National Debate over the Costeffectiveness of Imprisonment, 9 Brookings Review 28 (Fall, 1991).
  • Fleisher, Mark A., Beggars and Thieves: Lives of Urban Street Criminals, (University of Wisconsin, Madison Press, 1995).
  • Freeman, Richard B., Why Do So Many Young American Men Commit Crimes and What Might We Do About It? 10 Journal of Economic Perspectives 25 (1996).
  • Gillard, Darrell and Allen J. Beck, Prison and Jail Inmates at Midyear 1997 (Bureau of Justice, No. NCJ167247, 1998).
  • Hagan, John, “Children of the Prison Generation: Collateral Consequences of Imprisonment for Children and Communities,” (unpublished paper, 1998).
  • Miller, Ted R., Mark A. Cohen, and Brian Wiersema, Victim Costs and Consequences: A New Look, (National Institute of Justice, 1996).
  • Moore, Mark, Policies to Achieve Discrimination on the Effective Price of Heroin, 63 American Economic Review 270 (1973).
  • Moore, Mark H. and Anne Morrison Piehl, Reckoning the Value of Prisons: The Strengths and Limitations of Benefit/Cost Analyses of the Crime Reduction Effects of Imprisonment, (Unpublished paper, Harvard University, 1998).
  • Piehl, Anne Morrison and John J. DiIulio, Jr., Does Prison Pay? Revisited, 13 Brookings Review 20 (Winter, 1995).
  • Sourcebook of Criminal Justice Statistics 1995 (Kathleen Maguire & Ann L. Pastore, eds., U.S. Department of Justice, Bureau of Justice Statistics, USGP0, 1996).
  • Sourcebook of Criminal Justice Statistics 1996 (Kathleen Maguire & Ann L. Pastore, eds., U.S. Department of Justice, Bureau of Justice Statistics, USGP0, 1997).
  • U.S. Department of Justice, Crime in the United States 1997, (FBI Uniform Crime Reports, USGPO, 1998).
  • Zedlewski, Edwin W., Making Confinement Decisions, (National Institute of Justice Research in Brief, 1987).
  • Zedlewski, Edwin W., New Mathematics of Imprisonment: A Reply to Zimring and Hawkins, 35 Crime and Delinquency 169-73 (Jan. 1989).


* Our study had its origins in deliberations of the Council on Crime in America. We are grateful to the Council; the foundations that supported our independent research; the Manhattan Institute’s Center for Civic Innovation; Mr. Peter Widulski, Ms. Pauline Spaulding and Ms. Melinda Ogle, who were instrumental to our New York, New Mexico and Arizona survey administrations, respectively; and the officials and prisoners in each state who cooperated with our study. The data analyses and opinions expressed herein, however, are solely those of the authors.

  1. Roper Starch Worldwide, reporting survey results for August 1998, as summarized in The American Enterprise, (January/February 1999) at 93.
  2. The State of Violent Crime in America, at i.
  3. Council on Crime in America, Preventing Crime, Saving Children (The Manhattan Institute for Policy Research, 1997).
  4. Thomas P. Bonczar and Allen J. Beck, “Lifetime Likelihood of Going to State or Federal Prison,” (Bureau of Justice Statistics, No. NCJ160092, 1997).
  5. The critical assumption is that the current distribution of inmates approximates the “steady state.” However, in a period of rapid population growth, an extrapolation based on cross-sectional data may yield estimates that overstate the lifetime likelihood of going to prison. This is because there will be a large number of “first-time” offenders at all ages in the early stages of prison growth. As people age in this new regime, there will be fewer additional first-time offenders due to the larger pool of people with prison experience. Therefore, estimates based on the early experience in a new sentencing regime would tend to produce an upward bias in the lifetime likelihood of incarceration. On the other hand, Bonczar and Beck (1997) used data collected in 1991.  The increase in the incarceration rates since 1991 may offset this upward bias.
  6. See Ted R. Miller, Mark A. Cohen, and Brian Wiersema, “Victim Costs and Consequences: A New Look,” (National Institute of Justice, 1996); or Mark Moore, Policies to Achieve Discrimination on the Effective Price of Heroin, 63 American Economic Review 270 (1973), for more details.
  7. This wide band for costs reflects two things.  First, the actual cost of corrections varies considerably from state to state.  Second, measuring the cost of corrections is notoriously difficult, and there is no standard or accepted method.  This makes inter-state comparisons risky. In any case, according to one calculation, Arizona’s per inmate costs are just below the average, whereas those in both New York and New Mexico are among the highest in the nation (The Corrections Yearbook 74-75 (Criminal Justice Institute, 1997)).
  8. Zedlewski determined the cost of a single crime by dividing the total costs of the criminal justice system in a year by the number of crimes committed in that year.
  9. Edwin Zedlewski, New Mathematics of Imprisonment: A Reply to Zimring and Hawkins, 35 Crime and Delinquency 171 (Jan. 1989).
  10. John J. DiIulio Jr. and Anne Morrison Piehl, Does Prison Pay? The Stormy National Debate over the Costeffectiveness of Imprisonment, 9 Brookings Review 28 (Fall, 1991); and Anne Morrison Piehl and John J. DiIulio, Jr., Does Prison Pay? Revisited, 13 Brookings Review 20 (Winter 1995).
  11. All relevant previous analyses, from Zedlewski (1987) through Piehl and DiIulio (1995) have imputed zero social cost to illicit drug sales. To illustrate why, consider that in Piehl and DiIulio (1991), the prisoner at the 75th percentile (i.e., 3 out of 4 of the prisoners in the sample imposed greater social costs than he did) was a high-rate property offender.  In effect, the empirical evidence on offending patterns suggests that incapacitating high-rate property offenders via incarceration reduces the total number of property crimes because one high-rate house burglar or car thief is not simply “replaced” by another, nor do non-incarcerated high-rate property offenders respond to the incarceration of another high-rate property offender by doing all the crimes they would otherwise have done plus “his.”  In fact, his incarceration may even reduce their marginal rate of offending (i.e., a deterrence effect).  But most of the empirical evidence suggests that the incapacitation value of incarcerating the median drug offender is zero or close to zero.  There are definitely conditions under which street-level anti-drug law enforcement makes drugs more expensive and deters both consumption and sales; for example, see Mark A. R. Kleiman, Against Excess: Drug Policy for Results (Basic Books, 1992). But there is as yet little evidence of an inverse relationship between, for example, quadrupling prison commitments from 19 to 80 per 1,000 drug arrests (as we did in America from 1980 to 1994) and the number of felonious street-level drug transactions. James Q. Wilson, in an article otherwise devoted to exposing the empirical fallacies of arguments for legalizing drugs, captures the academic consensus on the subject when he observes that a “robber taken off the street is not replaced by a new robber who has suddenly found a market niche, but a drug dealer sent away is replaced by a new one because an opportunity has opened up.” See Wilson, What to Do About Crime, Commentary (Sept. 1994) at 31.
    As explained in the text, our other social cost estimates are derived from relevant analyses of jury awards. There are no other social cost estimates in the literature for drug sales.  We here impute a social cost of $5 per drug crime as a considered response to critics of previous analyses who argue that imputing zero cost to these crimes “analytically decriminalizes” or “legalizes” drugs, and wrongly assumes that all street-level drug crimes always and everywhere, from small towns to inner cities, are subject to an instant and perfect “replacement effect.”  But the $5 imputation does not in any way fundamentally change our bottom-line results from what they would be had we instead imputed zero cost as per the practice in the literature to date, nor does it preclude others from imputing larger costs and reanalyzing the data accordingly.
  12. The correlation between using the numbers we report below in calculating “social costs” (defined below) and using the numbers some of us have used in the past is 0.993 (c.f. the estimates in DiIulio and Piehl 1991; and Piehl and DiIulio 1995). The point of doing this is that some economists argue that stolen property is not destroyed, it is “transferred” to others. Therefore, the value of the stolen property should not count as a social cost. Taken seriously, this should argue for subtracting the entire value of property taken from social loss estimates. This seems extreme, because we know that the “resale” price of stolen property is much lower than the replacement value.  We have taken an admittedly arbitrary discount of 25%. Note that because the value of the property is quite small compared to the total costs per crime, precision on this dimension makes little substantive difference.
  13. Among new court commitments to prisons in 1992, 4.1% were convicted of homicide (includes murder, nonnegligent manslaughter, and negligent manslaughter). These figures come from data collected in 38 states, covering the vast majority of new inmates.  Sourcebook of Criminal Justice Statistics 1996, Table 6.33 (Kathleen Maguire & Ann L. Pastore, eds., U.S. Department of Justice, Bureau of Justice Statistics, USGPO, 1997).  As detailed in the notes to Table 2, the omission of murderers drops 40 people from our analyses.
  14. It is possible that some of the people in this row have been sentenced for a crime they committed more than four months before their arrest, which could account for the large number of people who were sentenced for crimes listed in Table 1 but who do not self-report committing any crimes. Of course, there is no way to determine the extent of this phenomenon, and even if there were there is no way of ascertaining the number of Table 1-listed crimes each person committed. Therefore, people who were sentenced for, but did not self-report, Table 1-listed crimes are excluded from further analysis.
  15. From admissions data to prisons nationwide, we learn that 22% of offenders committed to prison had a “most serious offense” that was not specifically asked about in the selfreport section of our survey. These offenses include: kidnapping (0.6%), drug possession (6.7%), other drug offenses (nontrafficking) (4.8%), public order offenses (including weapons violations and driving while intoxicated) (8.8%), and other offenses (1.1%). Sourcebook at Table 6.33.
  16. It should be kept in mind that these social costs are for a subset of the entering offenders. Excluded are murderers and offenders who did not report having committed a crime that we asked about.
  17. While the samples of women inmates in the three states were small, it appears that the incapacitation benefits are lower for women than they are for men. An important dimension of this finding is that the proportion of “drug-only” offenders is as high or higher for women than men. This is especially the case if “drug possessors” are included in the category of drug-only offenders.
    Given the small sample size, the number of dimensions that can be analyzed, as well as the reliability of the estimates, are limited. Here we must rely primarily on the Arizona data, which include 71 women inmates with crimes associated with social costs. Relative to the men in Arizona, at each point in the distribution (c.f., Table 4), the social costs associated with the women were lower than for the men, though these differences were not always statistically significant.
  18. While there was some difference in New Mexico, given the huge variation in social cost and the small number of African Americans in our sample (n=23), we cannot say anything definitive.
  19. Alfred Blumstein and Allen Beck, “Factors Contributing to the Growth in U.S. Prison Populations,” (Crime and Justice Conference, Washington, 1998).
  20. Mark Moore, Policies to Achieve Discrimination on the Effective Price of Heroin, 63 American Economic Review 270 (1973).
  21. Thomas P. Bonczar and Allen J. Beck, “Lifetime Likelihood of Going to State or Federal Prison,” (Bureau of Justice Statistics, No. NCJ160092, 1997).
  22. Richard B. Freeman, Why Do So Many Young American Men Commit Crimes and What Might We Do About It? 10 Journal of Economic Perspectives 25 (1996).
  23. Mark A. Fleisher, Beggars and Thieves: Lives of Urban Street Criminals, (University of Wisconsin, Madison Press, 1995).


Center for Civic Innovation.



This study examined the criminal histories of incoming prisoners in three states to determine if non-violent offenders were taking up prison space that might better be used to house violent criminals.


The Criminal Justice Sorting Machine

Benefit/Cost Analysis: Estimating The Criminality Of Incoming State


The Cost of Crime

Inmate Surveys

Policy Alternatives: Parole Violators, Repeat Offenders, and Drug-Only Offenders

Parole Violators

Criminal Histories

Drug Offenders


Offenders with Social Costs

Imprisonment of Drug-Only Offenders

Penal Harm


Highlights from Surveys

Lives of Inmates Prior to Incarceration
Meaning of Imprisonment

Expectations of Post-Release Life

Inmates’ Understanding of Sentencing Policy

Survey Responses





Home | About MI | Scholars | Publications | Books | Links | Contact MI
City Journal | CAU | CCI | CEPE | CLP | CMP | CRD | ECNY
Thank you for visiting us.
To receive a General Information Packet, please email
and include your name and address in your e-mail message.
Copyright © 2015 Manhattan Institute for Policy Research, Inc. All rights reserved.
52 Vanderbilt Avenue, New York, N.Y. 10017
phone (212) 599-7000 / fax (212) 599-3494