sign in sign up
<<
Home , Criminal procedure

T O EN F J TM U U.S. Department of R ST A I P C E E D

B

O J Office of Justice Programs C S F A V M F O I N A C I J S R E BJ G O OJJ DP O F PR National Institute of Justice JUSTICE National Institute of Justice R e s e a r c h i n B r i e f Jeremy Travis, Director November 1999 Issues and Findings Effects of ’ Sentencing Discussed in this Brief: This study sought to determine the degree to which judicial sentencing decisions Decisions on Criminal Careers affect subsequent criminal careers. It By Don M. Gottfredson examined the criminal careers of 962 felony offenders in Essex County, New Jersey, sentenced in 1976 and 1977 Despite the absence of adequate data on sions based on case and offender variously to confinement and noncus- todial programs. The 18 participating the effects of incarceration and other sanc- characteristics. tions on convicted offenders’ subsequent judges exercised considerable discre- ● Risk—differently sentenced groups may tion in making sentencing decisions. , sentencing trends over the past represent different risks of future offend- The data collected included judicial two decades have moved toward increased ing at the time their sentences are im- perceptions, the judges’ predictions of determinacy, greater use of mandatory sen- the offenders’ future criminal behav- posed (usually called the a priori risk). ior, the judges’ sentencing purposes, tences with longer terms, and reduced judi- offender backgrounds, execution of cial discretion. A better understanding of ● Incapacitation—offenders are confined sentences, and offenders’ and how different felony sanctions impact the for differing periods of time as a result of charges during the 20 years after future behavior of offenders is needed to the imposed on them for the cur- sentencing. Also measured were the provide a basis for evaluating the efficacy rent offense and they commit later. judges’ selection of different sanc- tions, the validity of subjective and of current sentencing policies. These factors must be considered when objective predictions of future criminal To this end, the National Institute of Jus- assessing the effects of punishment, treat- behavior (risks), and the offenders’ ment, specific deterrence, and incapacita- time in the community (free of the in- tice sponsored research that examined the capacitating effects of jail or ). control effects of sentences, over a tion. The results of controlled studies of 20-year period, on 962 felony offenders different punishments cannot be compared Key issues: Rigorous tests of sen- sentenced in 1976 or 1977 in Essex because experiments with different sen- tencing policy changes are rare tences for equivalent groups of offenders— because serious obstacles impede County, New Jersey. This Research in assessments of the crime control Brief summarizes findings of the study. for example, by random assignment of effects of such policies. The results sentences—rarely are believed to be fea- of controlled experiments involving sible and are not often done. different punishments of equivalent Research issues groups of offenders cannot be Three main factors bias comparisons of Assessment of sentencing policies ordi- compared because such studies, outcomes for differently punished groups narily is hindered by several problems rarely believed to be feasible, are that cast doubt on the relevance of not done. of offenders: selection of sentence, risk of reoffense, and time in confinement. Fair findings about sentencing choices: Other obstacles to sentencing re- comparisons of sentencing choices must search include: judges’ reluctance ● The most rigorous research methods, take these sources of variation in outcomes to provide detailed records of their in which offenders would be randomly reasons for choosing particular into account: assigned to various sentencing alterna- sentences; fixed or mandatory sen- tives, usually are precluded. tencing that limit sentencing ● Selection—groups given different alternatives available to judges; and sentences are not equivalent because ● Judges rarely are willing to record in the demands of careful, long-term judges make their sentencing deci- detail their views on the seriousness followup analysis of sentencing effects on offenders’ subsequent criminal careers. continued… R e s e a r c h i n B r i e f

Issues and Findings of the offense or the offender’s prior were allowed much discretion in their continued… record, their assessment of the risk that decisionmaking. Besides documenting the offender will commit future crimes, the sentences they imposed, the judges This study included careful statistical controls for biasing factors (judicial their ratings of other case characteris- recorded their perceptions of the serious- selection of sentences, offender risk tics, or their specific reasons for select- ness of the crime and the offender’s past of reoffense, time in confinement, ing the sentences they impose. and their assessment of and time in the community), de- the likelihood that the offender would ● As changes in sentencing policy have tailed recording by judges, availabil- commit future crimes. ity of broad sentencing discretion, taken place over the past two decades, and study of offenders’ criminal judicial discretion often has been re- The central issue of this study was the effects careers over the 20 years after duced by greater determinacy in sen- sentencing. of different sanctions on the offenders’ subse- tencing structures or stricter mandatory quent criminal careers. Secondary issues Key findings: The judges’ subjec- sentencing laws or guidelines. As a concerned the measurement of judicial se- tive risk assessments of offenders’ result, judges have not had as much lection of different sanctions, the validity of likelihood of recidivism, although choice in imposing sentences as they only modestly valid, had a substan- subjective and objective predictions of risks tial influence on their sentencing did previously. for reoffending, and the amount of time the choices. More formal, empirically ● Perhaps most important, researchers offenders were in the community free of the derived methods provide better incapacitating effects of jail or prison. The measures of the risk of reoffending. must carefully trace the offenders’ records through any periods of confine- effects of sentences on later arrests and Available sentencing choices had ment resulting from the sentence (and charges were assessed through statistical or little effect—other than that of inca- from later crimes) to determine the “quasi-experimental” methods that adjusted pacitation—on recidivism as mea- sentence as actually executed and then for the nonequivalence of differently sen- sured by new arrests and charges: observe new arrests, , and tenced groups. ● Whether the offender was con- sentences after their release into the fined or given noncustodial sanc- Data on the imposition and actual execution community. Sufficient time must elapse tions made no difference. of sentences and postconviction arrests, after sentencing to enable most offenders ● charges, sentences, and confinements over Where the offender was con- to complete any terms of confinement fined (i.e., in jail, a youth facility, the next 20 years were examined. The and then to experience enough time free or prison) made little difference. followup data collection ended in February in the community to permit a fair assess- ● 1997. It relied on the New Jersey Depart- The length of the offender’s ment of the effects of the sentences on maximum imposed sentence ment of Offender-Based Correc- their criminal careers. made no difference. tional Information System and record files, ● The length of time the offender These problems were mitigated in this the U.S. Department of Justice Interstate actually was confined made little study by using careful statistical controls Identification Index, the National Crime In- difference. for biasing factors, detailed recording formation Center Wanted Persons File, and ● A “split” sentence of jail and by the judges, the availability of broad sen- the New Jersey PROMIS/GAVEL Prosecu- made no difference. tencing discretion, and the detailed study tors Case Tracking System. Staff familiar ● Fines or made no of offenders’ criminal careers during the with each of these systems coded the data. difference. 20 years after they were sentenced.1 No records after sentencing could be found for 37 persons (4 percent) who were counted Some important lessons for current 2 policy are clear from this study; it Study methods as having no known subsequent arrests. offers little support, aside from incapacitation, for increased use of Beginning in May 1976, 18 judges in the Measures of sentence selection and the confinement, emphasis on longer Essex County, New Jersey, recorded offender’s a priori risk (risk as estimated terms, or more acceptance of their judgments when sentencing 962 from data available at the time of sentenc- specific deterrence as a crime felony offenders to confinement and non- ing) were developed using multivariate control strategy. custodial programs. They continued to statistical methods. Time variables were Target audience: Judges, legisla- code the cases they sentenced during the calculated from recorded dates. Differences tors, , attorneys, next year. Under New Jersey’s then inde- in later arrests and charges (at 5 years and commissions, terminate sentencing , the judges 20 years after sentencing) associated with corrections administrators, research- ers, and others concerned with sentencing policy. 2 R e s e a r c h i n B r i e f different sentences were analyzed with Exhibit 1. Average percentage of points assigned by each to each statistical controls for measures of the purpose of incarceration* biasing factors.3 The variability in new arrests and charges was analyzed to 1 55 assess sentence effects, and survival 2 56 without over time was examined 3 56 for differently sentenced groups. 4 12 5 47 6 56 Number of Cases Crimes, demographics, results 7 59 8 58 The offenders had been convicted of 9 52 crimes typically considered to be felo- 10 57 nies. These crimes were classified by 11 58 12 60 legal definitions and groupings based Participating Judges on a multidimensional approach to 13 59 14 60 4 scaling offense seriousness. By legal 15 56 offense classifications, 37 percent were 16 36 convicted of violent crimes, 24 percent 17 59 of drug offenses, 23 percent of 18 60 956 crimes, 10 percent of weapons offenses, 0% 20% 40% 60% 80% 100% and 6 percent of other crimes. By the Mean Percentage of Points Assigned to Purposes by Judges behavioral classification based on judg- ments of seriousness, one in six was Retribution Rehabilitation Incapacitation convicted of an offense involving inter- General Deterrence Specific Deterrence Other Purpose personal confrontation or physical vio- lence, although property crimes were * Data were missing for 6 cases. Points assigned by judge 7 were prorated to total 100. (Ratings did not most common. Fifteen percent were always total 100.) serious drug offenders whose crimes included the sale of drugs other than of violent crimes. One in 10 had been bation. The main choices, in terms of marijuana. committed to a juvenile facility. executed sentences, were a noncusto- dial sentence (42 percent), jail (29 On other measures, the study sample Research results described the sen- percent), prison (19 percent), or a was fairly typical of sentenced offender tences as imposed (announced) and as youth facility (10 percent). About 10 populations. The average age was 29. executed (carried out), the sentencing percent received “split” sentences of Most were black males; one-fourth were purposes of the judges, and later arrests probation with jail. Others received white; one-eighth were women. About and charges. Next, the validity of the special conditions, including fines and half had served prior jail terms; 16 per- judges’ predictions and the measures of restitution. cent had been in prison; and 80 percent a priori risk and sentence selection had prior placements on probation. On needed for the analyses of effects were Sentencing purposes. Judges most average, they were first arrested at age considered, as described below. often reported a crime control aim as 22, had 2.4 prior arrests and convic- the main reason they imposed the sen- tions, and were currently charged with Sentences tences they did. Rehabilitation and two counts and convicted of one. One in specific deterrence were prominent 10 was reported to have used heroin or Most offenders were sentenced to in- considerations. Other sentencing goals barbiturates as a juvenile, and 29 per- carceration with terms ranging from included incapacitation and general cent had a recorded use of these drugs 1 month to life (typically 6, 12, or 18 deterrence. Retribution was the pri- in the 2 years prior to their arrest. One- months in jail or 5, 7, or 10 years in mary purpose in about one in five third had histories of drug offenses; prison). Sentences were suspended in cases. Judges typically had more than nearly half had a record of property whole or in part for many offenders, one purpose. The distribution of offenses; and one-third had a history who then usually were placed on pro- ratings is shown in exhibit 1. The

3 R e s e a r c h i n B r i e f cases were assigned sequentially ac- Charges classified by legal offense study, are shown in exhibit 2. The num- cording to the availability of the judge, groups. In terms of the legal offense bers and percentages are those for the and all judges participated. It cannot classifications, in the 20 years after most serious charges among any made be assumed that the judges received their initial sentencing, the offenders after the first arrest in the followup entirely equivalent cases. were charged with 40 criminal homi- period. The total numbers of offenses cides ( or manslaughters), 455 charged during the followup period, Offenses after sentencing , 752 assaults, 928 burglar- grouped into these general classes, are ies, 18 rapes, 507 weapons offenses, shown in exhibit 3. Thirty percent of the offenders were 682 receiving stolen property offenses, not arrested in the 20 years after sen- and 16 kidnapings—to list some Judges’ predictions tencing. More than half of all, how- charges out of 9,346 allegations of new ever, were rearrested in the first 5 crimes. The median number of charges Judges’ subjective predictions of years after they were sentenced. The was 5, while the mean was 9.7. whether the offenders would commit 962 offenders were arrested 5.3 times any new property, violent, or any type on average in the 20 years after sen- Charges classified by behavioral of crimes were valid but modestly so. tencing. The arrest rate for all offend- offense groups. The offenders in this Predictions of “any new arrest” ac- ers participating in the study, not sample were arrested 5,117 times during counted for about 6 percent of the dif- including the time spent incarcerated, the 20 years after sentencing. Most of- ferences in new arrests. The validity of was 0.28 arrests per year. The rate for ten, by far, they were charged with prop- some judges’ predictions compared fa- offenders who were rearrested at least erty crimes or nuisance offenses, but vorably with some empirically derived, once was 0.36. For those rearrested, they also were charged with a substantial formal prediction methods. Yet, the the average time to a new arrest was number of serious personal confrontation predictions of 4 of the 18 judges were 3.5 years; half of those with new ar- crimes and drug offenses.5 The charges invalid as to whether the offenders rests were rearrested within 2.2 years. at the first arrest after sentencing, classi- would commit any new property, vio- fied into kinds of offenses as used in this lent, or any crimes; would be charged with new offenses; or would be rear- rested. The predictions of seven judges Exhibit 2. Charges at first arrest after sentencing, by behavioral offense as to whether the offenders would com- Number of Percentage of mit any new crimes had no validity. Type of Charge Persons Total The validity of the judges’ predictions None 288 29.9 of property crimes was similar to Nuisance 269 28.0 35 3.6 that for any new offense. Predicting Property 248 25.8 whether the offenders would commit Serious drug 7 0.7 crimes against other persons proved to Personal confrontation 115 12.0 be a more difficult problem because 962 100.0 these crimes are relatively infrequent. The judges’ predictions accounted for less than 2 percent of the differences in Exhibit 3. Number of charges in the followup period, by behavioral offense new arrests for offenses involving inter- Number of Percentage of personal confrontation. The predictions Type of Charge Charges Total of violent crimes by 12 of the judges had Nuisance 3,584 38.3 no statistically significant relationship to Fraud 376 4.0 new violent offense arrests or charges. Property 3,642 39.0 Serious drug 93 1.0 The judges who would be better Personal confrontation 1,175 12.6 predictors, of course, could not be Other 476 5.1 identified beforehand. The judges’ 9,346 100.0 predictions appeared to be influenced

4 R e s e a r c h i n B r i e f

about whether to order confinement Exhibit 4. Factors considered by judges in making subjective predictions of risk of any new crime by an (“in-out”), selection of the four main sentencing alternatives (noncustodial, jail, youth facility, or prison), selection Long arrest record of the three main custodial sanctions, and selection of “split” sentences. Serious offense Selection for any confinement Low social stability (in-out). In deciding whether to order confinement, judges appeared to be Probation officer recommends custody influenced mainly by their assessment Judges’ of rehabilitation as an important aim of predictions of sentencing in the particular case, the any new crime recommendation of the probation of- Not a person crime ficer for confinement (unless the pro- bation officer and the judge used the Aggravating factors same information in arriving at the decision), their own prediction of Long record whether the offender would commit future crimes, and the seriousness of Age (younger) the offense (see exhibit 10). These and other items enabled correct

Exhibit 5. Validity of judges’ mainly by their assessments of the crime (property or serious drug offense). predictions of any new crime seriousness of the offense, their judg- In combination, these and other factors ments of the offender’s social stability, shown in exhibit 6 accounted for 23 Percentage With New Arrests (20-Year Followup) and the length of the offender’s arrest percent of the differences in new arrests 100 and conviction records. The factors (see exhibit 7). The same measure 91 90 that explain the judges’ risk predic- substantially correlated with other out- 84 comes, such as the total number of 82 tions for any new crime are shown in 80 exhibit 4.6 The validity of the judges’ arrests and charges. 76 predictions is illustrated in exhibit 5.7 Risk measure 2 included most of the 70 measures in risk measure 1, plus such 60 57 Empirical measures of risk factors as the offender’s sex, number of Two more formal risk measures were prior probation sentences, and prior 50 developed for use in statistically con- incarceration for probation or trolling for the offenders’ a priori risk violations (see exhibit 8). The scores 40 of new arrests. Each was based on on this measure accounted for 31 per- information available at the time of cent of the variation in the any new 30 sentencing. Risk measure 1 was based arrest criterion (see exhibit 9). on the whole sample. Risk measure 2 20 was based on probationers only.8 Selection of sentences by 10 N=225 N=216 N=187 N=148 N=184 Risk measure 1 included, as predictors, judges measures of age, the judges’ ratings of Measures were devised to control 0 1 or 2 2 or 3 4 or 5 6 or 7 8 or 9 the offender’s arrest record, race, heroin for judicial selection of sentences in Score Assigned by Judge or barbiturate use in the 2 years prior to the statistically designed studies of (Two Cases With Missing Data) arrest, alcohol abuse, and the type of effects.9 These included the decision

5 R e s e a r c h i n B r i e f classifications (“predictions”) of con- Exhibit 6. Factors in an empirically derived measure of risk of any new arrests finement in 88 percent of the sample. (risk measure 1) Selection of type of sanction. Selection of a noncustodial sentence, Age (younger) jail, youth facility, or prison appeared to be influenced mainly by the judges’ Long arrest record predictions of whether the offender would commit any future crime, the Race: not white perceived importance of rehabilitation for the offender, the seriousness of the Any heroin or barbiturate use charge, the offender’s criminal record, in the past 2 years Empirically the number of counts in the convic- Number of prior probation sentences derived tion, and the recommendation of the prediction of any probation officer. The sentence type Alcohol use as problem drinker new arrests also appeared to be influenced by the (in record) age of the offender and whether the Less serious offense felony was a property crime.

Selection of split sentences. In Number of prior jail sentences ordering split sentences, the judges seemed to be influenced by their own Property crime predictions of whether the offender would commit future crimes, aggravat- Sale of drugs (current offense) ing factors, and the relative importance of retributive and rehabilitative aims.

Effects of sentences confinement were rearrested during small but statistically significant effect the 20-year followup period. When on new arrests during the first 5 years The effects of different sentences were adjustments were made for sentence after the offender was sentenced. After studied using statistical designs that selection, a priori risk, and the time adjusting for selection, a priori risk, controlled for the measures of sentence spent in the community (not incapaci- and incapacitation, those offenders selection, the a priori risk, how long the tated), however, there was no statisti- sent to the youth facility had the high- offender was confined as a result of the cally significant difference in the est adjusted percentage of new arrests. sentence, and how long the offender percentage of offenders with new subsequently had an opportunity to arrests between the two groups during When persons given noncustodial reoffend in the community (20 years that period. Similarly, there was no sanctions were excluded from the minus time served on the sentence and difference due to the type of confine- analysis, a statistically significant ef- incapacitation time for later confine- ment when only the first 5 years after fect of placement in a jail, youth facil- ments).10 Typically, a naive interpreta- sentencing were considered (see ity, or prison was found in the 20-year tion of the observed outcomes without exhibit 11). When selection, risk, followup. The adjusted values for jail consideration of the effects of selection, and incapacitation were taken into and prison were the same, but those risk, and incapacitation was compared account, confinement had no effect on offenders sent to the youth facility had with adjusted values taking into ac- the offenders’ arrest rates, the total higher adjusted values for new arrests count the potentially biasing factors. number of arrests or charges, or (see exhibit 12). charges for specific crime categories. Effects of the in-out decision. Sev- The choice among the four sentence enty percent of offenders who were not Effects of the type of sentence. alternatives had a small effect, mainly sentenced to confinement and 82 per- The type of sentence—noncustodial, due to higher adjusted values for new cent of those who were sentenced to jail, youth facility, or prison—had a arrests for those sent to a youth facility.

6 R e s e a r c h i n B r i e f

Age, selection, a priori risk, or Exhibit 8. Factors in an empirically derived measure of risk of any new arrests incapacitation could not explain this (risk measure 2) for the probation sample only effect. Placement in noncustodial pro- grams, jail, or prison had no effect. Age (younger) Effects of the time served in con- finement. Factors helping to explain Long arrest record the time actually served in jail or prison are listed in exhibit 13. After adjusting Race: not white for selection, a priori risk, and inca- pacitation, the amount of time the of- Any heroin or barbiturate use Empirically fender actually served in confinement in the past 2 years derived had a small but statistically significant prediction of any Number of prior probation sentences effect on new arrests, accounting for new arrests less than 2 percent of the differences in (offenders granted Male new arrests. A comparison of the actual probation) and adjusted values for new arrests is Incarceration for probation shown in exhibit 14. or parole violation Checks fraud or burglary Exhibit 7. A priori risk groups and the (current offense) percentage with new arrests in the Low social stability followup period (risk measure 1) (judges’ ratings)

Percentage With New Arrests Sale of drugs (current offense) (20-Year Followup)

100 95 92 87 90 Survival in the community without indicates the rates of arrests at par- arrest. The offender’s survival in the ticular times, given no arrest until that 80 community over time (after serving any time. That is, it is an arrest rate per 70 incapacitating sentences) without ar- unit of time. Analyses of survival, 70 rest did not differ between those who however, indicated that the effects of were confined and those who were not confinement in the youth facility or 60 confined. This result is reflected in in prison may change over time and 12 50 exhibit 15, which shows the “hazard should be further investigated. rates” for offenders who were sen- 40 Other effects of sentences. No 40 tenced to confinement and those who were not.11 The curve shows new ar- effect on new arrests, beyond incapaci- 30 rests from the time of release from the tation, was found for the following sen- relevant confinement (if the sentence tencing choices: the maximum sentence 20 called for incarceration) or the date of imposed, “split” sentences of jail and the sentence (if the offender was not probation, fines, restitution, and com- N=192 N=193 N=192 N=193 N=190 10 incarcerated). Whether the offender pliance with fine or restitution orders. was confined or not makes little or no 0 12345 difference to survival without arrest Implications Offender Group* when the survival or hazard function is The judges’ diverse selection purposes calculated on the basis of the a priori for sentencing support the * Offenders were divided into groups of about risk, total time incapacitated, and equal size for analysis. Data were missing for need for greater clarity and consistency two cases. length of incapacitation on the present in sentencing aims. The conflict between sentence. The hazard function shown

7 R e s e a r c h i n B r i e f

Exhibit 9. A priori risk groups and the • Except for the effect of incapacita- trol effects), little justification for differ- percentage with new arrests in the tion, whether the offender was sen- ences in sentences was found from a followup period (risk measure 2) tenced to confinement made no crime control perspective. The different for the probation sample only difference. sanctions, varying in severity of punish- ment and incapacitating effect, may • Where the offender was confined Percentage With New Arrests have served as a warning to others or as made little difference—except 120 deserved punishment, but there was perhaps for the unfavorable effect little or no of other effects on of placement in a youth facility. 98 crime control objectives. 100 • The length of the maximum sentence 91 imposed made no difference. This study found no evidence to justify the belief that the addition of jail time 79 • The length of time actually confined 80 to a probation sentence has a specific made a slight difference. 71 deterrent effect. Unless it is believed • When jail was imposed along with that jail time is required for punish- 60 probation, it made no difference. ment, or the hope of an effective warn- ing to others is maintained, this study • Fines or restitution made no would support abandoning the use of difference. 40 split sentences. Aside from general deterrence (not 27 studied) and incapacitation (in this Similarly, if considerations of general deterrence and deserved punishment 20 N=93 N=93 N=94 N=93 N=93 study it mainly provided a correction for the investigation of other crime con- are set aside, it must be concluded

0 Exhibit 10. Factors helping to explain judges’ decisions about whether to 12345 confine offenders Probationer Group*

* Probationers were divided into groups of about Age (younger) equal size for analysis. utilitarian and retributive perspectives Not a nuisance offense was apparent in this study, despite a general preference on the part of judges Serious offense for utilitarian crime control. Clarity Judges’ predictions of any could be increased if there were an in- new crime Decision to confine or not confine ternally consistent sentencing theory and Low importance of if it were consistently applied. rehabilitation in this case (jail or prison vs. noncustodial Despite their modest validity, the Low social stability sanction) judges’ subjective risk judgments sub- stantially influenced their sentencing No mitigating factors choices. The use of more formal, empirically derived methods would Any aggravating factors enhance sentencing rationality when sentencing theory incorporates risk as a ’s recommendation for confinement relevant and justifiable consideration. Probation officer’s recommendation The main sentencing choices available for confinement to these judges had little effect on crime control aims:

8 R e s e a r c h i n B r i e f

that confinement or increased length of Exhibit 11. Actual and adjusted percentage of offenders with any new arrests in the incarceration served the crime control 5 years after sentencing, by sentences to confinement or noncustodial sanctions purpose of incapacitation but had little or no effect as a “treatment” with reha- Percentage With New Arrests bilitative or specific deterrent effects. 60 Exhibit 16 illustrates that there is little 58.6 or no effect of sentences on crime control 58 objectives—other than by incapacita- 56 tion—as measured in this study. It also 55.0 55.1 portrays the large area of unexplained variability in new offenses by the per- 54 sons who were sentenced. 51.5 52 Policy implications 50 How is a study of 20-year-old sentenc- ing practices relevant to current de- 48 bates? Some important lessons for present day policy are clear. These 46 results offer little support for the policy Not Confined Confined Not Confined Confined trends, prominent since this project be- (Actual) (Actual) (Adjusted) (Adjusted) gan, that have supported increased use of confinement as a sentencing choice, emphasized longer terms, or accepted specific deterrence to reduce offenders’ Exhibit 12. Percentage of offenders with new arrests in the 20-year followup recidivism. Another lesson concerns period, by sentences to jail, youth facility, or prison specific procedures for the sentencing decision process: When the risk of new Percentage With New Arrests offenses is seen as an appropriate ele- 100 94.7 92.8 ment in sentencing policy, empirically 90 83.8 derived methods are equal or superior 81.2 81.2 76.6 to subjective assessments. 80

The societal stake in crime control 70 effects of sentences, the limitations of 60 statistical designs, and the results of this study suggest the importance of 50 investigating these and similar ques- tions by rigorous experimental meth- 40 ods whenever they are determined to 30 be ethically justifiable. 51.5 20

Study limitations 10

Some sentencing concerns were be- 0 yond the scope of the project. This Jail Youth Facility Prison study did not examine sentencing for purposes of retribution or general de- Actual Percentage Adjusted Percentage terrence—whether punishments were deserved or whether they served as an

9 R e s e a r c h i n B r i e f effective warning to others. The rela- other times. The study does not show, of sanctions beyond the length of sen- tive costs of sentencing choices, in for example, that well-designed inter- tences were available. The analyses either human or monetary terms, were ventions implemented with fidelity to a reported here could not show whether not assessed. clear, coherent theory cannot be effec- different kinds of sanctions are differ- tive in meeting crime control objec- entially effective for different kinds of Conclusions from this study cannot be tives. No measures of the quality of offenders. If either objectives or sanc- generalized to other or rehabilitative programs or the severity tions have changed, for example, to emphasize aims of restorative justice or Exhibit 13. Factors helping to explain time served in jail or prison elaborated “intermediate sanctions,” (sentences as carried out) this report must remain silent about the newer goals and alternatives.13 Judges’ ratings of the most serious offense charged Nonjudicial decisions may have af- Judges’ ratings of the fected the results. For persons sen- importance of incapacitation Time served in jail tenced to prison, the time actually or prison served in prison reflected decisions by Number of counts charged (sentences as a parole board as well as the judges. executed) Person offense Statistical designs such as those used in this study always provide less defini- Judges’ ratings of the importance of retribution tive results than do true experiments, which ensure that offenders compared after subjection to different sanctions Exhibit 14. Actual and adjusted percentage of offenders with new arrests, by may be considered equivalent in all number of months served in jail or prison on the present sentence respects at the outset. Perhaps other, as yet unknown, selection factors affect Percentage With New Arrests either judicial decisions or risk and 100 94.4 93.5 therefore influence the outcomes. 87.8 87.0 90 85.7 83.8 Other limitations are due to an often 80.2 81.4 80 used but nevertheless crude measure- 74.3 75.2 ment of outcomes. Counts of arrests 70 and charges are not adequate to assess 60 the full spectrum of costs and harms associated with either the sanctions 50 imposed or the new crimes committed in the community. 40 Similarly, a lack of information limited 30 measurement of the main independent 20 variables of the study: Adequate descriptions of community treatment 10 programs and jail and prison programs 0 were not available, nor were measures Up to 12 12–18 18–36 36–84 More Than of the quality of rehabilitative programs Months Months Months Months 84 Months and the severity of sanctions within the Number of Months Served in Jail or Prison main types of sentences imposed.

Actual Percentage Adjusted Percentage Related research results ameliorate some of these limitations. The case of

10 R e s e a r c h i n B r i e f

the judges’ predictions provides an ex- Exhibit 15. Hazard of arrest after serving required sentence, confined and ample. In the past 70 years, 136 studies not confined* have documented 617 distinct compari- sons of the two basic measures of predic- 1.0 Function at Mean of Covariates tion (subjective or “in the head” versus more formal or statistical measures). 1.4 Some of these studies were conducted in 1.2 the arena. In criminal justice decisions and in all other areas 1.0 of behavioral prediction, the evidence clearly shows that more formal objective .8 measures are superior to subjective evaluations.14 Similarly, related research .6 since the 1960s does not support the ex- Cumulative Hazard pectation that the length of time some- .4 one is incarcerated in prison is related to repeated offending, particularly when .2 relevant offender characteristics are 0.0 taken into account.15 -10 -5 0 5 10 15 20 25 30 Years to Next Arrest After Time Served Notes Confined Not Confined 1. This study started more than 20 years ago in collaboration with the judges of the Essex * The covariates were the a priori risk, incapacitation, and time served. County, New Jersey, Court, particularly with Judge John A. Marzulli, then the assignment Office of the , and John McCarthy, 4. Gottfredson, S.D., and R.B. Taylor, judge. The support of Robert D. Lipscher, then deputy director for criminal practice for the “Community Context and Criminal Offenders,” director of the New Jersey Administrative New Jersey Administrative Office of the Courts, in Communities and Crime Prevention, ed. T. was essential. The followup phase of the project Hope and M. Shaw, London: Her Majesty’s described in this Research in Brief relied on Stationery Office, 1988; and Gottfredson, S.D., Exhibit 16. Explained and unexplained the expertise of an advisory committee com- and D.M. Gottfredson, “Behavioral Prediction variances in any new arrests during posed of McCarthy; Stan Repko, deputy com- and the Problem of Incapacitation,” Criminol- the 20 years after sentencing* missioner for research and planning, New ogy 32 (3) (1994): 450–451. Jersey Department of Corrections; and Wayne Fisher, deputy director, New Jersey Division of 5. “Nuisance” offenses include, for example, Selection Criminal Justice. The full report of the study, parole and probation rules violations, posses- 6% including references, is available from the Na- sion or use of drugs, marijuana offenses, disor- tional Criminal Justice Reference Service. See derly conduct, prostitution, and gambling. Unexplained A priori risk Gottfredson, Don M., Choosing Punishments: 75% 8% Crime Control Effects of Sentences, report to the 6. The items listed are from regression analysis National Institute of Justice, Sacramento, CA: results. In combination, these items resulted in Justice Policy Research Corporation, 1998. a multiple correlation of 0.79 (N=931) with the judges’ ratings, which were scores of 1 to 9. The 2. Resources did not permit a search for death weights of each item (i.e., the standardized re- Incapacitation records, and the total number of deaths during the gression coefficients) are not reflected in the (due to sentence followup period is unknown. Twelve persons are exhibit but are given in the full report cited. or as a result known to have died during the followup. Two had of later crimes) received noncustodial sentences; four received jail 7. The point biserial correlation of these scores 11% sentences; four were sent to a youth facility; and with any new arrests was 0.22 (N=960). two went to prison. Before their deaths, none of * Neither variation in the time served on the 8. The items listed are from regression analyses. these offenders had been arrested after sentencing. present sentence nor the type of sentence The point biserial correlation of risk measure 1 (noncustodial, jail, youth facility, or prison) ex- 3. Typically, these were analyses of covariance scores with any new arrest was 0.48 (N=960); for plained differences in outcomes beyond their with the measures to be controlled for as risk measure 2, it was 0.56 (N=466). (For standard- incapacitative effects. covariates. Analyses of “survival” (the time that ized coefficients, see the full report.) The results of had elapsed when arrest occurred) also were used. logistic regression analyses, theoretically more

11 appropriate for the dichotomous dependent vari- model, and there was a significant effect from Gottfredson, D.M., M. Neithercutt, J. Nuffield, able, did not differ substantively. This is not atypi- a time x treatment interaction. Further analysis and V. O’Leary, Four Thousand Lifetimes: A cal; see, for reviews and reasons, Copas, J.B., and indicated the hazard rate increased more rapidly Study of Time Served in Prison and Parole Out- R. Tarling, “Some Methodological Issues in Mak- over time for those sentenced to the youth insti- comes, Davis, CA: National Council on Crime ing Predictions,” in Criminal Careers and “Career tution, less so for those imprisoned. Survival in and Delinquency, 1973; Gottfredson, D.M., Criminals,” ed. A. Blumstein, J. Cohen, J. Roth, the community without arrest after release from M.R. Gottfredson, and J. Garafalo, “Time and C. Visher, Washington, DC: National Acad- confinement in the youth facility decreased more Served in Prison and Parole Outcomes Among emy of Sciences, 1986; Farrington, D., and R. rapidly than did survival without arrest for per- Parolee Risk Categories,” Journal of Criminal Tarling, Predicting Crime and Delinquency, Al- sons who had been imprisoned. Justice 5 (1) (1977). bany, NY: State University of New York Press, 1985; Gottfredson, D.M., and M.R. Gottfredson, 13. For recent reviews, see Cullen, F.T., J.P. “Data for Criminal Justice Evaluation: Some Re- Wright, and B.K. Applegate, “Control in the sources and Pitfalls,” in Handbook of Criminal Jus- Community: The Limits of Reform?” in Choos- Don M. Gottfredson is Richard J. tice Evaluation, ed. M.W. Klein and D.S. Teilman, ing Correctional Options That Work, ed. A. Hughes Professor Emeritus at Harland, Thousand Oaks, CA: Sage Publica- Beverly Hills, CA: Sage Publications, 1980. In the School of Criminal Justice, both the ordinary least squares regression and the tions, 1996: 113–115; Petersilia, J., “Under- logistic regression, the items most helpful in pre- standing Probation and Parole: Theory and Rutgers University, and President diction, in the context of the other variables in- Practice,” in Oxford Criminology Handbook, of the Justice Policy Research cluded, were age, arrest record, race, and heroin or ed. M. Tonry, London: Oxford University Press, Corporation. This study was barbiturate use in the past 2 years. The race vari- 1998; Petersilia, J., Community Corrections: funded in part by grants 82–IJ– able (“white”) was included and retained because Probation, Parole, and Intermediate Sanctions, the prediction method was not developed for any London: Oxford University Press, 1998. CX–0054 and 95–IJ–CX–0118 by operational classification use but only for the pur- the National Institute of Justice. pose of statistical control in the other analyses de- 14. Meehl, P.E., and W.M. Grove, “Comparative Efficiency of Informal (Subjective, Impressionis- scribed. More complete specifications of these and Findings and conclusions of the research re- other analytical results discussed here are included tic) and Formal (Mechanical, Algorithmic) Prediction Procedures: The Clinical-Statistical ported here are those of the author and do not in the full report of the study. necessarily reflect the views or endorsement of Controversy,” Psychology, Public Policy, and the Essex County Court, the judges who partici- 9. Selection measures relied on discriminant Law 2 (2) (1996): 293–323. pated in the study, the New Jersey Administra- functions with n-1 equations where n=the tive Office of the Courts, the U.S. Department of number of groups to be discriminated. The 15. Glaser, D., The Effectiveness of a Prison Justice, or any other individuals or agencies. equations, including the standardized coeffi- and Parole System, New York: Bobbs-Merrill, cients, are given in the full report of the study. 1964: 302–303; Department of Institutions, State of Washington, Research Monograph No. The National Institute of Justice is a component of the Office of Justice 10. Most analyses relied on analyses of 27, Olympia, WA: Department of Institutions, Programs, which also includes the Bureau covariance, others on regression methods. The State of Washington, 1967; Department of Cor- of Justice Assistance, the Bureau of Justice “adjusted values” shown in the exhibits are the rections, State of California, Long Jail Terms Statistics, the Office of Juvenile Justice and adjusted means from the analyses of covariance. and Parole Outcomes, Sacramento, CA: Depart- ment of Corrections, State of California, 1967; Delinquency Prevention, and the Office for 11. This result is based on a Cox regression Assembly Committee on , Victims of Crime. analysis. (Committee Report). Sacramento, CA: Assem- bly of the State of California, 1968; Kolodny, This and other NIJ publications can be 12. A Cox regression analysis based on the S., Parole Board Reform in California: Order found at and downloaded from the NIJ groups given noncustodial, jail, youth facility, Out of Chaos, Report of the Select Committee Web site (http://www.ojp.usdoj.gov/nij). and prison sentences did not meet assumptions on the Administration of Justice, Sacramento, of proportionality required for the use of the CA: Assembly of the State of California, 1970; NCJ 178889

U.S. Department of Justice PRESORTED STANDARD Office of Justice Programs POSTAGE & FEES PAID National Institute of Justice DOJ/NIJ PERMIT NO. G–91 Washington, DC 20531

Official Business Penalty for Private Use $300