Journal of Criminal Law and CriminologyVolume 69Issue 4 Winter Article 17

Winter 1978

Mandatory Prison Sentences: Their ProjectedEffects on Crime and Prison PopulationsJoan Petersilia

Peter W. Greenwood

Follow this and additional works at: https://scholarlycommons.law.northwestern.edu/jclc

Part of the Criminal Law Commons, Criminology Commons, and the Criminology and CriminalJustice Commons

This Criminology is brought to you for free and open access by Northwestern University School of Law Scholarly Commons. It has been accepted forinclusion in Journal of Criminal Law and Criminology by an authorized editor of Northwestern University School of Law Scholarly Commons.

Recommended CitationJoan Petersilia, Peter W. Greenwood, Mandatory Prison Sentences: Their Projected Effects on Crime and Prison Populations, 69 J.Crim. L. & Criminology 604 (1978)

9901-4169/78/6904-0604$02.00/0THE JOURNAL OF CRIMINAL LAW & CRIMINOLOGY Vol. 69, No. 4Copyright © 1978 by Northwestern University School of Law Pn'nted in U.S.A.

CRIMINOLOGYMANDATORY PRISON SENTENCES: THEIR PROJECTED EFFECTS ON CRIME

AND PRISON POPULATIONS*

JOAN PETERSILIA** AND PETER W. GREENWOODt

INTRODUCTION

Reform of sentencing statutes has recentlyemerged as a major issue of national debate. Newlegislation is being considered at both federal andstate levels to modify criminal proceedings, partic-

ularly sentencing. At the beginning of 1977, thefederal government and the legislatures of thirtystates were contemplating major revisions of theircriminal codes.'

Two concerns have apparently prompted thisinterest in sentencing reform. First, criminologists,legal scholars and political leaders have expresseddiscontent with excessive disparities in the sen-tences imposed and served under present statutes.It has been repeatedly shown that persons of similarcriminal history convicted of similar crimes aretreated differently by the courts. For example, theaverage prison sentence for persons convicted ofbank robbery is eleven years nationwide, but fiveand one-half years in the Northern District ofIllinois and seventeen years in Georgia. 2 Much ofthe blame for the disparity has been placed on thewide latitude allowed judges under current sen-

* This paper was prepared under Grant Number 77-NI-99-0053 from the National Institute of Law Enforce-ment and Criminal Justice, Law Enforcement AssistanceAdministration, U.S. Department of Justice. Points ofview or opinions stated in this document are those of theauthors and do not necessarily represent the officialposition or policies of the U.S. Department of Justice.The authors wish to thank Hans Zeisel, James Q. Wilson,Daniel Glaser and Jan Chaiken for reviewing an earlierdraft of this article.

** Joan Petersilia is a Research Associate in the RandCriminal Justice Program, and is the coauthor of CRIMI-

NAL CAREERS OF HABITUAL FELONS (1978) and THE CRIM-

INAL INVESTIGATION PROCESS (1977).t Peter Greenwood is the Director of the Rand Crim-

inal Justice Program, and coauthor of CRIMINAL CAREERSOF HABITUAL FELONS (1978), THE CRIMINAL INVESTIGA-

TION PROCESS (1977) and PROSECUTION OF ADULT FELONYDEFENDANTS (1976).

'Compiled from a recent, unpublished survey con-ducted by the Rand Corporation in 1977.

'CRIME CONTROL DIGEST, June 13, 1977, at 3.

tencing statutes. In addition to obvious jurisdic-tional differences, some empirical evidence suggeststhat a significant part of the disparity reflectsjudicial prejudice, conscious or subconscious. Forexample, a study of sentences for larceny andassault disclosed that in state courts 74% of theblacks convicted of larceny were sentenced toprison, while only 49% of the whites with similarrecords were imprisoned. Many criminologistshave called for substantial changes in sentencing

on the grounds that the current practice is grosslyinequitable.

Second, there is mounting public distress overnot only the high rate of violent crimes but alsoover the rise in property crimes. Encouraged by themass media, the public has begun to blame judicialleniency for the high level of such crime. Manycitizens believe that a "get tough" policy in thecourtroom would (1) help protect them againstserious criminals by imprisoning such persons forlonger periods, and (2) deter other persons fromcrime because of the harsher sentences they wouldexpect to receive if caught. This notion exists notonly in the popular realm but it is also advancedby respected law enforcement personnel. For in-stance, former U.S. Attorney General Edward H.Levi recently cited the failure of the courts toimprison enough criminals as a primary reason forrising crime rates.4

Many advocates of reform are convinced thatmore certain, more widely publicized and more.severe prison sentences for serious offenders willenhance public protection. They cite recent empir-ical evidence that most serious crime is committedby repeaters -and that these recidivists, althoughrepeatedly arrested and convicted for seriouscrimes, are not consistently imprisoned. Statisticscompiled by the Rand Corporation reveal that

3ADMINISTRATION OFFICE OF THE U. S. COURTS, FED-

ERAL OFFENDERS IN THE UNITED STATES DisrRiar COURTS77 (1971).

" BOSTON EVENING GLOBE, Sept. 16, 1975, at 4.

MANDATORY PRISON SENTENCES

60% of those arrested for robbery have a priorfelony conviction, but only 48% of those with aprior conviction are sentenced to prison.5

Because of the low probability of incarcerationafter conviction, many experts believe that thegrowth of crime can be substantially explained bythe fact that, given our present use of sanctions,"crime pays." Therefore, an object of any newsentencing scheme must be to raise the "costs" ofcommitting crime. If offenders were certain tl.atthey would surely receive a prison sentence if con-victed for a serious offense, some might judge thepenalty too great to risk. In fact, every empiricalstudy relating the certainty of imprisonment to thecrime rate has shown that the higher the probabil-ity of imprisonment for a major offense, the lowerthe rate for that offense 6

Even if the deterrent effect failed to reducecrime, increasing the proportion of offenders whogo to prison should reduce the level of crime be-cause more offenders would be unable to commitcrimes while imprisoned. Researchers have recentlybegun to estimate the probable reduction in crimeif a larger proportion of convicted felons wereimprisoned. The estimates vary widely dependingon the assumptions made. For instance, it has beensuggested that the rate of violent crimes could bereduced by as much as 80% if every person con-victed of a violent crime were imprisoned for fiveyears.7 But another study has concluded that "in-capacitation makes only a small and modest im-pact on the violent crime rate; a 4.0 percent dropis the highest estimate obtained in this re-search."8 Regardless of the continuing debate overhow crime might be reduced by incapacitation, itappears inevitable that policymakers will alterstatesentencing codes in the hope of assuring morecertain and equal justice. If this alteration alsoreduces crime, so much the better.

Two major reform proposals designed to limitthe latitude of sentencing judges are receiving se-

5 Based on an analysis of the 1973 California OffenderBased Transactional System arrest file, maintained bythe California Bureau of Statistics.

6Tittle & Rowe, Certainty of Arrest and Crime Rates, 5-Soc. FORCEs 455 (1974); Tullock, Does Ihniment DeterCrime? 36 PuB. INTRmr 103 (1974). Although thesefindings are not inconsistent with deterrence theory, therare other possible explanations for these effects thatcannot be sorted out with the data currently available.

7 Shinnar & Shinnar, The Effects of the CriminaljusticeSystem on the Control of Crime A Ojantitatire Appmada, 9 LAw& Soc'y REv., 581, 605 (1975).

8 Van Dine, Dinitz & Conrad, The Irapacitation of theDangerous Offeder. A Statistical Exp iment, 14 J. RasxicCRIME & DELNQUENCY 22,31 (1977).

rious attention. In the first proposal, called "flat-time sentencing," the legislature sets a specificsentence for each crime, or degree of crime, to beimposed by the judge and served in full, althoughreductions for "good behavior" are possible. Thesecond proposal is called the "mandatory-mini-mum' sentence and it requires that a minimumperiod of incarceration be served. The mandatory-minimum scheme appears to be the more popularof the two. During the Ninety-fourth Congressalone (1975-76), more than thirty separate bills orresolutions calling for mandatory-minimum sen-tences were introduced. Several'states, includingMassachusetts and Connecticut, have already en-acted statutes requiring mandatory-minimumprison sentences for conviction of certain offenses.

Most of the state reform proposals limit themandatory sentencing to specified crimes or cate-gories of criminals. The most common category isthe repeat criminal because of the general beliefthat the greater number of crimes an offendercommits, the more severe his sentence should be.Some states begin mandatory sentencing with thesecond offense,9 while others begin it at the third'0

or fourth." Other states have abandoned the"quantity of convictions" punishment principle byfocusing on particularly dangerous aspects of acrime, such as the use of a weapon, 2 while a finalgroup combines the two philosophies and looks forrepetition of more violent crimes.'3 Whether theprior offense need be a felony or a lesser crime isanother point of variation." Similarly, the age ofthe defendant at the titne of the prior offense 5 andthe time span between offenses' 6 are treated differ-ently in various states.' 7

If these reforms are instituted, more criminals

9 E g., N.Y. PENAL LAw (McKinney 1975) § 70.06.'0 Eg., Cowo. REv. STAT.§ § 16-13-101 (1973).'Eg., VT. STAT. As. tit. 13, § 11 (1974).'2 MAss. ANN. LAws. eh 265, § 18B (Michie/Law. Co-

op Supp. 1978).UjFg., S.C. CODE 17-25-40 (1977); TENN. CODE ANN.

§40-2801 et req. (1975).' Eg., MAss GrE. LAws ANN. ch. 279, § 25 (West

1968); Mo. ANN. STAT. § 556.280, 290 (Vernon Supp.1975).

5 Compare Ky. REv. STAT. ANN. § 532.080 (Baldwin1975) (over eighteen) with OR. REv. STAT. § 161.725(1975) (sixteen).

'6 Compare LA. Rav. STAT. § 15.529.1 (West 1967)(convictions cannot have occured more than five yearsafter the maximum sentence for the previous convictionhas expired) with Munm STAT. ANN. § 609.155 (West 1964)(convictions cannot be more than ten years apart). -

17Se Note, 45 FoymNntAm L REV. 76 (1976), for furtherdiscussion of the above statutes.

19781

PETERSILIA AND GREENWOOD

will undoubtedly be sent to prison. However, sincemany state prisons are now filled close to capacity,many prison facilities would probably have to beenlarged. Greater prosecutorial and investigatoryresources will likely be required in a system withless plea bargaining and more severe sentences.Thus, the proposed sentencing reform impliesgreater system costs. It is imperative, therefore, thatsuch reforms be preceded by an evaluation of thepotential benefits in reduced crime that wouldoffset these increased costs. Specifically, we mustdetermine how much crime is prevented by impos-ing sentences of varying length on specific classesof offenders and must predict the impact thatmandatory penalties will have on the prison pop-ulation. This article attempts to do both.

DATA BASE FOR THIS STUDY

We obtained data on a random sample of de-fendants convicted of serious offenses over a two-year period in the Denver, Colorado DistrictCourt.' 8 The sample population consists of 625persons who were convicted between mid-1968 andmid-1970 of burglary, robbery, rape, aggravatedassault, homicide, auto theft, selling drugs andgrand larceny. The sample represents 42% of thepopulation available for study.

For each person in the sample, information wascollected on personal characteristics, prior criminalrecord, court disposition of the current offense, andrecidivism during a two year follow-up period. Allof the data were obtained from secondary sources,that is, the various criminal justice records. Theserecords included probation files for personal char-acteristics, district court files for the status of cur-rent cases and final court disposition, parole filesfor the length of time served on prison sentenceand the recidivism rates of those released fromprison, and police department files for rearrest andreconviction rates.

The collection of data was completed in March1974; thus four to six years had elapsed since thetime of the conviction that placed a person into thesample.

Property-related offenses were the most commontype of crime committed by persons in the sample.Over one-third were prosecuted for burglary andanother 20% for robbery. About 19% of the sample

1 The data were originally collected for the Denver

High Impact Anti-Crime Program, reported in Charac-teristics and Recidivism of Adult Felony Offenders inDenver (1974). We are grateful to John Carr for givingus access to this data base.

members were charged with assaultive acts, andanother 10% with drug offenses. The remaining18% of the sample were charged with theft (includ-ing auto theft) or miscellaneous offenses. Of the625 convicted persons, 78.9% were found guilty ascharged; 17.8% were guilty of a lesser offense; 1%were guilty of a more serious offense; and 2.2%unknown. The average age of the sample wastwenty-six years. Other characteristics are pre-sented in the next section.

CRIME-REDUCTION EFFECTS

In attempting to estimate the effect that man-datory minimum prison sentences would have onthe adult crime rate, we focused on the followingquestions:

Crime-Reduction Potential. If defendants convicted ofcertain felonies were given mandatory prison sen-tences, how much would the overall crime rate bereduced? The violent crime rate? The burglary rate?Optimal Length of Confinement How long asentence-one, three, five, more than five years-isnecessary to reduce significantly the overall level ofcrime? The violent crime rate? The burglary rate?Optimal "Target Population". Relying on official crim-inal justice records, which characteristics define sub-populations of offenders for whom an incapacitationpolicy would be justified?

In projecting the impact of a particular man-datory-minimum sentence policy, we assumed thatall convicted defendants who meet the criteriaspecified by the policy will receive at least themandatory-minimum term. Those who receivedlonger terms in the past will continue to receivelonger terms in the future. Moreover, when esti-mating incapacitation effects, we counted only

those prevented crimes that are attributable to theextra incarceration time that results from the man-datory-minimum term. We did not include inca-pacitation effects attributable to sentences thatwould have been imposed in the absence of amandatory-minimum policy.

METHODOLOGY

Two analytic techniques have been developedfor estimating the incapacitative effects of alter-native sentencing policies. The first, a modelingapproach, uses some form of mathematical modelto estimate the relationship between sanctions andcrime rates. Parameters for these models, such asindividual crime or arrest rates, are logically de-rived from aggregate data. Analyses using this

(Vol. 69

MANDATORY PRISON SENTENCFS

approach have been reported by Clarke,19 Green-berg,20 Shinnar and Shinnar,21 and Ehrlich,2 andhave been critically analyzed by Cohen.23 WhileClarke, Greenberg and Ehrlich all conclude thatincapacitation effects are quite small, Shinnar andShinnar claim to demonstrate that they are sub-stantial. In her review of these different models,Cohen demonstrates that they all lead to similarconclusions if similar estimates of the parametersdescribing the offender population are used. Ap-proximately a 5% increase in prison population isrequired to achieve a 1% reduction in crime.

The second approach, which is the one adoptedin this study, relies on career histories to estimatethe probable incapacitation effects if offenders hadbeen sentenced differently in the past. The proce-dure takes a cohort of arrested or convicted offend-ers, examines their past convictions and determinescase by case whether each offender would havebeen imprisoned at the time of his current offenseif the sentencing policy being examined had beenapplied at the time of his last conviction.u This isthe technique used by Van Dine and his colleaguesto support their conclusion that incapacitationstrategies, even when the sentences considered aresevere, have a modest impact on violent crime2 5

For our analysis, we proceeded as follows:

1. The sample was divided into three "offense co-horts" based on the offense with which they were

19 Clarke, Getting 'Em Out of Circulation: Does Incarcerationofjuvenile Offenders Reduce Crime? 65 J. CRIU. L. & C. 528(1974).

' Greenberg, The Intcapacitative Effect of Imprisonment:Some Estimates, 9 LAw & Soc'Y REv. 541 (1975).

21 Shinnar, supra note 7, at 581.' Ehrlich, Participation in Illegitimate Activities: An Eco-

nomic Analysis, 81 J. PoL ECON. 521 (1973); reprinted inESSAYS IN THE ECONOMICS OF CRIME AND PUNIsHMENT 68(1974).

' Cohen, The Incapacitative Effect of Imprisonment: A Crit-ical Review of the Literature, in DETEReNcE AND INCAPACI-TATION: ESTIMATING THE EFFECTS OF CRIMINAL SANCriONS

ON CRIME RATES 187 (1978).2 We will sometimes speak of preventing the "cohorts'

crimes." Those crimes reflect the offenses for which thedefendants were originally charged-although approxi-mately 20 percent did not result in conviction. It isassumed that persons in the cohort, whether found guiltyor not, did in fact commit the charged crime. Thus, aperson officially charged with burglary who was con-victed of grand theft is assumed to have committed theburglary. If he had been given a mandatory prisonsentence on his immediately preceding conviction of afelony, the resulting imprisonment was counted as havingprevented the burglary.

2 Van Dine, supra note 8, at 29.

officially charged in the 1968-1970 period:

Violent: offenders charged with robbery, rape,aggravated assault, homicide, and kidnap-ping.

Burglay: offenders charged with burglary.Other Felonies: offenders charged with auto theft,

drug offenses, grand larceny, forgery, andmiscellaneous offenses.

2. Each offender's criminal record was examined todetermine whether he fell into. one of severalsubgroups of interest. For example, did he haveone, two, or three or more prior adult felonyconvictions? Did he have any prior adult convic-tions for violent crimes?

3. For each offender, the time interval between theimmediately preceding adult felony convictionand the arrest date for the current crime wasdetermined.

4. Several possible sentencing options were hypo-thetically applied to the defendant's immediatelypreceding conviction. For example, every personconvicted of a robbery who had at least one prioradult conviction was hypothetically given a one-year mandatory prison term for the prior convic-tion. Then we ascertained whether or not theseoffenders would have been in prison at the timeof their current offense. If so, the current offensewas counted as having been prevented by thatsentencing policy.

The diagram below illustrates the analytic tech-

nique employed for an offender convicted of rob-

bery on January 1, 1966, and January 1, 1967, andindicted for robbery on April 1, 1969.

5-year sentence

3-year sentence

1-year

I I I I I I I IAge 196618 Con-

victedJan. 1

1967Con-victedJan. 1

1968 1969 1970Indicted-

Apr. 1

1971 1972

This offender's 1969 robbery would be counted asnot having been prevented by the 1-year sentencebut having been prevented by a 3- or 5-year sen-tence since he would have been in prison on AprilI, 1969!7

' We sometimes refer to the sample as a whole as theall-felonies cohort.

'7 For those defendants with more than one priorconviction, as in the example above, we focused only onthe most recent prior conviction. However, if the man-

PETERSILIA AND GREENWOOD

SAMPLE CHARACTERISTICS

In dividing the sample into cohorts based uponthe charge of their current crime, we might expectthat the persons in the cohorts would differ signifi- ocantly with respect to prior criminal record or courtdisposition of the current offense. Tables I and 2show the criminal history and disposition of the c Ecase for each offense cohort. The tables show that 1 c q ,the average offender in the burglary cohort was -'<slightly younger than his counterpart in the othercohorts; that the "other felonies" cohort was .slightly more likely to have a adult record, proba- .) g . o

bly because it was the oldest group; and that the < t-

violent cohort received the most severe sentencing "treatment by the courts. The other differences were '-insignificant.. - -0 0 --

FINDINGS ^

The Violent Cohort

We measured the extent that the violent cohort's -*. _crimes would have been prevented by the imposi- C < -

tion of mandatory prison sentences for their pre- -ceding adult felony conviction. Seven hypotheticalsentencing policies were considered: co- -In o -C

<0Sentencing Option I: Each offender convicted of an 1 Z-u

adult felony, violent or not, is sentenced to a 0 c-Vmandatory prison term of one, three, or five L%

0years.

<

Sentencing Option II: Each offender convicted of an -adult felony whose criminal record shows at least 0 -t-

i -- 'ione previous adult conviction is sentenced to a -mandatory prison term of one, three, or five <years on each conviction after the first. On the E, -

first adult conviction the penalty structure re- C) 1 E E 2.mains as under present law. ,t <

EC

Sentencing Option III: Each offender convicted of an 0cadult felony whose criminal record shows at least cq -0two previous adult convictions is sentenced to a 0

mandatory prison sentence of one, three, or five -,

years after the second conviction. On the first - 0-,. - .5two convictions the penalty structure remains as j .. -0 -c 'aunder present law.

Sentencing Option IV.- Each offender convicted of a ' ._ _ _

violent felony is sentenced to a mandatory prison 0 ", -Rterm of one, three, or five years. .

. 0 C) u u C)

datory-minimum policy had been in effect over the de- ".R ,C7C)-fendant's whole career, the second prior conviction might -U w

have been prevented by a sentence imposed on the firstprior conviction. Thus our simplification will result in aslightly inflated estimate of the true incapacitation effect.

[Vol. 69

MANDATORY PRISON SENTENCES

Sentencing Option V." Each offender convicted of aviolent felony whose criminal record shows atleast one previous adult conviction (not neces-sarily for a violent crime) is sentenced to amandatory prison term of one, three, or fiveyears after the first conviction. On the first adultconviction the penalty structure remains as un-der present law.

Sentencing Option VI: Each offender convicted of aviolent felony whose criminal record shows atleast two previous adult convictions (not neces-sarily for violent crimes) is sentenced to a man-datory prison term of one, three, or five yearsafter the second conviction. On the first twoconvictions the penalty structure remains as un-der present law.

Sentencing Option VII: Each offender convicted of a

violent felony whose criminal record shows atleast one previous adult conviction for a violentcrime is sentenced to a mandatory prison termof one, three, or five years after the first convic-tion. On the first adult conviction the penaltystructure remains as under present law.

Figure 1 shows the percentage of the violentcohort's offenses that would have been preventedif the offender had been sentenced alternativelyunder each of the seven mandatory sentencingoptions for his immediately preceding felony con-viction.

Under option I, 10.9% of the cohort would havebeen prevented from committing their violent of-fense in the 1968-1970 period by a one-year man-datory prison sentence; 22.2% would have beenprevented by a three-year sentence, and 31% by a

TABLE 2

DISPOSmON OF CURRENT OFFENSE, BY COHORT

Cohort Percentage of Specified Cohort Av. Length of Incar-ceration (years)*

Convicted on Given Given Given Given GivenCohort Charges Probation Jail Reformatory Prison Other

Violent 96 27 9 25 36 2.5 1.3Burglary 69 28 8 34 25 4 1.1Other felonies 71 6 25 32 28 8 .9

All felonies 79 22 13 32 30 4 1.1'These averages represent a combination of the percentage of persons who were convicted but did not serve any

institutional time, and the average time served by those sentenced to jail, reformatory, and prison.

50

40 1(ay felony)

.30.2

. |l(arr fIony, I prir)

120 12

/ IIl(any futony, 2 pion)- _-vIa (,fofy. f.l,

- 1 3 5 >5

FIGURE 1Percentage of violent cohiort offenses prevented by mandatory sentencing options

19781

PETERSILIA AND GREENWOOD

five-year sentence. The data suggest that the inci-dence of violent crime might be lessened by one-third if every adult defendant convicted of a felony,regardless of prior record, were imprisoned for aperiod of five years. The maximum crime reductioneffect possible under such a sentencing scheme is40%. The "> 5 years" category on each of thegraphs can be interpreted as the maximum benefitderived under the different policies.

The less stringent sentencing options, options IIthrough VII, impose mandatory prison sentencesonly on defendants whose criminal records containprevious adult convictions. For example, option IImandates imprisonment for persons previouslyconvicted of at least one adult felony; under thisoption, a five-year sentence would have prevented15.8% of the violent offenses.

Even more restrictive sentencing options-forexample, those that require the defendant to pos-sess two priors or have convictions for violent of-fenses-prevent many fewer crimes. Such policies,even with a five-year imprisonment, reduced theviolent crimes of the cohort by less than sevenpercent.

On the surface, these data tell us only that acertain percentage of the crimes for which themembers of this cohort were officially chargedwould have been prevented by a particular man-datory-minimum sentence. Given the existing lowrates of crime clearance and conviction, this num-ber represents a very small proportion of the totaladult crime reported in a given period. However,the use of this small percentage as a measure ofincapacitation implicitly assumes that each of theoffenders in the sample committed only onecrime-the crime for which he was officiallycharged. It further assumes that other active of-fenders, not arrested during the two-year period inquestion, did not experience earlier convictionsthat would have resulted in their incapacitationduring the period.8s The use of these extremelyconservative assumptions clearly leads to a largedownward bias in the resulting incapacitation es-timates.

A more appropriate assumption would be thatthe offenders in this sample represent a random

2' In other words, if we use only the crimes for whichthe offenders in our sample were convicted as a measureof incapacitation, we would be ignoring the crimes ofdefendants who were released and not subsequently con-victed during the period (i.e., the "successful" offender).Given the low probability of arrest and conviction forproperty crimes, there is reason to believe that theirnumber is substantial.

sample, in terms of prior record characteristics, ofall offenders who were criminally active during the

period of the study. In other words, if 30% of our

sample have prior felony convictions, we assumethat 30% of the crimes which were not cleared byarrest were also committed by offenders with prior

felony convictions.Theoretically, there are two reasons to suspect

that a random sample of persons arrested or con-victed is not truly representative of all persons who

are engaged in crime in a particular period. On the

one hand, any random group selected on the basis

of an arrest or conviction will tend to overrepresent

offenders with higher crime rates; by definition

they have a greater likelihood of entering the sam-ple. This biasing effect would tend to make the"random sample" assumption conservative, over-

representing the high-rate offender. On the other

hand, we know that arrestees and convicted persons

overrepresent the offenders who commit crimesagainst victims who can identify them and who

therefore have a higher probability of arrest.2 This

bias would tend to make the "random sample"assumption too liberal, in overrepresenting the less

sophisticated offender. Since the available evidenceis inadequate for sorting out these two conflicting

effects, we cannot be sure in which way the final

estimate will be biased. Yet, the "random sample"assumption appears to provide a more reasonable

basis for interpreting those results than any other

assumption that we can make at this time.In the rest of this article we will adopt the

"random sample" assumption and will refer to the

percentage of cohort crimes prevented as the per-

centage of all reported crimes prevented. Also, inthe model adopted here, there are no assumptions

concerning the criminal lifetime of offenders or of

a Poisson process for crime commissions, as in most

studies of incapacitation effects.

The Burglary Cohort

The sentencing schemes applied to the burglary

cohort resembled those imposed on the violent

cohort. Options I, II, and III were identical; optionsIV, V, and VI substitute burglary convictions forviolent convictions as follows:

Sentencing Option IV. Each offender convicted of a

burglary is sentenced to a mandatory prisonterm of one, three, or five years.

2 See VERA INSTITUTE OF JUSTICE, FELONY ARRESTS:THEIR PROSECUTION AND DisPosmON IN NEw YORK CITY'SCOURTS (1977).

[Vol. 69

MANDATORY PRISON SENTENCES

Sentencing Option V- Each offender convicted of aburglary whose criminal record shows at leastone previous adult conviction (not necessarilyfor burglary) is sentenced to a mandatory prisonterm of one, three, or five years after the firstconviction. On the first adult conviction thepenalty structure remains as under present law.

Sentencing Option VI: Each offender convicted of aburglary whose criminal record shows at leasttwo previous adult convictions (not necessarilyfor burglary) is sentenced to a mandatory prisonterm of one, three, or five years after the secondconviction. On the first two convictions the pen-alty structure remains as under present law.

No option VII was applied to the burglary cohort.Figure 2 shows the percentage of the 1968-1970

burglary cohort offenses that would have beenprevented by each of the six mandatory sentencingschemes. Comparison of the effects of the identicaloptions I, II, and III in Figures 1 and 2 suggeststhat mandatory sentencing may have greater po-tential for reducing the incidence of burglary thanof violent crimes.

Under option I, 42% of the offenders in theburglary cohort would have been prevented fromcommitting their current burglaries if they hadbeen imprisoned five years for their last adultfelony conviction; 31% would have been prevented

by a three-year imprisonment; 14% because of aone-year imprisonment. The most stringent option(option I, five-year imprisonment) would have pre-vented nearly one-half of the burglaries.

Option II, which limits mandatory prison sen-tences to offenders with at least one prior adultfelony conviction, would have prevented few bur-glaries with one- or three-year imprisonments (1.9%and 8.5%, respectively). Options IV-VI, which aremore conservative, also would have prevented onlya few of the burglaries (approximately 1-12%).

Entire Sample

Figure 3 presents the percentage of all 1968-1970felonies that would have been prevented if man-datory prison sentences had been imposed underoptions I, II, and III for the defendants' immedi-ately preceding adult conviction.

It has been suggested that every person convictedof a serious felony should receive some imprison-ment. Sentencing option I measures"the number ofthe entire sample's crime that would have beenprevented under this principle. In applying themost stringent form of this option (sentencing everyperson convicted of a felony to five years' imprison-ment), we found that only 45% of the crimescommitted by our sample would have been pre-vented. Option II, which sentences only personspreviously convicted of an adult felony to five years

[(any felony)

/ II(cmy felony, I prior)

//

// --/ ,(any felony, 2 prior,)

V(burglary, I prior)

...... IV(bglry). - VI(burglry, 2 priors)

I 3 5 >5Sentence length (yr)

FIGURE 2Percentage of burglary cohort offenses prevented by mandatory sentencing options

. .. .. 6111978]

PETERSILIA AND GREENWOOD

I (any felony)50

40

X 30 II (any felony, I prior)

20

u20 I1l(ony felony, 2 priors)

10

3 5 >5Sentence length yr)

FIGURE 3Percentage of entire sample's offenses prevented by mandatory sentencing options

in prison, would have prevented only 18% of thecrimes.

These results cause us to doubt that mandatory-minimum sentences can easily reduce the crimescommitted by adults. To reduce the level of crimein half, every person convicted of a felony, regard-less of prior criminal history, would have to beimprisoned for five years. If the length of imprison-ment were three years, the crime level could bereduced by a third; if it were one year, the reduc-tion would be 15%. Moreover, to reduce violentcrime by one-third, every person convicted of arobbery, rape, aggravated assault; homicide, andkidnapping would have to be imprisoned for fiveyears. Even if violent offenders were imprisoned formore than five years, violent crimes would only bereduced by about 40%. Finally, the burglary ratecould be reduced by nearly half (42%) if all de-fendants were imprisoned five years for their pre-ceding adult conviction. With a three-year im-prisonment, crime would be reduced by 31%.

Some may regard the policy of sentencing everyconvicted person to prison as too harsh, since itdoes not take into account the offender's priorinvolvement in crime. However, these data suggest.that if only defendants who have a prior adultconviction are imprisoned, the crime-reduction ef-fect would be just about one-half that of the re-duction produced by sentencing every convicted

felon to prison. The crime reduction effects dis-cussed here may not be-as large as one would haveexpected. But, it must be remembered that thefigures presented above represent only the effectsdue to incapacitation and do not take into accountadditional benefits resulting from deterrence orrehabilitation. Additionally, the policies considereddeal only with adult sentencing, ignoring any al-ternative policies aimed at juveniles.

EFFECT ON PRISON POPULATION

Before opting for sentencing reforms, policymak-ers must weigh their projected benefits in crimereduction against their likely costs in increasedprison population. Advocates of new mandatorysentencing schemes have generally ignored thelikely extra burden on the correctional system. Inthis section, we consider the effect that the hypo-thetical sentencing options discussed in the preced-ing section would have on the prison population. 3

0

30 For simplicity, we do not consider the total prisonpopulation but only that deriving from our sample ofoffenders. Our calculations measure the additional man-years of p. *son time to be served by the offenders sampled.For convenience, we translate these additional man-yearsinto increases in the prison population. This translationassumes that enough time passes to achieve a new steady-state prison population, and that the prior offense char-acteristics of persons convicted do not change signifi-cantly.

[Vol. 69

MANDATORY PRISON SENTENCES

Specifically, we estimate the percentage increase inthe prison population if every convicted felon inthe sample had been given a prison sentence ofone, three, and five years, and we predict thepercentage increase if every convicted felon with atleast one prior adult conviction had been given aone, three, or five year prison sentence.

In projecting the impact of these mandatoryminimum policies on future prison populations,one should take into account that some offenderswould have received sentences in excess of theminimum mandated by the new policy. Therefore,applying these mandatory sentences to them wouldtend to reduce rather than increase the prisonpopulation. We have made no attempt to predictthe "reduction" effect here. For the most part, theeffect would be minimal, as the hypothetical sen-tences are considerably harsher than those nor-mally imposed.

METHODOLOGY

To estimate the percentage increase in prisonpopulation, we proceeded as follows:

1. For each hypothetical sentence (&g., all personsconvicted of a violent felony with one prior adultconviction will receive a three-year sentence), wedistinguished three parts of the Denver sample:

Those offenders who did not qualify for thesentence (not convicted of a violent felonyor had no prior record).Those offenders who qualified for the sen-tence but who were already sentenced toterms longer than the mandatory-mini-mum, and hence would be unaffected bythe new policy.Those offenders who qualified for the sen-tence, had currently received less than themandatory-minimum sentence, and whowould be assumed to serve exactly the man-datory-minimum.

2. Members of the third group were the only oneswhose sentences would be increased under thenew mandatory-minimum policy. The total(minimum) increase in time to be served by thisgroup is the difference between the mandatoryminimum and the average time now served (S- T3), multiplied by the number of offenders inthe group (N3).

No attempt is made to distinguish between prison andjail time. Since some mandatory prison terms wouldresult in a transfer of prisoners from jail to prison, theactual percentage increase in prison population will belarger than estimated. And since jails contain mostlymisdemeanants and detentioners awaiting trial, the ac-tual increase in total incarcerated population will be lessthan estimated.

3. The percentage increase in prison populationowing to the third group can be estimated as theincrease in time to be served (calculated in step2), divided by the total time to be served by theentire sample, i.e.,

(S - T)N3T iNt + T2N2 + T.N.'

where S = mandatory minimum sentencelength

Ti = average time served by inmatein group i

Ni = number of inmates in group i.

FINDINGS

Figure 4 shows the percentage increase in theprison population if the entire sample, regardlessof prior criminal history, had been given a man-datory minimum sentence of one, three, or fiveyears. The results are shown separately for thoseconvicted of a violent felony, burglary, and allfelonies. At the extreme, if every person convictedof a felony had been sentenced to a minimum offive years, the prison population would have in-creased by 450%; if each were sentenced to a three-

1 3 5- , , Ien th (yr)

FIGURE 4Increase in prison population with mandatory mini-

mum sentences for every convicted felon

PETERSILIA AND GREENWOOD

year minimum, a 230% prison increase would haveoccurred; and if each were sentenced for only oneyear, a 50% increase would have resulted.

If every person convicted of a burglary had beensentenced to a minimum of five years, the prisonpopulation would have risen by 150%; if each weresentenced to three years, there would have been a75% increase; and if each were sentenced for oneyear, a 25% increase would have occurred. Finally,if every person convicted of a violent felony hadbeen given a five-year mandatory minimum sen-tence, the prison population would have risen by160%; if each were given a three-year sentence,there would have been an 80% increase; and ifeach received a one year sentence, a 25% increasewould have resulted.

Figure 5 shows the impact of a more selectivepolicy, under which those who had been convictedof at least one prior adult felony received manda-tory prison sentences. As the figure shows, limitingthe mandatory-minimum sentence to persons pre-viously convicted seems to be a more plausiblepolicy option because the impact on prison popu-lations is more reasonable. If every person in thesample who had at least one prior felony convictionhad been sentenced to one year in prison, theprison population would have risen about 15%; ifeach were sentenced to three years, an 80% increasewould have occurred; and if each were sentencedto five years, there would have been about a 190%increase.

CONCLUSIONS AND POLICY IMPLICIATIONS

We have analyzed the impact of various man-datory-minimum sentences on the adult crime rate

1 3Sentence length (yr)

FIGURE 5Increase in prison population with mandatory mini-

mum sentences for at least one prior conviction

and on the prison population. The results aresummarized in Figures 6 and 7 for the all-feloniescohort and for the violent cohort respectively. InFigure 6, the lower curve is generated by varyingthe mandatory-minimum prison term (one year,three years, five years) under sentencing Option I,which imposes the mandatory-minimum term onall convicted felons regardless of prior record. Theupper curve corresponds to Option II, which im-poses the mandatory-minimum term (ofaltemative

100

00 I- I(..y lony, I p"l)

any fo )

Length of mwentoymInimun sentence

o I yer0 5 years

o Sy t i t

0 50 100 150 200 250 300 350 400 450Percentage increase in prison population

FIGURE 6Effect on crime rate and prison population of sentenc-

ing options I and II and a range of mandatory minimumsentences-all felonies cohort

00 V (violent felony, I prioer)

IV (Violet relony)

90

1 (oy felony, I prior)

~70S Length of maondotory-inimum sant e-a

* year60 3yer

0 5 years

050 10O 150 200 250 300 350 400 450 500

Parenotog. increase in prisn poplatIon

FIGURE 7Effect on crime rate and prison population of sentenc-

ing options I, II, IV and V and a range of mandatoryminimum sentences-violent cohort

[Vol. 69

MANDATORY PRISON SENTENCES

lengths) only on convicted felons- having at leastone prior felony conviction.

Figure 6 clearly reveals the trade-offs betweencrime rate and prison population under the sen-tencing alternatives considered. For example, thecrime rate would be reduced 15% (compared to85% for the current level) and the prison popula-tion would (eventually) increase 50% if every per-son convicted of a felony were imprisoned for oneyear (Option 1). To lower the crime rate by thesame 15 percent under Option II, it would benecessary to impose a mandatory-minimum termof approxiamtely four years on felons with a priorconviction, which would raise the prison popula-tion by 125%.

Figure 7, showing the relation between crimerate and prison population for the violent cohort,addresses four sentencing options. We see that themore restrictive the prior-record condition thatdefines the group to which the option applies, thesmaller its effect on crime rate. For example, athree-year mandatory-minimum term produces a22% reduction in crime rate under Option I, and11% reduction under Option II, a 3% reductionunder Option IV and 1% reduction under OptionV-with the corresponding increases in'prison pop-ulation being, respectively, 225, 87, 87, and 27%.Contrast these results with those in Figure 6, whichshowed that a three-year mandatory-minimumterm reduces the crime rate by 34% under OptionI and 12% under Option Il-while raising thepopulation 225 and 87%, respectively. -

This finding-that a better crime-reduction/prison population trade-off would result from im-prisoning all felons for a short period regardless ofprior record, may go against expectations thatcrime could be most effectively controlled by con-centration on the offenders with prior convictions.At this time, we can only speculate as to why thisrelationship is found. We know that under currentsentencing policies, convicted offenders are muchmore likely to be sentenced to prison if they havea prior criminal record. Within this sample, only57% of those defendants convicted with no priorrecord were incarcerated, while 90% of those withprior records were incarcerated. We also know thatalthough the recidivism rate increases with anyprior record, this increase is minimal when com-pared with the increase in likelihood of prisoncommitment.

We have referred earlier to the disparity in thesentences received by defendants with similar crim-inal records, convicted of similar crimes. Some ofthe disparity reflects the different philosophies heldby different judges. However, disparity is found in

the sentencing decisions of even a single judge.With respect to a series of similar defendants con-victed of similar crimes, he may give some defend-ants a straight probation, others a prison term, andstill others a combination of probation and jailtime. In making his sentencing decisions, a judgetries to assess the risk an offender will pose to thecommunity if he is not confined and the impactthat imprisonment would have on him and hisfuture behavior. In so doing, the judge avoids anunjustly mechanical application of sentencing lawsbased only on the convicted charges and priorrecord.

For example, not all convicted armed robberswith prior records receive a prison sentence. Indistinguishing the cases in which a prison term wasimposed from those in which it was not, we wouldbe likely to find systematic differences related tothe judge's appraisal of the risk to the community.Those not sentenced to prison would have lessserious prior records or better family and commu-nity ties, better employment records, etc., thanthose given prison terms. If there is a positivecorrelation between the risk estimated by thejudgeand the actual risk posed by the defendant, wewould expect to find that, on the average, defend-ants with serious prior records who are not givenprison sentences have a lower recidivism rate thanthose convicted of similar offenses with minor priorrecords who are not given prison sentences. Whilewe cannot prove that the relationships hypothe-sized above are in fact true, they are at leastconsistent with the crime reduction prison popu-lation curves in Figures 6 and 7 and they do deserveattention in future studies.'

In summary, mandatory-minimum sentence pol-icies can reduce crime through incapacitation ef-fects, but substantial increases in prison popula-tions will be required to achieve modest reductionsin adult crime. Our analysis indicates that for aone percent reduction in crime, prison populationsmust increase by three to ten percent, dependingon the target population to be sentenced. Thesefigures are consistent with the five to one ratioestimated by Cohen in her review of other inca-pacitative models.31 Our analysis suggests furtherthat mandatory-minimum sentencing policies thatfocus only on defendants with prior records, al-though they may accord better with the notion ofjust deserts, are less efficient. Such policies result inless crime reduction for a specified increase inprison population than those policies that applyregardless of prior record.

31 Cohen, supra note 23, at 188.