Berkeley Law Berkeley Law Scholarship Repository Faculty Scholarship 11-1992 Between Two Extremes: An Examination of the Efficiency and Effectiveness of Community Service Orders and Their Implications for the U.S. Sentencing Guidelines Malcolm M. Feeley Berkeley Law Richard Berk Alec Campbell Follow this and additional works at: http://scholarship.law.berkeley.edu/facpubs Part of the Law Commons Recommended Citation Malcolm M. Feeley, Richard Berk, and Alec Campbell, Between Two Extremes: An Examination of the Efficiency and Effectiveness of Community Service Orders and Their Implications for the U.S. Sentencing Guidelines, 66 S. Cal. L. Rev. 155 (1992), Available at: http://scholarship.law.berkeley.edu/facpubs/1332 This Article is brought to you for free and open access by Berkeley Law Scholarship Repository. It has been accepted for inclusion in Faculty Scholarship by an authorized administrator of Berkeley Law Scholarship Repository. For more information, please contact jcera@law.berkeley.edu. BETWEEN TWO EXTREMES: AN EXAMINATION OF THE EFFICIENCY AND EFFECTIVENESS OF COMMUNITY SERVICE ORDERS AND THEIR IMPLICATIONS FOR THE U.S. SENTENCING GUIDELINES MALCOLM M. FEELEY,* RICHARD BERK** & ALEC I. CAMPBELL*** INTRODUCTION American penal policy has for years been forced to choose between two extremes, the harshness of imprisonment and the leniency of proba- tion. A vast gulf separates these two forms of punishment, with little in between. In recent years frustration with these extremes has led many to search for something to fill the vacuum-intermediate punishments that would contribute to a greater range of sanctions and facilitate a more rational sentencing system.' One such alternative is community * Professor of Law, Boalt; Visiting Professor, Hebrew University. B.A. 1964, Austin College; Ph.D. 1969, University of Minnesota. Research for this study was supported by gifts to the Center for the Study of Law & Society through the auspices of the U.S. Parole & Probation Service and the Daniel & Florence Guggenheim Foundation. The authors appreciate this help and the comments offered by a number of people, most especially Loren A.N. Buddress, Chief Parole & Probation Officer for the U.S. District Court for the Northern District of California; the Hon. William W Schwarzer, Judge of the U.S. District Court for the Northern District and currently Director of the Federal Judicial Center; and Sheldon Messinger, Professor of Law at the University of California (Berkeley). We also appreciate the research assistance of Charles Lester, Sean Gallagher, and Terri Winter. Finally, we are indebted to Donald Rubin for suggestions on our statistical procedures and to Bob Kagan for comments on an earlier draft. The authors are solely responsible for the contents and conclusions in this study. ** Professor of Sociology & Statistics, University of California (Los Angeles). B.A. 1964, Yale; Ph.D. 1970, the Johns Hopkins University. *** Doctoral Candidate in Sociology, University of California (Los Angeles). B.A. 1988, Columbia University; M.A. 1991, University of California (Los Angeles). 1. See, ag., NORVAL MORRIS & MICHAEL TONRY, BETWEEN PRISON AND PROBATION: INTERMEDIATE PUNISHMENTS IN A RATIONAL SENTENCING SYSTEM (1990). HeinOnline -- 66 S. Cal. L. Rev. 155 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 service-the imposition of an obligation to work for the community, either in lieu of imprisonment or as an additional condition of probation. Putting criminal offenders to work is not, of course, a new idea. It has a long history predating the founding of the United States. And the rise of the convict lease system and chain gangs in the post-Civil War South ushered in an especially inglorious history of American penal policy.2 However, contemporary community service orders (CSOs) differ substantially from these earlier forms of penal servitude. Apart from differences in the conditions under which work is performed, this new form of labor as punishment is not coerced, nor is it even performed under the supervision of correctional officials. Typically, under CSOs, offenders are required as a condition of probation to perform a specified number of hours of work for a community service agency. The offenders usually undertake this work while living at home and schedule it on a part-time basis around existing employment, school, and family commitments. Although judges have imposed such sentences on an ad hoc basis for decades, it was not until the 1960s and '70s-first in England and Canada and then in the United States-that community service sentences began to be institutionalized as formal sentencing options and viewed as alternatives to incarceration.3 By the early 1990s, CSOs were statutorily authorized options in many, if not most, American jurisdictions. And more recently Norval Morris and Michael Tonry have mounted a vigorous defense of community service as a significant, new, and potentially widely used form of punishment.4 Despite this growth in both numbers and popularity, there is little research that systematically examines the administration and impact of CSO programs.5 Although a more substantial body of research on CSOs exists in England and Canada, it tends to be anecdotal and unsystematic. Systematic research on American programs is even more sparse. Most of it falls into one of two categories: either it uncritically touts community service sentences as the answer to the failure of traditional rehabilitation 2. See, eg., BLAKE McKELVEY, AMERICAN PRISONS: A HISTORY OF GOOD INTENTIONS (1977). 3. See, eg., MORRIS & TONRY, supra note 1, at 156; A.A. VASS, THE PRISON CRISIS AND ALTERNATIVES TO CUSTODY (1990); WARREN A. YOUNG, COMMUNITY SERVICE ORDERS: THE DEVELOPMENT AND USE OF A NEW PENAL MEASURE (1979); Ken Pease, Community Service Orders, in 6 CRIME AND JUSTICE: AN ANNUAL REVIEW OF RESEARCH (1985). 4. MORRIs & TONRY, supra note 1, at 160-74. 5. VASS, supra note 3, at 168. HeinOnline -- 66 S. Cal. L. Rev. 156 1992-1993 1992] BETWEEN TWO EXTREMES programs and the problems of prison crowding or it warns that alternative sentencing schemes unwittingly widen the net of social control.' This study takes a different tack. It examines this new form of sentencing in light of three enduring issues that are raised in all sentencing schemes: (1) the relative severity of the sanctions, (2) the problem of implementation, and (3) the question of deterrence. We have formulated these concerns as hypotheses that we have investigated in the context of one well-established CSO program, the one administered by the U.S. District Court for the Northern District of California. Although these issues are obvious, of general significance, and have been raised many times before in the context of various types of sentence reforms,' few if any systematic studies address them as they apply to CSOs.8 Obviously, a single study of a single program cannot provide definitive answers to such questions, but it can clarify issues, expose unfounded assumptions, and perhaps demonstrate that certain things can be done. We begin by spelling out more fully each of our three central concerns: 1. The problem of sentence severity is concerned with an enduring question in the administration of criminal justice, namely, what constitutes an appropriate punishment? More particularly, we seek to shed empirical light on an important and continuing debate about the efficacy of "alternative" punishments. Conventional wisdom holds that the feasible "outer bound" of the length of community service is 240 hours-six weeks of full-time employment?-an assertion which, if accurate, seriously curtails the likelihood that CSOs can be considered significant enough punishment to be seen as viable alternatives for anything other than the shortest of jail terms. Our study challenges this conventional wisdom about outer bounds and shows that it is possible to impose substantially longer terms of community service than is generally thought. 2. The problem of implementation has emerged in recent years as a major-perhaps the central-issue in policy analysis. The literature in 6. James Austin & Barry Krisberg, The Unmet Promise of Alternatives to Incarceration, 28 CRIME & DELINQ. 374, 380 (1982). 7. See, e.g., MALCOLM M. FEELEY & DEBORAH KELLEY, REPORT ON CLIENT SPECIFIC PLANNING: A PROJECT OF THE NATIONAL CENTER ON INSTITUTIONS (1983); DOUGLAS C. MCDONALD, PUNISHMENT WITHOUT WALLS (1986); AND ALTERNATIVES MORRIS & TONRY, supra note 1. 8. It is worth noting that the two most comprehensive surveys of research on community service sentences directed at American audiences focus almost entirely on programs in Great Britain, Canada, and Continental Europe. See, eg., MORRIS & TONRY, supra note I; Pease, supra note 3. 9. MORRIS & TONRY, supra note 1, at 173. HeinOnline -- 66 S. Cal. L. Rev. 157 1992-1993 SOUTHERN CALIFORNIA LAIW REVIEW [Vol. 66:155 this field suggests that policies are most likely to be implemented when the distance between policy articulation and policy administration is short and the process of translating policy into action is tightly controlled.10 It hypothesizes that programs that depend on weak, decentralized, or loosely coupled structures for implementation are not likely to succeed.1 1 Yet we have found that at least one such program works tolerably well. This observation is not earth-shattering, but in the face of much skepticism about CSOs and the administration of decentralized programs and the widespread belief that "nothing works" in criminal justice reforms, 2 the counterexample-the instance of a single successful program-is significant. We examine reasons why this program seems to be distinctive, but of course it remains to be seen whether other programs will work on a regular basis and in a variety of settings. 3. The problem of deterrence and impact involves one of the classic questions about criminal sanctions, namely, what is the relative deterrent value that this form of punishment has on the individual offender? Our response is to compare the impact of CSOs with the impact of other forms of punishment, such as straight probation and imprisonment. By approximating a random-assignment research design, we compare the recidivism rates of those offenders receiving CSOs to those receiving these other forms of punishment. We conclude that CSO sentences compare favorably-as indicated by recidivism rates-with these other types of sentences. The database for this study is the CSO program in one federal district court, the U.S. District Court for the Northern District of California. Since its establishment in the 1970s, this court has imposed community service on selected federal offenders in lieu of sentences of straight probation or incarceration. The court expanded this practice in light of the Comprehensive Crime Control Act of 1984,13 which explicitly authorized community service sentencing. We briefly trace the history of this program, examine the structure under which it operates, and then turn to the three issues posed above. Part I briefly introduces the CSO program in the federal courts and describes the process of sentencing offenders to community service in the Northern District of California. The data for this discussion come from 10. JEFFREY PRESSMAN & AARON WILDAVSKY, IMPLEMENTATION (1977). See, ag., id.; MALCOLM FEELEY & AuSTIN SARAT, THE POLICY DILEMMA: FEDERAL CRIME POLICY AND THE LAW ENFORCEMENT ASSISTANCE ADMINISTRATION (1980). 11. 12. 13. Robert Martinson, What Works?, PUB. INTEREST, Spring 1974, at 22-54. Pub. L. No. 98-473, tit. II, 98 Stat. 1837, 1976 (1984). HeinOnline -- 66 S. Cal. L. Rev. 158 1992-1993 1992] BETWEEN TWO EXTREMES observation and interviews undertaken during the summer and fall of 1989. During this period we observed sentencing practices and interviewed five of the court's twenty-three judges, read presentence reports filed by all eight of the court's probation officers, and interviewed the probation officers about their views on CSOs. Part II examines the first of the three issues outlined above, the problem of sentence severity, and compares CSO sentences in the Northern District of California with those elsewhere. In light of this discussion of the nature of community service sanctions, we ask whether the court's orders impose conditions that can be regarded as significant punitive sanctions and hence serve as alternatives to incarceration. If we do not answer this question, we do at least hope that we have advanced discussion of it. Part III assesses the problem of implementation. Discussion here is based upon in-person and telephone interviews with community-based supervisors of 129 CSO participants and visits to ten participating agencies. Part IV explores the issue of deterrence and impact by employing a propensity score analysis that approximates a random-assignment research design to test the effects of treatments. Here we compare the recidivism rates of those offenders receiving CSO sentences to those receiving straight probation and those receiving CSOs and periods of incarceration. Part V explores the policy implications of our findings. In particular we explore the consequences of adjusting the U.S. sentencing guidelines in order to create a presumptive sentence of community service. We find that with only marginal adjustment, a substantial number of offenders could be redirected away from sentences involving short periods of imprisonment toward community service punishments. The final part summarizes the central findings of this study and explores the policy implications for intermediate punishments. II. THE PROCESS OF COMMUNITY SERVICE SENTENCING IN THE FEDERAL COURTS A. BACKGROUND AND STRUCTURE In the 1960s federal judges began experimenting with community service as an alternative to straight probation and incarceration, and by the late 1970s probation dep;artments in seventy-one of the ninety-four federal judicial districts had d~veloped community service programs. By the mid-1980s nearly nine percent of all federal offenders on probation HeinOnline -- 66 S. Cal. L. Rev. 159 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 were involved in community service as a condition of their probation.14 This growing interest in community service in the federal system was reinforced by title 18, section 3563 of the Comprehensive Crime Control Act of 1984,15 which made express provision for community service as a 16 condition of probation for selected groups of felony offenders. Although the implementation of the federal sentencing guidelines in 1987 severely restricted the sentencing discretion of judges, judges did retain discretion over sentences of offenders who were convicted of less serious offenses and who had relatively clean criminal histories. Since this group constituted the pool from which those sentenced to community service had come, the adoption of the guidelines did not significantly curtail the existing community service sentencing practices of federal judges. The guidelines did, however, severely limit the ability of federal judges to expand the scope of this sentencing option, and this remains one of the judiciary's criticisms of the guidelines. 17 Still, within these limits, judges in the Northern District of California have continued to be aggressive in experimenting with CSOs and have remained adamant in their criticism of the guidelines. Since the mid-1980s a substantial portion of all sentences have involved community service, either as a condition of probation or as part of a split sentence involving community service following a period of incarceration." Like all sentencing options, CSOs reflect a host of different interests and diverse objectives. 1 9 Some view them as facilitating rehabilitation. 14. JUDICIAL CONFERENCE COMM. ON THE ADMIN. OF THE PROBATION SYS. (SUBCOMM. ON COMMUNITY SERv.), COMMUNITY SERVICE: A GUIDE FOR SENTENCING AND IMPLEMENTATION 187 (1986). 15. Pub. L. No. 98-473, tit. II, 98 Stat. 1837, 1976 (1984). 16. 18 U.S.C. § 3563(b)(13) (1988). 17. Title II of the Comprehensive Crime Control Act of 1984 mandates that for offenses punishable by more than one year in prison, the offender, if sentenced to probation, must also be ordered to pay a fine, make restitution, perform community service, or perform some combination of these actions. This has led to an increase in community service orders throughout the federal system, including the Northern District of California. 18. Although the U.S. sentencing guidelines have had some effect on the administration of CSOs, the impact in the Northern District of California--as throughout the federal system-has been minimal. As stringent as they are, the guidelines allow for considerable discretion at the lenient end of the continuum, and offenders within this discretionary zone make up the bulk of CSO participants. Although the guidelines set a (low) upper limit on who can receive a CSO sentence in lieu of prison, as a practical matter the guidelines have not significantly affected this option, because even before their introduction the overwhelming majority of offenders who received CSOs came from this less serious group. However, in our conclusion we make an argument that the guidelines should be altered to allow more serious offenders to receive community service sentences. 19. See, eg., MORRIS & ToNRY, supra note 1; Frederick G. Allen & Harvey Treger, Community Service Ordersin FederalProbation:Perceptionsof ProbationersandHost Agencies, FED. PROBATION, Sept. 1990, at 8; Pease, supra note 3. HeinOnline -- 66 S. Cal. L. Rev. 160 1992-1993 1992] BETWEEN TWO EXTREMES Others see them as providing symbolic restitution or reparation, returning something to the victimized community and making it whole again. Still others view CSOs as inexpensive intermediate sanctions, locating them somewhere between incarceration and probation. All of these views are represented among judges and probation officers in the Northern District of California, and for one reason or another all the judges have embraced the idea of community service sentencing and have imposed CSOs regularly. Furthermore, there is a consensus among both judges and probation officers in the Northern District of California that CSOs are a useful sentencing option.20 This view is held by other federal judges and probations officers as well.2 1 Community service sentences in the Northern District of California are imposed in much the same way that other sentences are imposed; they are usually recommended by probation officers in their presentence reports. Although the introduction of the federal sentencing guidelines in November 1987 has somewhat limited who can be sentenced to community service, the guidelines did not significantly affect existing CSO sentencing practices. Perhaps the biggest difference is that under the guidelines courts are prohibited from issuing community service orders in lieu of custodial sentences. However, it is not clear that community service was an explicit alternative to a custodial sentence prior to the guidelines, and, even after the introduction of the guidelines, judges continue to possess sentencing discretion in borderline cases. In recent years most probation officers have come to recommend community service on a nearly routine basis, largely because it is promoted by the Crime Control Act of 1984. In interviews all eight probation officers in the Northern District of California indicated that they made frequent use of the option. All of them recommended community service in at least half the probation cases, and some indicated that they recommended it "nearly all the time."'22 They expressed a consensus about the appropriate use of CSOs, agreeing that CSOs should be used only when offenders have no records of violent or unstable behavior. Other factors about which there was general agreement, though not a consensus, were (1) whether the offender was able to pay a fine (those unable to pay might receive a recommendation for community service), (2) the type of offenses, and (3) the type of offender. Other, less 20. Memorandum from Charles Lester, Researcher, to Authors, Community Service Orders 1 (June 27, 1988) (on file with authors). 21. See, eg., Allen & Treger, supra note 19. 22. Lester, supra note 20, at 3. HeinOnline -- 66 S. Cal. L. Rev. 161 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 frequently mentioned considerations were (1) the nature of the offense in relation to the community (offenders who "take advantage of the community" should "give something back"), (2) the skills of the offender ("Do they have skills likely to be of use to an agency?"), and (3) the offender's sense of responsibility (one probation officer reported that he reserved community service recommendations for offenders he believed 23 were "responsible enough to complete the service"). Probation officers only recommend; judges sentence. However, seven of the eight probation officers in the Northern District of California agreed that, "for the most part," judges follow their community service recommendations at least seventy-five percent or more of the time. This view was supported by interviews with five judges and confirmed by a comparison of sentence recommendations in several dozen presentence reports with actual sentences imposed by judges. As one probation officer put it, although judges and probation officers might weigh factors differently, "we all play in the same ballpark."'2 4 He was emphasizing that even when there are differences, they are not likely to be great. For instance, in our sample, judges completely accepted the probation officers' recommendations two thirds of the time. When there were differences it almost always had to do with the number of hours to be served and not whether community service should be imposed. This high rate of interagreement may be a consequence of a high correlation of independent decisions, an ability of probation officers to anticipate what judges will do, or a willingness of judges to follow recommendations. However, there appears to be an informal "common law" of sentencing that is understood and shared by judges and probation officers. B. COMMUNITY SERVICE IN THE NORTHERN DISTRICT OF CALIFORNIA: PARTICIPANTS AND STRUCTURE Since the outset of the Northern District's formal community service program in the mid-1980s, nearly 500 offenders have been sentenced to perform community service. Table 1 gives descriptive statistics for the population of offenders placed on probation with community service from 1984 to 1988 in the Northern District of California. Several features stand out. Almost none (seven percent) had been sentenced to jail in the past. Only four percent were involved in offenses in which weapons were used. Twenty-seven percent were African-American; eleven percent were Asian; three percent were Latino, less than one percent 23. 24. Id. Id. HeinOnline -- 66 S. Cal. L. Rev. 162 1992-1993 1992] BETWEEN TWO EXTREMES TABLE 1 Descriptive Statisticsfor Offenders Placed on Probationwith Community Service (1984-1988), Northern Districtof California(N = 477) VARIABLES No. of Defendants Dollar Amount in Crime No. of Juvenile Convictions No. of Adult Convictions No. of Times Incarcerated Less than One Year No. of Times Incarcerated One to Five Years No. of Times Incarcerated More than Five Years No. of Priors No. of Prior Revocations Months Employed Monthly Income (Dollars) Male Black White Hispanic American Indian Asian One Act in Offense Acted Alone No. CJ Status This Offense No. CJ Status Prior Offense Crime Was a Felony Age Crime Was a Drug Offense College or More Education Lives w/Another Person History of Drug Abuse Needs Employment Alcohol Dependent Needs Shelter Illiterate Vocational Training Physically Infirm Needs Psychological Help Needs Financial Help No Weapon Involved in Offense Never Been to Jail MEAN STD DEV 1.55 142,373.83 0.04 0.60 1.61 2,339,281.81 0.27 1.39 0.09 0.48 0.00 0.03 0.27 0.00 0.01 0.09 0.14 8.84 1220.27 0.64 0.27 0.03 0.00 0.11 0.45 0.64 0.92 0.91 0.47 37.09 0.17 0.49 0.41 0.17 0.22 0.08 0.00 0.03 0.09 0.03 0.16 0.15 0.96 0.93 0.10 0.40 0.90 4.65 1646.85 0.48 0.44 0.16 0.09 0.32 0.50 0.48 0.27 0.28 0.50 10.31 0.38 0.50 0.49 0.38 0.41 0.27 0.06 0.18 0.29 0.16 0.36 0.36 0.20 0.25 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 20.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 MINIMUM MAXIMUM 17.00 50,000,000.00 3.00 9.00 2.00 5.00 9.00 12.00 18,000.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 73.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 were Native American. The rest (forty-two percent) were non-Hispanic whites. Roughly two thirds (sixty-four percent) were male. One half (forty-nine percent) had at least some college education. The mean number of prior convictions was 0.09, indicating that the vast majority had no priors at all. The mean monthly income was $1220, which translates into a yearly income of $14,640. Seventeen percent had a history of HeinOnline -- 66 S. Cal. L. Rev. 163 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 drug abuse and eight percent were alcohol dependent. The mean age of offenders was 37. The typical offender is difficult to define by such demographic characteristics as age, race, and sex, because there is considerable variation on these characteristics. However, the characteristics that do emerge are that these offenders have no or only minor prior criminal records, virtually no prior jail time, and are unlikely to have committed crimes of violence. Nearly 500 offenders have been sentenced to community service in the Northern District of California since 1984. For several reasons this piece of our study is based on only a portion of them. Work records of participants who completed their sentences early in the program were not available because records on work schedules are discarded shortly after probation is completed. Thus, information on most of the early cases in this program was not available. Other cases were winnowed out for different reasons. Fourteen offenders who were initially sentenced to community service subsequently opted to pay fines and were successful in requesting that their community service order be voided. Another group had been sentenced too recently to have completed a substantial portion of their service (we set fifty percent as the cutoff point), so they were dropped from our study. Thus, eighty-one cases of community placement either had not yet been selected at the time of our study or the file lacked sufficient information to be able to trace the community service placement. Another 144 cases were excluded because we were unable to complete interviews with key supervisors. The single largest reason was the lack of record keeping by placement agencies. Although we were always able to locate the agency with which the offender had been placed and identify the participant's supervisor there, the agency often did not have records of work histories of people who had completed their assignments, supervisors could not always recall particular persons, or the supervisor was no longer working at the agency. For these and similar reasons, these cases were dropped from our analysis. Thus, the first part of this report is based upon an analysis of the remaining cases. Our primary database consists of responses to interviews of the CSO supervisors in these 129 cases. Despite this high dropoff rate, we believe that these responses are not skewed and that the HeinOnline -- 66 S. Cal. L. Rev. 164 1992-1993 BETWEEN TWO EXTREMES 1992] information on these participants represents fairly well the experience of community service placements in the Northern District of California.2 5 Because of dismantled records, we had to focus on the most recent cases. One consequence was that many of the participants were still serving their sentences. However, of this group, most (sixty-four percent) had completed all of their work. Thus, we are confident that our assessment rests upon a solid foundation of experience. C. ARRANGING FOR COMMUNITY SERVICE At the time a community service order has been recommended by a probation officer and imposed by a judge, the details of placement have rarely been worked out. Only after sentence is imposed is a community placement arranged, usually through one of three ways: (1) a probation officer directs the offender to one of several volunteer "clearinghouses," which in turn locate a suitable job, (2) a probation officer negotiates directly with a community agency that uses volunteers and then sends the offender directly to it, or (3) an offender locates his or her own placement and then seeks to have it approved by a probation officer. All three forms of placement were used with some frequency by the offenders in our study. Thirty-nine percent obtained placements through a clearinghouse, twenty-eight percent located their own placement, and in the balance of cases-thirty-three percent-probation officers worked with offenders and other officials to locate placements, usually in community service programs run by county courts. Each of these placements is different. Clearinghouses facilitate volunteer services for a host of people, not just offenders sentenced to community service by the federal courts. Indeed, one of their primary jobs is to locate community service placements for offenders sentenced by state courts, and another is to recruit volunteers from a variety of sources and match them with agencies in need of volunteers. 26 Offender-initiated 25. Initially, we were concerned that this drop-off would bias our sample, perhaps leading us to oversample the more reliable organizations and offenders. However, after discussions with staff at a number of agencies, we do not believe this is the case. Almost all the agencies are underfinanced and understaffed and experience high staff turnover. Given the nature of these organizations, this is not surprising. This does not necessarily reflect on either the quality of supervision or the work performance of community service workers. Indeed, even within the same organization, depending on staff turnover and how recently the community service order had been completed, we could obtain assessments on some participants but not others. 26. Our data show that those clearinghouses the court in the Northern District of California has come to rely on most heavily include Volunteer Centers of Alameda County (and its component units, the Community Service Alternatives programs in Oakland and Hayward), the Volunteer Center of Contra Costa County, the Volunteer Center of Sonoma County, the Sentencing AlternaHeinOnline -- 66 S. Cal. L. Rev. 165 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 placements are often located as a result of prior association with a community organization, from lists of agencies provided by the Office of Probation, or as follow-ups on suggestions given by probation officers. Offenders have an incentive to locate their own placement to secure a convenient location and the like. III. IS COMMUNITY SERVICE A SIGNIFICANT SANCTION? In a recent book, Between Prison and Probation, Norval Morris and Michael Tonry have argued that community service is a viable alternative to imprisonment.2 7 In putting forth their argument, they also embrace a widespread belief that there is an upper limit, or "outer bound," to the number of hours offenders can reasonably be expected to spend performing community service.28 Drawing on figures put forward by others, they argue that this outer limit is 240 hours, preferably served in short and compact time periods so as to facilitate effective administration.29 They arrive at this figure by considering two sets of factors, one a practical administrative concern, the other a more formal consideration drawn from laws and court rulings in several countries. The practical issue has to do with administrative concerns over compacting service into a short period of time so as to facilitate effective administration of the program. This certainly has strong intuitive appeal, and there is little doubt that the longer service is stretched outeither by part-time work or in the total hours to be served-problems of nonappearance, supervision, and administration increase. Yet the figure does not appear to rest upon any substantial empirical base. Rather, it appears to be based upon speculation and practices in a handful of British and European programs that impose service for periods of not more than 240 hours.30 In presenting this figure, Morris and Tonry also draw on the experience of the Vera Institute's community service program in New York City, which, in order to minimize administrative problems, self-consciously limited work to seventy hours served over a concentrated, two-week period.3 1 The second component of their argument for an outer limit of 240 hours rests upon frequently used formulas for finding the equivalents tives Program in San Jose, Sunnyvale Volunteer Center, Community Options in Santa Cruz and Watsonville, Project 20 in San Francisco, and the San Mateo County Court Work Program. 27. MORRIS & TONRY, supra note 1. 28. 29. 30. 31. Id. Id. Id. Id. at at at at 173. 168-69. 153. 155. HeinOnline -- 66 S. Cal. L. Rev. 166 1992-1993 1992] BETWEEN TWO EXTREMES between imprisonment and community service. Drawing on the Vera Institute's program, Morris and Tonry accept the assertion that seventy hours of community service is the equivalent of one month's jail time.32 Similarly, they cite approvingly rulings by the Court of Appeal in England, which have held that 190 hours of service is the equivalent of one year's imprisonment.3 3 In addition, they accept statutory presumptions in Germany regarding community service, which equate twenty days of community service with six months in jail.34 In light of these and other considerations, Morris and Tonry conclude that 240 hours is the outer limit for community service and that this limit ought to be treated as the equivalent of a six- to twelve-month jail sentence.35 We are skeptical of such assertions. First, they rest at least to some extent upon untested empirical foundations. Second, in our opinion, such liberal translations of work for time served and such low upper limits are likely to reinforce the idea that community service cannot be-at least not in the American context-a significant punitive alternative to incarceration. Our concerns have some basis in experience. Existing studies of English programs, where community service is generally restricted to 200 hours or less, suggest that the vast majority of offenders receiving CSOs would not otherwise have received prison terms.36 Estimates by informed American observers suggest the same thing. They warn that the primary effect of such programs may be to "widen the net" of social control rather than to limit reliance on incarceration. 37 And the most highly publicized and carefully evaluated American community service program to date, the Vera Institute's program in New York City, was specifically aimed at petty (although chronic) offenders, whose offenses normally were not likely to warrant jail time. 31 In short, although Morris and Tonry's argument is intuitively appealing-surely shorter, more compact periods of time are more easily supervised and completed than longer, more leisurely paced terms-it does not rest on a foundation of experience. Furthermore, the outer limit of only 240 hours is low 32. Id. 33. Id.at 173. 34. Id 35. Id 36. Fg., Pease, supra note 3, at 79. 37. Eg., Austin & Krisberg, supra note 6; Thomas Blomberg & Karol Lucken, Intermediate Punishment and Social Control: A Preliminary Assessment 20 (unpublished manuscript, on file with authors). 38. MCDONALD, supra note 7, at 81. HeinOnline -- 66 S. Cal. L. Rev. 167 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 enough to lead some to question whether community service is or can be a serious and viable alternative to imprisonment. Although it is beyond the scope of this study to determine what constitutes "appropriate" penalties for given offenses or to devise a system of equivalency between community service and time served in custody, the issue of the viability of sentence lengths or outer bounds is at least in part an empirical question. If so, the experience in the Northern District raises questions about some of the conventional wisdom on these upper bounds. Our data reveal that a substantial portion of offenders in the Northern District's program were ordered to make much larger commitments of time than was the case in programs in England, Germany, and New York City, or than is deemed practical by Morris and Tonry. In light of their discussion of equivalencies, the community service sentences imposed by the court in the Northern District of California constitute intermediate sanctions with a bite, since on average the length of service in the Northern District is one and a half times the upper limit Morris and Tonry posit. Figure 1 displays the frequency of CSO hours actually sentenced and served, which range from a low of 50 to a high of 2000 hours. The mean was 387 hours, the mode 100, and the median 250. Thus more than half of the offenders in the CSO program in the Northern District of California received sentences greater than those regarded as the upper limit by the country's leading experts on intermediate sanctions.3 9 In terms of a forty-hour workweek, the sentence terms in Figure 1 show that half of the offenders were sentenced to the equivalent of ful-time work for six and a half weeks or longer. Furthermore, just under thirty percent were sentenced to work for the equivalent of ten or more weeks. Translated into the equivalencies put forward by Morris and Tonry, the orders in the Northern District ranged from thirty days in jail to more than five years in prison, with the average (387 hours) being more than one and a half years in prison. Although we are not convinced by this "equivalency" scheme, to the extent that it does represent conventional thinking about CSOs, the experience in the Northern District of California raises the possibility that CSOs can amount to much more substantial punishment than is generally thought and indeed can be regarded as an alternative to imprisonment.40 39. Allen and Treger examined the CSO program in the U.S. District Court of the Northern District of Illinois and found that the average CS0 sentence was 300 hours, but they presented no figures on the range of sentences in their sample. See Allen & Treger, supra note 19. 40. There are not enough careful studies of CSO programs for us to do anything more than hazard a guess about the possible "outer bounds" of sentences. The Vera Institute seized on a HeinOnline -- 66 S. Cal. L. Rev. 168 1992-1993 BETWEEN TWO EXTREMES 1992] a E 0) 0 ._>0, 00) 0crI :c3 E a E4 a) ._ to 0 c'm N IL0U HeinOnline -- 66 S. Cal. L. Rev. 169 1992-1993 to 0 SOUTHERN CALIFORNIA LAW REVIEW A. [Vol. 66:155 THE NATURE OF COMMUNITY SERVICE WORK Most community service is pursued on a part-time basis, with offenders fitting their court-ordered job around existing schedules of work, school, child care, and so forth. The average number of hours per week devoted to community service varied widely in our study, but most were within a range of five to twenty hours. Although fifteen percent of the offenders worked intermittently from five to twenty hours per week, forty-three percent worked from five to twenty hours on a regular basis and only twelve percent worked more than twenty hours per week. In thirty percent of the cases, work schedules and hours varied during program participation in order to accommodate shifting child-care, employment, or school schedules, as well as holiday and vacation schedules. But here, too, in most cases the numbers of hours worked varied within the range of five to twenty hours per week. Work assignments varied considerably. Table 2 provides a breakdown of the jobs held by CSO participants. TABLE 2 Work Assignments Under Community Service Orders (N ASSIGNMENT Clerical Laborer Education Telephone Hotline/Counseling Kitchen Help Transportation Health Service Assistant Veterinary Assistant Fundraising Other Total 129) PERCENT 36 18 10 6 6 6 4 3 2 9 100% The three most frequent types of assignments were clerical, manual labor, and education, followed by staffing telephone hotlines, kitchen help, and transportation and delivery service. compact period of 70 hours of service in part because it targeted chronic offenders who were likely to have long histories of failing to appear in court, failing to meet conditions of probation, and the like. It limited the number of hours of service to 70 because the program was specially designed for irresponsible petty offenders. We appreciate that the federal system is likely to have a "better class" of offenders, see infra part VI.C., and no doubt this goes a long way in explaining why judges in the Northern District were able to impose longer sentences and, as we shall see, do so successfully. HeinOnline -- 66 S. Cal. L. Rev. 170 1992-1993 19921 BETWEEN TWO EXTREMES Work assignments were more complicated than what is suggested in Table 2. One third of those ordered to perform community service worked for more than one organization. This usually occurred when someone was referred to a clearinghouse, which matched participants' skills and interests with organizations' needs. In many such instances the clearinghouse matched participants with more than one assignment, although they maintained supervisory interest in the participant. This arrangement also helps explain why in interviews, one third of the supervisors reported that they had only indirect contact with participants. They served as "brokers," sending community service participants to various agencies that had requested workers, and thus did not directly supervise CSO workers. As noted earlier, most community service jobs are structured on a part-time basis to allow participants to fit them into existing schedules. In most instances in our study (sixty-two percent of the time), a schedule of regular part-time hours was arranged by mutual agreement. Typically this consisted of regular work assignments of between five and twenty hours per week. However, in nearly a quarter (twenty-three percent) of the cases, participants were allowed to set their own hours on a flexible basis. In most such instances, the designation and duration of a task were estimated by the organizations, which would then leave it up to the participants to complete the assignments at their own pace. The "other" category in Table 2 consists almost entirely of situations in which the length of time to complete work was estimated, and the participants were free to work at their own pace. In a few "other" situations, a task was assigned and it was up to the participant to complete it before an agreedupon deadline. In the vast majority of cases (eighty percent), supervisors kept time sheets with a running log of hours worked. In the remaining twenty percent of the cases, they relied on informal methods. Most of these instances involved participants with shorter sentences (that is, fewer than seventy-five hours), who completed their assignments in a matter of a few 41 weeks, not months. 41. As we indicated earlier, we found that most of the organizations maintained records only while volunteers were working. Once the work commitment had been completed, they usually dropped related information from their records. HeinOnline -- 66 S. Cal. L. Rev. 171 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW B. [Vol. 66:155 SUMMARY Although we have not offered a theory of punishment or a system of determining equivalencies between community service and imprisonment, we have compared the lengths of community service sentences in the Northern District of California to those of other programs and to the literature discussing the outer bounds of service lengths. We found that sentences in the Northern District of California were substantially longer than what is frequently regarded as the natural upper limit for community service sentences. This suggests that community service can be a substantially more stringent punishment than many of its enthusiastic proponents think. Consequently, we believe that it has greater potential for serving as an alternative to imprisonment than is currently believed. IV. THE ADMINISTRATIVE INTEGRITY OF COMMUNITY SERVICE ORDERS CSOs pose a twin administrative challenge. First, criminal offenders as a group are unreliable. By definition they have failed to conform to social conventions and often fail to maintain steady employment or keep regular schedules. Indeed, this is often a causal factor in their criminality. Thus, the program has to be designed for a group which on the whole is unreliable.42 Second, community service programs tend to be loosely administered. Their administration depends heavily upon staffs of nonprofit social service agencies rather than on correctional officers or court employees. These agencies depend heavily on volunteers and are staffed by people who have no special interest in (and often are philosophically opposed to) performing a criminal justice supervisory function. As Douglas McDonald notes in his monograph on community service, 4 3 this poses a substantial challenge for those who design community service sentences. In New York City the Vera Institute met this challenge by compacting community service into the shortest time period possible and by relying on intensive supervision by a full-time staff whose only job was to monitor work performance of program participants. But most other programs depend heavily on less formal and less well structured arrangements, which spread community service out over a much 42. We recognize that as a group federal offenders may be a "better class" of criminal than offenders in state courts. See infra part VI.C. This is because a high proportion of federal offenders eligible for CSOs are middle or working class and employed. Many of them have been convicted of "white collar" or property offenses-typically fraud or income tax violations-rather than street crime. For a detailed demographic description of a similar group of offenders in the Northern District of Illinois, see Allen & Treger, supra note 19. 43. McDONALD, supra note 7. HeinOnline -- 66 S. Cal. L. Rev. 172 1992-1993 1992] BETWEEN TWO EXTREMES longer time period and depend heavily on existing nonprofit and public service agencies to place and supervise participants. The CSO program in the Northern District of California was such a program. Although probation officers maintained nominal supervision of offenders, they did not have day-to-day oversight of community service participants. This supervision was the responsibility of a variety of designated supervisors in a host of nonprofit and public organizations that had agreed to use the "volunteer" services of the CSO participants. Thus our question was, is it possible for CSOs to be administered effectively by third parties who are not directly accountable to the courts? To date only a handful of studies has addressed this question, and some of those studies" conclude that loosely structured programs pose serious risks for effective administration. These risks arise not from willfulness or incompetence of recipient agencies, but because of a division of labor. Although probation and parole offices have continuing responsibility for offenders undertaking community service, they rarely provide direct supervision of work performance. Instead they rely on reports from supervisors in the receiving social service agencies. Staff in most of these agencies have no reason to assume a law enforcement function. Furthermore, they have no formal incentive to compel recalcitrant volunteers to fulfill their obligations or indeed even to report their substandard work performance. Finally, staff of those agencies who arrange for placement are usually not the same people who supervise volunteers and CSO participants on a day-to-day basis. More generally, organizations that rely heavily on volunteers to provide services typically do not find it worthwhile to try to compel reluctant volunteers to fulfill their commitments. The only thing worse than no volunteer, they believe, is an unreliable volunteer. Thus, rather than pressuring unreliable participants, they concentrate on making the most of the services of those volunteers who are reliable. Probation officials may reinforce this practice; although they require clients to check in with them periodically, they expect placement agencies to notify them if problems develop and do not devote much time to making independent checks on compliance with conditions of the CSOs. Thus, there is a relatively weak system of oversight in CSO placements. As simple as the idea of community service is, then, it entails a complex process and depends upon a sequence of independent decisions of low visibility and low priority to those who must make them. For exam44. See, e.g., FEELEY & KELLEY, supra note 7; McDONALD, supra note 7. HeinOnline -- 66 S. Cal. L. Rev. 173 1992-1993 SOUTHERN CALIFORNLI LAW REVIEW [Vol. 66:155 pie, a probation officer may arrange for an offender to meet with someone on the staff of a community service clearinghouse to arrange for a suitable placement. The appointment must be kept, and the clearinghouse must match the offender with a registered organization and direct him or her to meet with someone at that organization. Someone at the organization must receive the offender, describe the work, arrange for hours, and the like. Still someone else in the organization is likely to be given responsibility for supervising or working with the offender. All this takes place in organizations with high turnover and limited resources. As in the game Secret, information can be garbled as it is passed along. Thus, conditions, hours, and expectations can be lost or transformed in the process. Furthermore, as discussed earlier, some staff members in placement organizations believe as a matter of principle that they should not do anything to distinguish between pure volunteers who work with them and those who are there as a consequence of a court order. (This belief was asserted from time to time in the course of conducting this study. Indeed, a few supervisors questioned the legitimacy of our study, and a few refused to cooperate on the ground that they do not regard offenders sentenced to community service any differently from other types of volunteers.) This arrangement creates what Jeffrey Pressman and Aaron Wildavsky have called the "complexity of joint action." 4 s In a chain of decisions, each decision is based on that which preceded it. To the extent that different decision makers misunderstand each other, have variable interests, or give the same goals different priorities, implementation of the original goals-however straightforward and simple--can be compromised. And as we have suggested, in implementing a community service sentence there is ample opportunity for failure. It was these concerns that led us to ask supervisors to assess participants' work records. To this end we contacted the supervisor with whom each offender had the most direct contact, and asked him or her to describe the type of supervision they gave and to assess the offender's ability and performance in the community service program. In most instances (fifty-nine percent), supervision consisted of "direct and continuing contact" with the community service worker during regular working hours. In another thirty-three percent of the cases, supervision was best characterized as "indirect." Such supervision consisted of periodic review of the participants' work performance, collation of others' reports, 45. PRESSMAN & WILDAVSKY, supra note 10, at 75-78. HeinOnline -- 66 S. Cal. L. Rev. 174 1992-1993 1992] BETWEEN TWO EXTREMES and the like. But even in these instances supervisors had a continuing responsibility for overseeing the participants' work, and although they did not maintain continuous contact, supervisors nevertheless felt confident that they had reliable information about attendance, quality of work performance, and the like of those they indirectly supervised. In many of these arrangements, participants were given tasks to complete and were free to proceed independently at their own pace, reporting to their supervisors upon completion of assignments. In the remaining cases (eight percent) supervising relationships varied as a result of the participant's shifting employment schedule or because community service assignments varied. In such situations, no single person maintained continuous overall supervision of the CSO participant. In a handful of these cases, the lack of continuing supervision appeared to result from poor planning or careless administration. Our findings regarding the nature of supervision in the Northern District of California compare favorably with those of other studies. Each community service placement file we examined in the Northern District contained the name of a supervisor. In almost all instances these supervisors, when contacted, had a clear understanding of their job. And in the vast majority of cases supervisors indicated that they had taken or were taking their job seriously. Even in those instances when direct supervision did not occur, it was usually because the offender was working in more than one job or location. In short, in all but a small handful of cases, a system of supervision was in place and functioning. These findings contrast sharply with findings of an earlier study by one of the authors of this Article. This previous study of community service found that fully fifty percent of those who had been identified by the court as "third-party advocates" (that is, supervisors and/or counselors) had no clear understanding of their role and that a substantial portion of them did not even know that this responsibility had been assigned or ascribed to them by a court. 46 However, our findings are consistent with those reported by Frederick Allen and Harvey Treger, 7 who in their study of community service sentencing in the Illinois federal courts compared host agencies' perceptions of the performance of CSO participants to their perception of the performance of other volunteers who had worked for them and found 46. See FEELEY & KELLEY, supra note 7, at 30-31. 47. See Allen & Treger, supra note 19, at 13. HeinOnline -- 66 S. Cal. L. Rev. 175 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 that in more than half of the cases CSO participants were rated as more effective. This initial assessment of the success of supervision is reinforced by other indicators as well. Nearly two thirds (sixty-four percent) of the supervisors in our sample reported that they supervised six or more CSO placements, while just twenty-two percent handled only single cases. Thus, the staffs at placement agencies appear to be familiar with and to have had experience supervising "volunteers" who are serving as a condition of a court order. No doubt this experience is one reason for the high rate of completion of community service, the generally high quality of supervision, and the relatively low proportion of seriously problematic placements. We pursued this issue in still another way, by asking supervisors to assess participants' work performance by three related measures: (1) the quality of the participants' work, (2) punctuality and attendance, and (3) the need to call a probation officer to deal with unreliable participants. Responses to these queries reinforce the picture drawn earlier. According to supervisors' reports, not only did most participants complete their scheduled work assignments, but the vast majority of them also had good work performance evaluations (eighty-seven percent were rated as either "excellent" or "satisfactory") and regularly showed up for work (seventy-four percent had no attendance problems, and only nineteen percent presented "continuing problems"). Finally, only a handful-eight percent-presented problems serious enough to lead supervisors to contact probation officers. Follow-up queries with supervisors of the cases rated as troublesome by one or more of the three indicators (ten to twenty percent) revealed a variety of problems. The single largest set of problems had to do with scheduling conflicts arising from child care, transportation, health, or conflicts with full-time employment. In most instances, supervisors reported that these problems were straightened out by reassignment or restructuring work hours. In the ten percent of the cases in which there were serious enough problems for a probation officer to be called, the reason usually was failure to show up for work, repeatedly leaving work early, or, in some instances, gross exaggeration of numbers of hours worked. These figures must be read with considerable caution. They are based upon supervisors' responses to our queries about "satisfactory" or "unsatisfactory" performance and "problems," terms subject to variable understanding. Furthermore, supervisors revealed in follow-up HeinOnline -- 66 S. Cal. L. Rev. 176 1992-1993 BETWEEN TWO EXTREMES 1992] TABLE 3 Three Measures of Supervisors' Assessment of CSO Work (N = 129) A. Quality of Participants' Work RATING PERCENT Excellent Satisfactory Unsatisfactory Other 57 30 10 3 100 B. Problems with Appearing for Work REPORT None Some (at the outset) Some (continuing problems) Other PERCENT 74 5 19 2 100 C. Need to Report Problems to Probation Officers REPORT No Yes Other PERCENT 90 8 2 100 responses that they varied considerably in their reaction to problems. Some indicated little or no tolerance for nonperformance of duties, and others indicated that they made considerable allowances for problems caused by "acceptable" excuses, such as scheduling conflicts and the like. Still, it must be remembered that CSO participants were not only supervised at work, they also had to report on a regular basis to their probation officer, who in theory could and occasionally in practice did initiate contact with CSO supervisors for a performance report. Furthermore, as Table 3C indicates, eight percent of the participants were reported to probation for repeated failure to comply with the conditions of the community service order. Taken together, and in light of discussions of other community service programs, Table 3 portrays a relatively well functioning community service program. Not only do the vast majority of participants regularly show up for work assignments and complete them, but supervisors rate participants' work favorably and report a minimum of serious administration problems. In short the CSO program in the Northern District of California has administrative integrity. HeinOnline -- 66 S. Cal. L. Rev. 177 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 There are not enough careful examinations of American CSO programs with which to compare this program and form a basis for understanding why it appeared to work so well. But we can offer some informed judgments based on two different sets of factors. First, federal courts are well off relative to state criminal courts. They have lower caseloads, more probation officers, and the like. In short, in comparison to state courts, they are resource rich. Federal courts also have a "better class of criminal" than do state courts; criminal defendants in federal courts are likely to be better educated, be employed and employable, have fewer problems with substance abuse, and by any number of indicators be more conventional and more likely to be responsive to conventional practices, including regularly showing up to work. This is likely to be especially true for candidates for community service sentencing. Contrast, for instance, the proffle of participants in the Northern District of California's CSO program presented in Table 1 with the profile of chronic offenders in the New York City program.4" Second, the program's supervision was relatively effective. Although the program depended upon supervision by a large number of noncourt personnel, supervisors generally understood their duties and the conditions of work imposed by the court, and they were in contact with participants' probation officers. Additionally, CSO participants had to report to the probation officers as well. As indicated, probation supervision in federal courts is much more intensive than the perfunctory forms of supervision of most state systems. As such, probation officers can keep abreast of the activities of each of their clients in ways that are rarely possible in state systems. V. THE IMPACT OF COMMUNITY SERVICE ON INDIVIDUAL DETERRENCE: A COMPARISON OF RECIDIVISM RATES AMONG OFFENDERS RECEIVING THREE FORMS OF SENTENCES A. THE PROBLEM If convicted offenders are assigned to community service programs that are well run, Part IV asks, what impact does community service have on these offenders? The concern is that CSOs lack sufficient restraints and that as a consequence they expose the community to undue risk of danger. Of course, such a concern requires consideration of the risks posed by the alternatives to CSOs. How do community service 48. See McDONALD, supra note 7. HeinOnline -- 66 S. Cal. L. Rev. 178 1992-1993 19921 BETWEEN TWO EXTREMES sentences compare with other kinds of sentences? Our response is to compare three groups of offenders, each experiencing a distinct kind of federal sentence: (1) "straight" probation, (2) probation plus community service, or (3) incarceration followed by probation with community service.49 The outcome variable for these comparisons is recidivism, as defined by violations committed by the offender while under supervision and reported to the court.50 There are a number of ways "violation" could be defined. Here, we define it as a new arrest or any other indicator of criminal activity that has been reported by a probation officer, excluding technical violations."1 More precisely, we define "violation" as criminal misbehavior that has come to the attention of a probation officer. This definition seems sensible on its face and is also the one employed in the 52 data set available to us for this analysis. B. RESEARCH DESIGN AND ANALYsIs METHODS As framed above, our concern calls for a classic experimental design, comparing the effects of three alternative treatments. Ideally this should be pursued by random assignment of each offender to one of each of the three sentences. However a full experimental design was impossible. As a consequence, this is a retrospective study, an "observational" study in which subjects were assigned to treatments by processes that we as researchers could not control.5 3 Under such circumstances, the impact of the treatments upon the outcome has to be assessed in light of the mechanisms by which subjects were assigned to these treatments. One must understand the latter before one can interpret the significance of the former. For example, if "good risks" were disproportionately assigned to probation plus community service compared with the other two treatments, recidivism might well be lowest for that group. However, the lower recidivism might have far more to do with which kinds of people 49. Since the vast majority of such sentences limit time behind bars to a year or less, these are the only sentences included in the analysis. 50. We appreciate that recidivism rates do not uncover the "dark figure of crime," but like others we believe that this indicator nevertheless has value. See Albert D. Biederman & Albert J. Reiss, Jr., On Exploring the Dark Figure of Crime, 373 ANNA.S AM. ACAD. POL. & Soc. SC. 1 (1967). 51. This definition follows the one used by the courts and recorded in the records of the Administrative Office of the U.S. Courts. 52. See infra part V.B. 53. See WiLAM G. COCHRAN, PLANNING & ANALYSIS OF OBSERVATIONAL STUDIES (1983). HeinOnline -- 66 S. Cal. L. Rev. 179 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 were in each of the three groups than with any treatment impact, and any computed differences in average recidivism could be potentially misleading. This implies the need for two models: one for the assignment process and one for the treatment impact. Basically, we must first try to represent accurately the assignment process for the three kinds of sentences, and we must then, using information from that first analysis, try to estimate accurately the impact of the treatments. To accomplish this we employed techniques that in principle compensate for possible selection biases via known covariates. In particular we employed "propensity scores" as the means of taking group composition properly into account. 54 The technique, in brief, is to compute for each offender the probability of being assigned to each possible sentence. This will naturally vary across offenders, since offenders with different backgrounds and crimes have different probabilities of receiving particular sentences. An offender with no prior record, for example, might have a probability of 0.25 of receiving a prison term, and another with a serious criminal record might have a probability of 0.50 of receiving a prison term. It is precisely this sort of variation that is the source of potential biases in any estimate of treatment effects, because the sentence received is confounded with characteristics of the offender and the crime, which, in turn, could affect the likelihood of recidivism. In practice, one proceeds by developing a statistical model for the assignment process that computes the necessary probabilities. In the second step, one uses these estimated probabilities as control variables in the analysis of treatment impact. That is, they are included in the analysis to help make the treatment groups more comparable on the average. In an analysis of covariance, for example, the estimated probabilities (propensity scores) would serve as the covariates. Consistent with these principles, we regressed membership in one of the three treatment groups on variables universally regarded as central factors accounting for sentence severity (the assignment process, using multinomial logistic regression): prior record, seriousness of the crime, and links to the community. From these equations, we then constructed for each offender the estimated probability of being assigned to each of the three groups. Finally, the estimated probabilities were used as control variables in an analysis in which a binary variable for recidivism 54. See Richard A. Berk & Phyllis J. Newton, Does Arrest Really Deter Wife Battery? An Effort to Replicate the Findingsof the MinneapolisSpouse Abuse Experiment, 50 AM. Soc. REv. 25362 (1985); Paul R. Rosenbaum & Donald B. Rubin, Reducing Bias in ObservationalStudies Using Subclassification on the Propensity Score, 79 L AM. STATISICAL ASS'N 516-24 (1984). HeinOnline -- 66 S. Cal. L. Rev. 180 1992-1993 19921 BETWEEN TWO EXTREMES (violation/no violation) was regressed on treatment group membership (using binary logistic regression). In effect, we relied on propensity scores to control for the assignment process. Statistical tests revealed that the three treatment groups were still not "balanced" on a number of possible confounding variables after propensity-score adjustments were made; possible confounding influences remained. This implied that we had not properly modeled the assignment process. In response, we broke down the assignment process further. In particular, the judges' choice between incarceration and no incarceration was separated from a later choice between straight probation and probation plus community service (among those not incarcerated). That is, the choice to append community service to the sentence of probation was nested within a prior decision whether to incarcerate the offender. This implied that there are two rather different processes underlying assignment of sentences, not one. Initially we had been in effect averaging across the two and obscuring differences between them. Taking this into account (with nested bivariate logistic regressions) led to three groups that were balanced for all of the potential confounding variables we examined. The end result was two distinct analyses. The first included all offenders and addressed the relative effectiveness of incarceration in reducing recidivism compared with the two types of probation combined. The second analysis focused only on those offenders not incarcerated and examined the relative effectiveness of probation plus community service compared with straight probation. It is beyond the scope of this Article to consider the trade-offs between the use of propensity scores to control for selection biases and other methods to control for selection biases. An accessible discussion can be found in Richard Berk and Phyllis Newton's 1985 article Does Arrest Really Deter Wife Battery?55 However it is important to stress that our methods depend on knowing and measuring the variables by which the assignment to sentences operates and on developing a proper statistical model of those processes (much as in an analysis of covariance). In this regard, we were fortunate to be able to collect our data from the forms filled out by probation officers from which sentencing recommendations were made. That is, we used the information on which sentencing recommendations were supposed to be based. In addition, we 55. Berk & Newton, supra note 54, at 253-62. HeinOnline -- 66 S. Cal. L. Rev. 181 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 questioned probation officers at some length about how they constructed their recommendations. Thus, we have considerable qualitative knowledge of how the information on the forms was used. The implication is that our model specification derives from extensive preliminary work. Moreover, the propensity-score form of analysis leads to specification tests for "balance" that often flag specification errors. In short, although statistical adjustments are no substitute for random assignment,5 6 they may be about as good as one can do in an observational study. C. DATA The analysis in parts I and II, dealing with the structure of the CSO program, required information about processes internal to the CSO program. Much of this information could be obtained only through interviews with CSO supervisors and field work and thus had to be restricted to relatively recent cases. In contrast, the inquiry in this section does not depend upon detailed knowledge about the internal operations of each of these treatments. We were able to base this portion of the analysis upon a much more substantial database-information routinely gathered for years by the Administrative Office of the U.S. Courts, which codes and collects a wealth of detail about each case, the nature of the offense, offender characteristics, case disposition, and sentence. These data were provided by the Administrative Office of the U.S. Courts in Washington, D.C., and covered all convictions in the Northern District of California between January 1984 and June 1988. Included was virtually all of the information recorded on the Report of Federal Sentence, and computer-entered follow-up actions on each supervised case. The outcome variable was "supervisory violations," which, according to the codebook obtained from the U.S. Sentencing Commission, 57 excludes purely technical violations. 56. Richard A. Berk et al., Social Policy Experimentation:A Position Paper, 9 EVALUATION REV. 387-401 (1985). 57. We relied entirely on the judgment of the Administrative Office (A.O.), which kindly supplied us with these types. It was beyond the scope of this study to verify independently the process by which the judgments reflected in these data were initially made, recorded, and collated from the A.O.'s tapes. Independent information supplied by the A.O. and court clerks indicates that these data are obtained, reported, and collated in a fairly consistent and standardized way nationwide. It is perhaps worth mentioning that without the Commission's codebook, the analysis would have been impossible. Codebooks maintained by the A.O. were seriously inaccurate in a number of places. HeinOnline -- 66 S. Cal. L. Rev. 182 1992-1993 1992] BETWEEN TWO EXTREMES D. RESULTS Table 4 shows the variables used in the analysis of which offenders were incarcerated. At the top, one can see that fourteen percent were incarcerated and eighty-six percent were not. Clearly, incarceration is not a popular sentence, but one must keep in mind that people with prison terms in excess of one year were excluded. This is because the relevant comparison group for offenders receiving CSOs was not all other offenders but only those receiving terms of incarceration for which community service was a viable alternative. Our cutoff point, reinforced by experimentation with construction of the propensity scores, was one year of incarceration. A number of other descriptive statistics are also of interest in their own right. For example, more than ninety percent of the offenders used no weapon, and about two thirds were male. Far more important are the logistic regression results shown in Table 5. This is a conventional analysis in which the outcome is defined as "l" if a person was incarcerated and "0" if a person was not. All of the logistic regression coefficients can be interpreted as odds multipliers when the constant (e) is raised to the power of the coefficients. Consider, for example, the coefficient for "crime was a felony." The odds multiplier for the coefficient of 1.47 is 4.34, which means that the odds of being assigned to prison if one possessed this characteristic are multiplied by 4.34 (a substantial increase). Analogously, the coefficient of -0.808 for "never been to jail" means that the odds of being assigned to prison are multiplied by .446 (a substantial reduction) if the offender had never before been committed to jail or prison."8 At this point, however, it may suffice simply to note the direction of each effect and whether it is statistically significant by conventional standards. This can be accomplished most easily by scanning the column in Table 5 labeled "t-value" and finding all t-values greater than 2.0 in absolute value. The sign of the t-value is the direction of the effect. For example, men are much more likely than women to be incarcerated, other things being equal. If the recidivism analysis that follows is to be credible, the results in Table 5 must be, by and large, intuitively plausible. By and large they are. In particular, the variables that one might associate with forecasts of 58. All of the effects represented in the table assume all other things are equal. These "other things" are all the variables included in the analysis (and Table 12). HeinOnline -- 66 S. Cal. L. Rev. 183 1992-1993 184 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 TABLE 4 Descriptive Statisticsfor Variables Used in the Assignment of Analysis for Incarceration Versus No Incarceration(N = 1520) KINDS OF SENTENCES PERCENT EXPLANATORY VARIABLES Constant No. of Defendants Dollar Amount in Crime No. of Juvenile Convictions No. of Adult Convictions No. of Times Incarcerated Less Than One Year No. of Times Incarcerated One to Five Years No. of Times Incarcerated More Than Five Years No. of Priors No. of Prior Revocations Months Employed Monthly Income (Dollars) Male Black White Hispanic American Indian Asian One Act in Offense Acted Alone No. CI Status This Offense No. CJ Status Prior Offense Crime Was a Felony Age Crime Was a Drug Offense College or More Education Lives w/Another Person History of Drug Abuse Needs Employment Alcohol Dependent Needs Shelter Illiterate Needs Vocational Training Physically Infirm Needs Psychological Help Needs Financial Help No Weapon Involved in Offense Never Been to Jail Jail No Jail 0.1395 0.8605 MEAN STD DEV 1.00 1.50 287,058.62 0.14 0.97 0.00 1.56 MINIMUM MAXIMUM 4,205,485.19 0.75 1.88 1.00 1.00 0.00 0.00 0.00 1.00 17.00 95,000,000.00 9.00 9.00 0.22 0.88 0.00 9.00 0.07 0.38 0.00 4.00 0.01 0.18 0.26 8.45 1264.96 0.66 0.31 0.02 0.00 0.09 0.45 0.62 0.88 0.87 0.55 37.37 0.17 0.42 0.38 0.21 0.20 0.11 0.01 0.05 0.10 0.02 0.16 0.12 0.94 0.88 0.16 0.00 0.75 1.20 4.83 2784.01 0.47 0.46 0.15 0.07 0.29 0.49 0.48 0.31 0.33 0.49 10.50 0.37 0.49 0.48 0.41 0.40 0.31 0.10 0.21 0.30 0.16 0.37 0.32 0.23 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 20.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 4.00 9.00 9.00 12.00 85,000.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 76.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 HeinOnline -- 66 S. Cal. L. Rev. 184 1992-1993 1992] BETWEEN TWO EXTREMES TABLE 5 Estimatesfrom Binary Logit Analysis of Assignment Process (IncarcerationVersus No Incarceration) LoGiT ESTIMATE STD ERROR Constant No. of Defendants Dollar Amount in Crim e No. of Juvenile Convictiions No. of Adult Convictionis No. of Times Incarcerat ed Less Than One Year No. of Times Incarcerat ed One to Five Years No. of Times Incarceratted -2.364 -0.007 0.6777 0.0508 More Than Five Yea No. of Priors VARIABLE t-VALUE p>t 0.000 0.1086 0.0595 -3.49 -3.49 0.43 -0.52 2.54 -0.076 0.1225 -0.62 0.534 -0.859 0.2574 -3.34 0.001 -1.317 0.042 0.107 0.013 -0.000 0.7848 0.1014 0.0664 0.0190 0.0001 0.662 0.2061 -1.68 0.42 1.62 0.70 -1.01 3.21 1.94 1.35 0.98 1.14 -3.02 -2.36 -1.52 -0.21 7.00 0.11 -1.34 0.65 -0.19 2.65 -0.08 0.11 -0.57 1.14 -0.30 -0.36 0.12 1.74 0.35 -2.66 0.093 0.674 0.106 0.482 0.314 0.001 0.052 0.178 0.326 0.256 0.003 0.018 0.130 0.835 0.000 0.916 0.179 0.514 0.852 0.008 0.937 0.910 0.572 0.253 0.764 0.718 0.905 0.082 0.724 0.008 0.000 0.0000 -0.056 0.151 No. of Prior Revocation Ls Months Employed Monthly Income (Dolla rs) Male 0.382 0.1970 Black White Hispanic 0.624 0.4638 American Indian 0.867 0.8824 Asian 0.346 0.3052 -0.531 0.1760 One Act in Offense Acted Alone -0.417 0.1771 No. CJ Status This Offense -0.497 0.3284 No. CJ Status Prior Offense -0.065 0.3118 Crime Was a Felony 1.475 0.2109 Age 0.001 0.0088 Crime Was a Drug Offe rse -0.332 0.2476 College or More Educat ion 0.114 0.1752 Lives w/Another Person -0.033 0.1765 History of Drug Abuse 0.543 0.2047 Needs Employment -0.017 0.2169 Alcohol Dependent 0.028 0.2538 -0.431 0.7634 Needs Shelter Illiterate 0.028 0.3643 Needs Vocational Training -0.089 0.2970 Physically Infirm -0.176 0.4891 Needs Psychological HelIp 0.026 0.2208 Needs Financial Help 0.432 0.2481 No Weapon Involved in Offense 0.118 0.3359 Never Been to Jail -0.808 0.3033 Table of Observed and PredictedOutcomes: OBSERVED Jail No Jail Total PREDICTED JAIL No JAIL 17 20 37 195 1288 1483 TOTAL 212 1308 1520 PERCENT CORRECTLY PREDICTED: 85.8553 HeinOnline -- 66 S. Cal. L. Rev. 185 1992-1993 0.878 0.670 0.600 0.011 SOUTHERN CALIFORNIA LAW REVIEW (Vol. 66:155 future problems are more likely to be found among incarcerated individuals.5 9 In principle, there are a number of ways the estimated propensities (or propensity scores) could be used to control for the impact of variables confounded with the sentence given. In the following analysis of the impact on recidivism of sentence given, we separate the propensity scores into five groups at the quintiles and include them in a binary logistic regression as dummy (binary) variables. While this discards some information, it has allowed us to proceed conservatively. That is, no functional form for the relationships between the propensity scores and the outcome (recidivism) needs to be assumed. Table 6 shows the variables used in the analysis of assignment to straight probation versus probation plus community service (that is, the second step in the nested sentencing process). Table 7 shows the logistic regression results for the assignment process. Overall, it is more difficult to explain the assignment between the two kinds of probation than the "in-out" assignment. Yet, the results seem to make sense. For example, offenders who were convicted of felonies were less likely to be assigned to community service, while those who were college educated were more likely to be assigned to community service. Table 8 shows the variables used in the impact assessment of incarceration. Table 9 shows the subsequent logistic regression results. "Time under observation" measures the elapsed time in months from the beginning of probation to the end of the probation period or the end of the study, whichever came first. It is basically time at risk in our data and a necessary control if the length of time under supervision (1) varied by treatment and (2) was correlated with the commission of new crimes. For offenders who committed new crimes, we use this nominal time measure because to use actual time under supervision would confound the time measure with the outcome (since people who committed new crimes would be under supervision for shorter periods). As expected, the regression coefficient is positive. In odds-multiplier terms, each additional month under supervision increased the odds of committing a new crime by a multiplier of 1.03, which works out to be 1.5 times greater 59. Since the statistical point is not to build a plausible causal model of the incarceration decision but to capture all the possible variability in that decision (akin more to a forecasting problem than a substantive problem), there are a few somewhat redundant variables in the question. This makes some of the regression coefficient point estimates unstable and occasionally leads to surprising signs. HeinOnline -- 66 S. Cal. L. Rev. 186 1992-1993 19921 BETWEEN TWO EXTREMES TABLE 6 Descriptive Statisticsfor Variables Used in the Assignment Analysis for Community Service Versus No Community Service (N = 1308) KINDS OF SERVICE PERCENT EXPLANATORY VARIABLES Constant No. of Defendants Dollar Amount in Crime No. of Juvenile Convictions No. of Adult Convictions No. of Times Incarcerated Less Than One Year No. of Times Incarcerated One to Five Years No. of Times Incarcerated More Than Five Years No. of Priors No. of Prior Revocations Months Employed Monthly Income (Dollars) Male Black White Hispanic American Indian Asian One Act in Offense Acted Alone No. CJ Status This Offense No. CJ Status Prior Offense Crime Was a Felony Age Crime Was a Drug Offense College or More Education Lives w/Another Person History of Drug Abuse Needs Employment Alcohol Dependent Needs Shelter Illiterate Needs Vocational Training Physically Infirm Needs Psychological Help Needs Financial Help No Weapon Involved in Offense Never Been to Jail Service Hours 0.3647 No Service Hours 0.6353 MEAN SD DEV 1.00 1.47 273 ,553.38 0.13 0.84 0.00 1.55 4,242,769.39 0.71 1.72 1.00 1.00 0.00 0.00 0.00 1.00 17.00 95,000,000.00 9.00 9.00 0.18 0.82 0.00 9.00 0.06 0.38 0.00 4.00 0.01 0.16 0.16 0.68 0.21 8.53 1288.03 0.64 0.31 0.02 0.00 0.09 0.47 0.64 0.90 0.89 0.51 37.20 0.16 0.42 0.37 0.19 0.20 0.10 0.00 0.04 0.10 0.02 0.16 0.12 0.94 0.90 1.08 4.80 2940.59 0.47 0.46 0.15 0.06 0.29 0.49 0.47 0.29 0.31 0.49 10.59 0.37 0.49 0.48 0.39 0.40 0.30 0.09 0.21 0.30 0.15 0.37 0.32 0.23 0.29 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 20.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 4.00 9.00 9.00 12.00 85,000.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 76.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 MINIMUM HeinOnline -- 66 S. Cal. L. Rev. 187 1992-1993 MAXIMUM SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 TABLE 7 Estimatesfrom Binary Logit Analysis of Assignment Process (Community Service Versus No Community Service) VARIABLE Constant No. of Defendants Dollar Amount in Crime No. of Juvenile Convictions No. of Adult Convictions No. of Times Incarcerated Less Than One Year No. of Times Incarcerated One to Five Years No. of Times Incarcerated More Than Five Years No. of Priors No. of Prior Revocations Months Employed Monthly Income (Dollars) Male Black White Hispanic American Indian Asian One Act in Offense Acted Alone No. CJ Status This Offense No. CI Status Prior Offense Crime Was a Felony Age Crime Was a Drug Offense College or More Education Lives w/Another Person History of Drug Abuse Needs Employment Alcohol Dependent Needs Shelter Illiterate Needs Vocational Training Physically Infirm Needs Psychological Help Needs Financial Help No Weapon Involved in Offense Never Been to Jail LOGIT STD ESTIMATE ERROR t-VALUE -0.147 0.047 -0.000 -0.366 -0.042 0.6168 0.0421 0.0000 0.1768 0.0636 -0.24 1.12 -0.74 -2.08 -0.66 p>t 0.811 0.263 0.459 0.038 0.507 -0.077 0.1772 -0.43 0.664 -0.208 0.2896 -0.72 0.472 -0.295 -0.201 0.063 0.008 -0.000 0.037 -0.406 0.232 1.441 0.112 -0.127 -0.041 -0.216 -0.144 -0.345 -0.002 0.079 0.348 -0.141 -0.000 0.274 -0.345 -0.929 -0.354 -0.019 0.116 -0.056 0.296 0.270 -0.462 0.5177 0.1439 0.0843 0.0142 0.0000 0.1342 0.1499 0.3882 0.8955 0.2066 0.1243 0.1409 0.3211 0.2953 0.1340 0.0063 0.1943 0.1255 0.1278 0.1754 0.1635 0.2205 0.7950 0.3323 0.2235 0.3829 0.1703 0.1804 0.2961 0.3463 -1.57 -1.40 0.75 0.57 -1.08 0.28 -2.71 0.60 1.61 0.54 -1.02 -0.29 -0.67 0.49 -2.58 -0.34 0.41 2.78 1.11 -0.00 1.68 -1.57 -1.17 -1.07 -0.09 0.30 -0.33 1.65 0.92 -1.34 0.568 0.161 0.452 0.571 0.282 0.778 0.007 0.550 0.107 0.586 0.307 0.770 0.500 0.624 0.010 0.731 0.683 0.006 0.269 0.997 0.093 0.117 0.242 0.286 0.930 0.761 0.739 0.100 0.360 0.182 Table of Observed and PredictedOutcomes: PREDICTED OBSERVED Srv No Srv Total I SRV 66 56 122 No SRV 411 775 1186 PERCENT CORRECTLY PREDICTED: TOTAL 477 831 1308 64.2966 HeinOnline -- 66 S. Cal. L. Rev. 188 1992-1993 19921 BETWEEN TWO EXTREMES TABLE 8 Descriptive Statisticsfor Key Variables of Impact Analysis, Incarceration Versus No Incarceration(N = 1520) VARIABLE Constant Probability of Jail Sentence, Quintile Probability of Jail Sentence, Quintile Probability of Jail Sentence, Quintile Probability of Jail Sentence, Quintile Time Under Observation (Months) Jail Treatment 1 2 3 4 MEAN STD DEV MINIMUM MAXIMUM 1.00 0.20 0.20 0.20 0.20 23.35 0.13 0.00 0.40 0.40 0.40 0.40 12.88 0.34 1.00 0.00 0.00 0.00 0.00 1.00 0.00 1.00 1.00 1.00 1.00 1.00 53.00 1.00 TABLE 9 Estimatesfrom Binary Logit Analysis of Treatment Impact (IncarcerationVersus No Incarceration) VARIABLE Locrr SD ESTIMATE ERROR Constant -3.145 0.3072 Probability of Jail Sentence, Quintile 1 -0.542 0.3323 Probability of Jail Sentence, Quintile 2 -0.134 0.2994 Probability of Jail Sentence, Quintile 3 -0.416 0.3089 Probability of Jail Sentence, Quintile 4 -0.460 0.3050 Time Under Observation (Months) 0.033 0.0077 Jail Treatment 0.417 0.2658 Table of Observed and PredictedOutcomes: t-VALUE p>t -10.24 -1.63 -0.45 -1.35 -1.51 4.50 1.57 0.000 0.102 0.654 0.177 0.132 0.000 0.116 PREDICTED OBSERVED Viol No Viol Total VIOL 0 0 0 No VIOL 114 1406 1520 TOTAL 114 1406 1520 PERCENT CORRECTLY PREDICTED: 92.5000 probability of committing a new crime for each additional year under supervision. The variables Quintile 1 through Quintile 4 are the binary variables for the propensity-score quintiles (with Quintile 5 deleted as the comparison). Their story is not terribly clear. Small differences exist among the four quintile effects, although the lowest of the four quintiles perhaps does worse than the highest. That is, those individuals with the greatest probability of being assigned to prison have the greatest likelihood of committing new violations while on probation. Finally, while the coefficient for "jail treatment" is not statistically significant at conventional levels, it is close enough to warrant some attention. The coefficient of 0.417, if taken seriously, means that the odds of committing a new violation are about 1.5 times higher among HeinOnline -- 66 S. Cal. L. Rev. 189 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 those offenders who served time behind bars. What is clear is that it is highly unlikely that incarceration improves offenders' chances of staying out of trouble. Tables 10 and 11 repeat the analysis process for the impact of straight probation versus probation plus community service. From Table 11, it is apparent that greater amounts of "time under observation" are associated with violations and that individuals with a higher probability of being assigned to probation plus community service are less likely to commit new violations. Perhaps in the eyes of probation officers, community service is reserved for the "good guys." Yet, there is not even a TABLE 10 Descriptive Statisticsfor Key Variables of Impact Analysis (Community Service Versus No Community Service) (N = 1308) OUTCOMES PERCENT Violation 0.0688 No Violation 0.9312 VARIABLE MEAN STD DEV MINIMUM MAXIMUM Constant Probability of Service Hours, Quintile 1 Probability of Jail Sentence, Quintile 2 Probability of Jail Sentence, Quintile 3 Probability of Jail Sentence, Quintile 4 Time Under Observation Community Service Treatment 1.00 0.19 0.20 0.19 0.20 23.29 0.36 0.00 0.39 0.40 0.39 0.40 12.90 0.48 1.00 0.00 0.00 0.00 0.00 1.00 0.00 1.00 1.00 1.00 1.00 1.00 53.00 1.00 TABLE 11 Estimatesfrom Binary Logit Analysis of Treatment Impact (Community Service Versus No Community Service) LOGIT ESTIMATE VARIABLE STD ERROR Constant -4.108 0.4643 Probability of Service Hours, Quintile 1 1.557 0.4334 Probability of Service Hours, Quintile 2 0.991 0.4554 Probability of Service Hours, Quintile 3 0.963 0.4556 Probability of Service Hours, Quintile 4 0.629 0.4808 Time Under Observation 0.024 0.0086 Community Service Treatment -0.204 0.2519 Table of Observed and PredictedOutcomes: OBSERVED Viol No Viol Total VIOL PREDICTED No VIOL 0 0 0 90 1218 1308 PERCENT CORRECTLY PREDICTED: TOTAL 90 1218 1308 93.1193 HeinOnline -- 66 S. Cal. L. Rev. 190 1992-1993 t-VALUE p> t -8.85 3.59 2.18 2.12 1.31 2.89 -0.81 0.000 0.000 0.029 0.034 0.191 0.004 0.418 19921 BETWEEN TWO EXTREMES hint of a treatment effect. The presence or absence of community service has no impact on recidivism once the assignment process is held constant. E. SUMMARY AND DISCUSSION These results represent the end of a long process of scrutinizing our initial findings using a variety of diagnostics. For the kinds of offenders represented in this data and for the particular implementations of the sentences examined, recidivism rates were perhaps highest for offenders who were sentenced to incarceration plus probation. Furthermore, it is virtually certain that incarcerated offenders did no better under supervision than those assigned to straight probation or probation plus community service. One central implication follows from these findings on the impact of these sentence alternatives: Alternatives to incarceration "do" at least no worse than incarceration. That is, recidivism rates are at least no higher, and the costs of the noncustodial sentence are lower. This is not to suggest that our study has proffered "the" answer on the issue. The three types of sentences need to be studied more rigorously with an even stronger research design. Ideally, a random assignment to treatment should be employed. Propensity-score adjustments such as those we have used are clearly a second (or third) best alternative, and they are no substitute for random assignment. In light of the suggestive findings of this report, we strongly encourage experimentation with this more sophisticated research design. Additionally, a stronger design should consider the possibility that different sentences are differentially effective; an effective sentence for one kind of offender is an ineffective sentence for another kind of offender. VI. POLICY IMPLICATIONS: RESTRUCTURING THE GUIDELINES TO CREATE A PRESUMPTIVE COMMUNITY SERVICE SENTENCE The findings outlined in the previous sections describe a relatively well functioning community service punishment system and indicate that we can reasonably regard community service as a punitive sanction in its own right. If these findings for the Northern District of California are generalizable-and we believe they are-they suggest the desirability of expanding community service as a form of intermediate punishment in its own right. In this part we suggest ways to think about how to accomHeinOnline -- 66 S. Cal. L. Rev. 191 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 plish this goal within the framework of the U.S. Sentencing Commission's guidelines. In particular, we engage in a thought experiment in which community service is treated as a presumptive sentence to be inserted into the guidelines between those offenses for which straight probation is the presumptive sentence and those offenses for which incarceration is the presumptive sentence. Most judges in the Northern District of California currently impose community service regularly, as a condition of straight probation or as part of a split sentence, in which community service is a condition of probation after serving a term in prison. If they continue to use it in this manner, only a small portion of offenders will receive community service in lieu of a custodial sentence. Even if community service sentences were used much more extensively, they would either be "add-ons" to straight probation or "pile-ons" to conditions of release after custody. Unless the guidelines are amended, community service quickly runs up against the presumptions of either straight probation or a custodial term, and the only way to expand it as a punishment of its own is to change the guidelines. However, as we show below, such changes do not have to be drastic to have a significant consequence. We wish to show how incremental adjustments in the guidelines could yield significantly expanded use of community service, and to explore related consequences of such a shift. More particularly, we think expanded reliance on intermediate punishments would be valuable in light of the following four sets of considerations: First, the findings discussed in parts I through IV reveal that community service is substantially more punitive-in terms of numbers of hours served-than is commonly believed, and if this is found to be the case generally, then community service should be more prominently featured in a system of punishments. Second, our findings on recidivism suggest that marginal expansion of eligibility for community service could lead to substantial increases in the use of community service without measurably jeopardizing public safety. Third, meaningful marginal adjustments in current guidelines are possible without disrupting the Commission's overall sentencing policy or altering widely accepted principles of sentencing. Fourth, marginal changes in the existing guidelines providing for presumptive sentences of community service could have more than a HeinOnline -- 66 S. Cal. L. Rev. 192 1992-1993 19921 BETWEEN TWO EXTREMES marginal impact on the federal sentencing process, on offenders, and on the prison population. We demonstrate this below. In sum, we think that community service punishments should not be regarded as idiosyncratic sentences to be imposed in rare and isolated instances. Rather, they should be regarded as an integral part of a rational sentencing system. A. How COMMUNITY SERVICE FITs INTO THE FEDERAL SENTENCING SCHEME Let us illustrate our concerns by considering the federal sentencing scheme in some detail. For purposes of illustration, we draw upon two tables, the Commission's standard Sentencing Table and Table M, reprinted in the Commission's annual reports. The Sentencing Table conforms to the by now familiar grid of all sentencing guidelines. The vertical axis arrays offense levels in fortythree categories, and the horizontal axis arrays criminal history (prior record) along six levels. The result is a grid with 258 different combinations of cells, or offense levels and criminal history categories. The heavy lines moving from the middle left to the upper right delineate punishment levels. Offenders (as defined by criminal offense level and criminal history) above the lines receive less severe punishments and those below, more severe. Unlike simpler state guidelines, which have only one such heavy line separating presumptively probationary sentences from presumptive imprisonment, the U.S. sentencing guidelines have three lines that distinguish four groups (groups A through D). In the discussion that follows, we propose to marginally redefine and adjust these four groups in order to create a middle group of offenders (as defined by offense level and criminal history) for which the presumptive sentence would be community service. But first let us explore the salient distinctions among the existing groups. Using the first column (the offense levels in combination with Category I of the criminal history scale) to identify, but not to define, these groupings, we find that Group A offenses consist of offense levels 1-6; Group B, levels 7-10; and Group C (which overlaps with B), levels 7-12. Group D, which includes the balance of the most serious offenses, constitutes the vast bulk of offense levels, 13-47. Note, however, that each of these groupings pushes upward as it moves to the right, indicating that within each offense level, as the offender's criminal history increases, so too does the presumptive sentence to be imposed. HeinOnline -- 66 S. Cal. L. Rev. 193 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW 194 [Vol. 66:155 TABLE 1260 SENTENCING TABLE (in months of imprisonment) Criminal History Category (Criminal History Points) Offense I II III IV V 153-21 18-24 15-21 18-24 21-27 24-30 27-33 30-37 33-41 37-46 41-51 46-57 51-63 57-71 63-78 70-87 78-97 87-108 97-121 108-135 121-151 135-168 151-188 169-210 188-235 210-262 235-293 262-327 292-365 324-405 360-hfe life 15-21 18-24 21-27 24-30 27-33 30-37 33-41 37-46 41-51 46-57 51-63 57-71 63-78 70-87 78-97 87-108 97-121 108-135 121-151 135-168 151-188 168-210 182-235 210-262 235-293 262-327 292-365 324-405 360-hfe 360-life life 12-24 21-27 24-30 27-33 30-37 33-41 37-46 41-51 46-57 51-63 57-71 63-78 70-87 72-97 87-108 97-121 108-135 121-151 13-168 151-188 169-210 188-235 210-262 235-293 262-327 292-365 324-405 360-life 360-life 360-life life 21-27 24-30 27-33 30-37 33--41 37--46 41-51 46-57 51-63 57-71 63-78 70-87 77-96 84-105 92-115 100-125 110-137 121-151 135-165 151-188 168-210 182-235 210-262 235-293 262-327 292-365 324-405 360-life 360-life 360-life 360-life life 24-30 27-33 30-37 33-41 37-46 41-51 46-57 51-63 57-71 63-78 70-87 77-96 84-105 92-115 100-125 110-137 120-1 0 130-162 140-17S 151-188 169-210 18-235 210-262 235-293 262-327 292-365 324-405 360-life 360-hfe 360-life 360-hfe 360-hfe lfe VI 15-21 1-24 21-27 24-30 27-33 30-37 33-41 37-46 41-1 46-57 51-63 57-71 63-78 7O-87 77-96 84-105 92-115 100-125 110-137 120-150 130-162 140-175 151-188 169-210 188-235 210-262 235-293 262-327 292-365 324-405 360-hife 360-life 360-life 360-life 360-life 360-hfe ie KEY A-Probation available B-Probation with conditions of confinement available C-New "split sentence" available 60. U.S. SENTENCING COMM'N, FEDERAL SENTENCING GUIDELINES MANUAL (Sentencing Table). HeinOnline -- 66 S. Cal. L. Rev. 194 1992-1993 242 (1992) 19921 BETWEEN TWO EXTREMES TABLE 1361 OFFENSE LEVEL BY CRIMINAL HISTORY CATEGORY For Guideline Cases With Complete Reports on the Sentencing Hearing Received (October 1, 1989 through September 30, 1990) CRIMINAL HISTORY CATEGORY [L OFFENSE L EVE LEVELP*n 1 12 2 365 3 152 1 0 OA VIT V IV III I I 1 0 14 04 3 456 23 10 236 12 24 34 11 5 110 53 27 47 1 6.59 83 :30 23 5 12 418 2I 1269 64 4 1,271 143 5 304 44 6 712 121 153 120 69 14 7 578 97 114 73 53 68 479 89 4 56 31__ 375 58 69 35 41 10 871 150 190 117 26 71 I1 348 64 79 58 53 12 562 114 122 75 33 13 242 41 45 36 14 481 95 91 IS 140 27 is 450 17 Is II ! 963 49 74 31 607 1.495 30 75 63 665 3.3 48 94 48 26 21 411 21 57 25 20 769 3-9 29 12 5 29 242 12 88 90 27 20 12 696 35 136 27 36 19 13 32 263 13 360 72 93 35 7 13 580 29 74 24 31 14 5 I1 159 08 408 95 36 27 21 653 33 92 20 66 71 12 3 379 83 64 31 22 21 600 30 73 14 is 22 4 6 137 07 142 6 1 126 a 62 121 40 69 53 399 126 9 73 48 60 49 1 27c 25 3 16 5 20 7 20 2 9 37 4 4 4 1 2 52 30 386 76 72 29 11 loo 674 31 27 7 5 2 0 4 45 32 409 102 70 37 7 132 33 20 9 6 1 1 1 io 10 1 20 21 22 1 23 24 O127 2e 21 27 28 211 S 34 n66 329 35 :36 37 341 n75 57 86 a6 o 07 1,183 72 976 63 580 757 1 38 38 02 555 28 26 3 5 2 0 31 67 03 162 44 32 23 3 4 268 13 22 7 4 1 1 63 98 23 7 3 4 167 1 2 1 2 26 64 28 i4 6 40 22 10- 13 9 1 2 41 g 4 1 2 0 0 1 42 25 3 7 0 I 2 F.S 43 It 3 3 0. 5 0 22 456 2,247 2,217 1 200 639 1.191 625 113 111 60 32 60 Of the 29,011 guidln condits - cases, the Cortinsston receried Reports on the Sentencing Hearing for 22,943 (79 1%) reponls, 216 Inixed law cases (both guidHe of the Io"w 34 02 102 lj 03 1 30 TTL12 59 04 49 03 29 rnisin andIpreguK~ehne counts) were excluded offense Wyal (2,574), missin In addition, 2.700 cases were cnnnat history category (2,628). 0508 _ 1 01 03 01 02 1 01 19.947 1000 Of the 22,943 cases with such excluded due to one or rmore or cases with no analogous guideline (43) Diter""o ( vanables used in this table are Provided in AppendixA, SOURCE U S Sentencing Cornnmson. 1990 Data Foe,MONFY90 61. Figures in Table M are for reporting year 1989-90, as supplied in the Commission's annual report for 1990. See U.S. SENTENCING COMM'N, ANNUAL REPORT (1990) (tbl. M) [hereinafter 1990 ANNUAL REPORT]. HeinOnline -- 66 S. Cal. L. Rev. 195 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 Currently some offenders within groups A, B, and C receive community service as a condition of probation. However, there are significant differences among offenders within each of these groups. Group A consists of offenders who, although subject to imprisonment, are almost all presumptively to be sentenced to probation. Few in this group are sentenced to prison; the vast majority receive straight probation. At times some offenders in Group A must perform community service as a condition of probation, but it is unrealistic to assume that this punishment is imposed in lieu of a prison term. That is, community service for Group A offenders is an "add-on" to probation. Although this is no doubt appropriate in many instances, our purpose here is to explore ways to use community service as a substitute punishment for imprisonment. As it is currently constituted, Group A is not a fruitful pool from which to draw candidates. Group B is more problematic. Offenders in this group are subject to one of three sentences: (1) a sentence of imprisonment, (2) a sentence of probation with an alternative custodial arrangement that substitutes for imprisonment, or (3) a sentence of imprisonment followed by supervised release. With respect to the second option, the law allows substitution of various forms of "community confinement" (for example, residence in a community treatment center, a halfway house, or a similar residential facility, or home detention) for a day-for-day exchange of custody in a federal prison. Although precise figures were not available to us, interviews with probation officers indicate that such substitutes are used with some frequency. Group B offenders with serious substance abuse problems are often sent to community corrections facilities in lieu of prison, and many others are confined to home detention, requiring confinement at night or incarceration on weekends. Of course, not all Group B offenders receive such alternative custodial arrangements. An undetermined number receive sentences that include some period of actual confinement in prison. Others may receive a combination of alternative custody and community service. And of course, although alternative institutions of confinement may be more flexible than prison, current guideline policy does not allow for any other alternatives. If in fact the guidelines were amended to include a welldeveloped intermediate punishment option, such as community service, the sentencing options of prison and custodial placement may not be required for Group B offenders. In short, some portion of offenders in Group B would be strong candidates for community service in lieu of either prison or alternative HeinOnline -- 66 S. Cal. L. Rev. 196 1992-1993 1992] BETWEEN TWO EXTREMES confinement. Those in Group B who now do time in prison might be eligible to serve part or all of this incarceration through community service. In addition, others who are serving time in other types of commu- nity custodial facilities might be directed to community service. Group C offenders are required by law to serve at least one half of their lower guideline time in prison.6 2 The possibility that the other half of their term could be served through community service would expand the options available to courts for sentencing. What would happen if the presumptive sentences in each of these groups were altered? One answer is that we cannot tell with any precision." The annual reports provide great detail about the sentences of offenders in various offense levels (for instance, we know how many offenders convicted of a level seven crime do time)." But they do not describe sentencing profiles for each of the four groups as a whole. Ideally, we would like to know how offenders in each cell of the 258-cell grid were sentenced. But at present we do not have even a collective portrait of offenders sentenced within each of the four major groups. Nevertheless, it is possible to formulate some crude estimates. For instance, Table M identifies the number of offenders within each of the 258 cells in the guidelines grid. And the annual reports indicate what portion of offenders in each offense level (but not controlling for criminal histories) were sentenced to prison. It is therefore possible to make some educated guesses about the possible impact of shifting presumptive sentences.65 Keeping in mind the problem with our database, let us engage in a thought experiment to trace some of the salient consequences of adjust62. Telephone Interview with Loren A.N. Buddress, U.S. Probation Office, Northern District of California (Dec. 16, 1992). 63. One of the frustrating and puzzling features of the highly detailed annual reports of the Sentencing Commission is that after having created a 258-cell matrix for purposes of rational classification of sentences, the great bulk of the text falls back on more conventional descriptions of offenses in reporting sentencing practices. Thus despite the wealth of detail in the reports, one cannot get even rough approximations of the proportion of offenders in, say, Group B who were sentenced to some period in prison. One can get some sense of the portion of offenders in level six who received some period of incarceration, but it is impossible to know whether they were group A, B, or C. 64. Buddress, supra note 62. 65. Our purpose here is not to provide a definitive "downstream" assessment of the consequences of substituting one form of punishment for another, but simply to demonstrate that marginal adjustments within the least serious offenders' portion of the guidelines grid would yield substantial reduction in the number of offenders sent to prison. A much more detailed analysis employing the JUSSIM model, designed by Prof. Alfred Blumstein and his colleagues at Carnegie Mellon University, would be valuable here. HeinOnline -- 66 S. Cal. L. Rev. 197 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 ments to the grid to allow for imposition of community service as an intermediate punishment. Recall that our goal is to create an imaginary group that incorporates portions of groups B, C, and D and to suggest that it be designated as a group for which the presumptive sentence is community service as a condition of probation. Specifically, we suggest that this new cluster consists of those offenders who are currently in groups B and C and a new grouping we will label C'. This latter grouping consists of a narrow band defined on the lefthand side of the table by the level thirteen offense class and moving to the right just one step down from the existing lower boundary for Group C. Figures for our imaginary new group are suggested by sentencing figures supplied in the 1990 Annual Report, for the reporting year October 1, 1989, through September 30, 1990.66 There were 3335 offenders in Group B, 1603 offenders in Group C, and 617 in our new Group C', for a total of 5555. It is this group that constitutes the core for our new presumptive sentence of community service. Let us examine this group in more detail to determine what the implications of this reformulation are and what its effects would be. First, we wish to point out that the changes required to create this grouping would be relatively modest. Involved would be a reclassification of only thirteen of the forty-three offense categories and thirty of the grid's 258 cells. Although these adjustments are minor, their consequences could be substantial. They would affect the sentences of twentyeight percent of all federal offenders sentenced during 1990.67 Second, the adjustments would affect most significantly those offenders with lighter criminal histories. As indicated in Table M, the overwhelming majority of offenders in groups B, C, and C' do not have lengthy criminal histories; most are clustered in categories I, II, and III. For instance, eighty-eight percent, or 4870, of the 5555 offenders in groups B, C, and C' are found in criminal history categories I, II, or III, and fully seventy-five percent are in categories I and II. In short, it is possible to reach a large pool of offenders in groups B, C, and C' without dipping deeply into those with substantial criminal histories. 66. These data are based on only 19,947 of the 29,011 guideline sentences handled during this reporting year. Several reasons account for this. The Annual Report notes, "Of the 29,011 guideline cases, the Commission received Reports on the Sentencing Hearing for 22,943 (79.1%)." 1990 ANNUAL REPORT, supra note 61 (tbl. M). Of the 22,943 cases with such reports, 216 mixed-law cases (guidelines and preguidelines counts) were excluded. In addition, 2780 cases were excluded on account of one or more of the following conditions: missing offense level (2574), missing criminal history category (2638), or no analogous guideline (43). 67. See 1990 ANNUAL REPORT, supra note 61, at 78 (tbl. M). HeinOnline -- 66 S. Cal. L. Rev. 198 1992-1993 1992] BETWEEN TWO EXTREMES In the discussion that follows we restrict ourselves to considering the 4870 offenders in groups B, C, and C' with criminal histories in categories I, II, and III. This number constitutes twenty-four percent of all sentenced offenders for the year and twenty-nine percent of all those whose presumptive sentences included prison. Even by focusing on those with the shortest criminal histories, we would still significantly affect a substantial portion of the prison-bound population. Let us examine this grouping more closely. In particular, we need to know what portion of those in the pool actually receive terms in prison. With respect to Group B offenders, we can only make an educated guess. However, for groups C and C' we can make reasonably firm estimates. But even employing conservative guesses and estimates, the results are impressive. We are unable to provide firm estimates for Group B (and the second half of any sentence pronounced in Group C) because, although all offenders in this group are required to serve half their term in "prison," by law the court has authority to substitute a variety of forms of "custody" for prison.6" In many such instances, we suspect, such alternatives are more appropriate than community service, and therefore community service would not be a reasonable alternative. For instance, "custody" within Group B can include custody in a community corrections facility, home detention, and the like, and such sentences may be more appropriate than either straight community service or confinement in prison. Still, some portion of Group B offenders do time in prison and still others do time in more secure facilities than might be warranted if there were a well-administered community service punishment alternative. Let us assume that we design a presumptive intermediate punishment for all those Group B offenders in criminal history categories I through III. This would affect 3042 of 3535 offenders in Group B. Assuming that the program would divert at least twenty-five percent of them from prison or some other form of publicly financed custody,69 this would result in 760 offenders placed in the pool for a presumptive sentence of community service. Assume further that each of them would serve an average of four months in custody. Redirecting them to community service punishment would result in a savings of 251 person years 68. Buddress, supra note 62. 69. We treat all figures from the guidelines grid in those categories where groups B and C overlap to be in Group B. Thus we overestimate the number of offenders in Group B, and provide a conservative estimate for the number of offenders in the "group" (which actually combines B and C) who do time, and do time in prison rather than in some other form of custody (for example, a halfway house or a treatment facility). HeinOnline -- 66 S. Cal. L. Rev. 199 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 in prison. These figures are of course only educated guesses, but, we emphasize, they are "educated" by discussions with federal probation officers. Our estimates of the consequences of redirecting offenders in Group C (which here we treat as that portion of Group C not overlapping with B-that is, those defined by offense levels eleven and twelve in criminal history category I) and Group C' away from prison and into community service punishments rest on more solid footing. Currently by law, offenders in these categories must serve at least one half of their sentences in prison, and no portion of this custodial sentence can be served in any other type of facility. 70 Thus figures for the total number of offenders in these groups provide a more reliable indication of offenders who received prison terms.7 1 In the 1989-90 reporting year, there were 2220 offenders in groups C and C', almost all of whom were sentenced to some period of imprisonment; 1828 of them were in the three least serious criminal history categories (I, II, and III). Minimum prison time for this group is half the minimum sentence, which means half of from eight to eighteen months.72 Assume that on average this group served six months in prison. This translates into 914 person years in prison. Thus, redirecting them to intermediate punishments would yield substantial savings. Combining these projections and estimates about the size and significance of the target population for a presumptive community service punishment drawn from groups B, C, and C' (criminal history categories I through III), we found an initial pool of almost 5000 offenders. Even after draining off some unlikely candidates for community service, the remaining pool included nearly 2600 offenders (760 plus 1828). Expressed in still another way, community service has the potential to redirect as many as 1165 person years away from prison (914 + 251 = 1165) and some portion of them into community service. However considered, these figures are not trivial. They imply the involvement of thousands of offenders. They suggest substantial financial savings. They have significant implications for construction needs. Even if these illustrative figures overestimate the consequences by a factor of two or three, the figures are still substantial. 70. Buddress, supra note 62. 71. There are a handful of exceptions, those whose sentences deviated from their presumptive sentence and who did not receive a prison term. Here, too, we cannot tell precisely how many would have been involved, but it is a small handful. 72. See supra tbl. 12. HeinOnline -- 66 S. Cal. L. Rev. 200 1992-1993 1992] BETWEEN TWO EXTREMES B. COST IMPLICATIONS Although these figures are only rough guesses, it is nevertheless instructive to at least briefly explore their cost implications. The Administrative Office estimates that the average cost for keeping someone in custody is $1492 per month, whereas the Office of Probation estimates that the average cost of supervision of those doing community service is only $115 per month.7 3 Using these figures as crude cost comparisons, the marginal increase of community service suggested here would translate into a net annual savings of $19,250,460 ((12 X $1492 X 1165) (12 X $115 X 1165)). Even if these guesses are wildly off the mark, a much-reduced amount would still be substantial. When this result (based upon 1989-90 figures) is considered in light of projected increases in prison populations, the attraction of intermediate punishments increases proportionately. The Federal Bureau of Prisons reports that as of September 1991, the federal prisoner population was 65,000. By 1995, the BOP estimates, this figure will be 100,000, and by 1999 it is expected to reach 137,000. 71 This constitutes an increase of more than 100% within eight years, from 1991 to 1999. Accordingly, in light of these projections the estimates on the pool for candidates for community service and the resulting cost savings discussed earlier should be adjusted upward. Such an exercise makes community service all the more impressive. Although they are only guesses and crude estimates, the figures discussed here are conservative and err on the side of caution. Four factors account for this: First, we based our numbers on U.S. Sentencing Commission figures that are incomplete. In the 1989-90 reporting year there were 29,011 guidelines cases. Yet the Commission received reports on only 22,943 of them and had to exclude still others for various technical reasons.75 The 1990 Annual Report is based on an analysis of only 19,947, or eightyseven percent, of all sentenced guidelines cases. Assuming that these missing data were proportionately distributed across the sentencing grid, 73. These figures were provided by the Administrative Office of the U.S. Courts and reported by Loren Buddress, Chief Probation Officer of the Northern District, in Office Memorandum No. 91-32 (Sept. 6, 1991) (on file with authors). 74. These figures were supplied to us by Loren Buddress, Chief Probation Officer for the Northern District of California, who in turn drew them from materials provided by the Federal Bureau of Prisons. See Memorandum from Loren Buddress to the Authors (Jan. 23, 1992) (on file with authors). 75. See supra note 66. HeinOnline -- 66 S. Cal. L. Rev. 201 1992-1993 SOUTHERN CALIFORNA LAW REVIEW [Vol. 66:155 we could increase the estimates presented earlier accordingly. Since the bulk of all cases are in the less serious categories, these "missing" cases would add substantially to their numbers. Second, we treated all offenders in cells that overlapped in groups B and C as if they were in the less serious group, B. We did this because the Commission's report does not identify what portion of those in the overlapping cells were in each group. No doubt most of the offenders in these overlapping categories were in Group B and only a small portion in C, but since we could not easily determine this, we conservatively estimated that all of them were in Group B, the less serious offense grouping. As such we may have underestimated average prison sentence length. Third, our discussion focuses on offenders with lighter offense histories. Had we taken those from criminal history categories IV, V, and VI, especially those in Group C', and included them in our presumptive community service grouping, the projected "impact" of intermediate punishments would have increased accordingly. But even without such more refined calculating we can make our central point: In terms of numbers of offenders, the redirection of possible person years in prison, and the dollar cost involved, even minor adjustments in the sentencing guidelines to create a presumptive community service punishment could have substantial consequences. Fourth, any expanded use of community service would probably involve split sentences, in that offenders would do community service after a period of confinement. Such expanded use of split sentences would probably involve reductions in prison time and community service for offenders in crime classifications beyond level thirteen, our cutoff level, and thus might expand the pool of offenders for whom intermediate punishment could serve as a (partial) substitute for confinement. C. POLICY IMPLICATIONS In Between Probationand Prison, Morris and Tonry quite properly warn advocates of intermediate punishment not to "oversell" their product, and they propose that the intermediate sentence alternative be judged on its own terms, justified by traditional defenses of sentencing punishment generally.7 6 Above all, they warn against basing an argu76. MORRIS & TONRY, supra note 1, at 232-37. The authors write, The cost savings claims are suspect for at least three reasons. First, comparisons of per capita costs are misleading. The marginal cost of one additional prisoner for a prison system is slight .... Second, to be valid, comparisons must be based on something other than the annual average costs of operating nonprison and prison programs .... Third, HeinOnline -- 66 S. Cal. L. Rev. 202 1992-1993 19921 BETWEEN TWO EXTREMES ment in favor of intermediate punishments on the promise of windfall cost savings, and they chide those advocates of alternatives who make such claims. For instance, they point out that New York City's highly regarded community service program cost about as much as keeping the same number of offenders in jail and that the small number involved in the program did not produce even a trace of an impact on the city's jail population." They warn that unless the numbers involved are substantial, the projected costs are illusory. Slight reductions in numbers incarcerated will not result in substantial enough numbers to close institutions or negate pressures to build new institutions, but rather will result only in marginal savings associated with slight decreases in the population. We are mindful of these warnings and agree that community service must be justified on its own terms and in light of traditional justifications of punishment, not in terms of cost savings. We also realize that if community service is to yield meaningful savings, it must be used for a critical mass of offenders. The purpose of our discussion, however, is not to pretend to present a cost-benefit analysis. Rather, we seek to demonstrate that even minor changes could have substantial consequences in ways that matter. Our exercise does not require any suspension of disbelief. Nor does it require that we abandon deeply held views of sentencing or that we embrace novel new theories of punishment. It is an investigation of the consequences that might flow from only marginal adjustments to the existing sentencing scheme. And further, as we next argue, such shifts might actually bring federal sentencing more closely in line withnot further away from-national sentencing norms. We should emphasize that we are not making a general claim about the value of instituting a widespread system of community service in both federal and state jurisdictions. Our argument is limited to federal sentencing for three reasons. First, federal offenders constitute a better sample for experimenting with community service, because on the whole federal offenders are a "better class" of offender. They are different, significantly different, from the offenders in state and local systems. They are more likely to be better connected with their communities in any number of ways. They are more likely to be employed or in school. They are more likely to be living in a stable environment. They are betcost/benefit assessments must be adjusted to take into account the time each offender is subject to control .... If the average intensive supervised client serves 12 months ($4000) but would otherwise have served 3 months in prison ($3000), the intensive supervision program is more, not less, expensive. Id. at 233-34. 77. Id. at 233. HeinOnline -- 66 S. Cal. L. Rev. 203 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 ter educated. On the whole, they recidivate at a lower rate than offenders in state courts. In short, a higher proportion of them are likely to be better risks for community-based punishments than are offenders in state courts. For these reasons the federal courts seem to be an especially promising laboratory for experimentation with community service punishments. Second, the federal system is distinctive in its ability to marshal resources to support and maintain a well-functioning but low-cost community service program that depends on volunteer third parties for supervising offenders' work. Third, as a number of commentators have noted, although the U.S. Sentencing Commission was charged with "codifying" existing federal sentencing policy in order to reduce disparities, in fact it has promulgated a system that substantially increases the harshness of sentences, especially for offenders with no or short criminal histories who are guilty of lesser offenses. Before the guidelines were promulgated, a substantial portion of this group received straight probation. Now many of them are required to serve some time in prison.7 8 Indeed, it is this group of lesser offenders, serving short terms, that accounts for much of the rapid increase in federal prison population. And it is precisely this group that we have targeted for community service. To the extent that we are correct to focus on this group, marginal adjustment of the guidelines to make room for community service in the ways we outline would not require major changes in justifications for punishments or even traditional concerns about sentencing. Indeed, if state and local systems are to help serve as bench marks for the general harshness of punishments, these changes might be regarded as bringing federal sentences back into line with national norms. Whatever the case, because of the distinctive situation of federal offenders and recent federal sentencing practices, the proposed changes illustrated here would not constitute any radical departure from conventional thinking about punishment. We have one final observation. One does not have to embrace postmodernist philosophy to agree with its emphasis on the power of the word. We are, as it has been said, what we read. There are important implications in this lesson for proponents of intermediate punishments. If there is no word, there is no concept, and the practice is invisible. Indeed, this may be the single, simple lesson of Morris and Tonry's enor78. See, eg., id at 55-56. HeinOnline -- 66 S. Cal. L. Rev. 204 1992-1993 BETWEEN TWO EXTREMES 1992] mously important book, Between Probation and Prison.79 In the best sense of the term, their book is a public relations campaign to create a concept, intermediate punishments. If they are successful, all those who share in the enterprise will have invented a concept, enriched our public vocabulary, and in so doing satisfied a precondition for change. The challenge is to reconstruct our concept of punishment, shifting our intuitive understanding from a binary alternative ("slap 'em on the wrist" [probation] and "lock 'em up" [prison]) to a tripartite continuum consisting of probation, intermediate punishment, and prison. We embrace this campaign and suggest that the U.S. Sentencing Commission join it as well. We emphasize this not to make an esoteric hermeneutical point at the close of this Article, but because of our experience rummaging through the Sentencing Commission's annual reports. As detailed and helpful as they are, one can read them from cover to cover and never have an inkling that substantial numbers of federal offenders are required to perform hundreds-at times thousands-of hours of community service, punishment that some authorities regard as the functional equivalent of six to twelve months or longer in jail. This failure to report on such punishment, we believe, constitutes a serious omission. Indeed, the failure to provide a name for this punishment renders it wholly invisible in the annual reports. We are not complaining that such information is unobtainable. Indeed, the Sentencing Commission is well known for its accessibility. Rather, we are suggesting that the failure of its most widely read, widely available documents, containing hundreds of pages describing federal sentences in detail, to feature existing community service punishments borders on irresponsibility. And this failure certainly does nothing to stimulate analysis of existing community service or to promote discussion of its expanded use. There is a simple explanation for this failure to describe the nature and extent of community service punishments in the federal courts. It is that community service is not a sentence in its own right, like prison and probation. Rather, it is a "condition of probation," and as such it is lost among all the other conditions. Despite this, it is possible to define, describe, and report on this important option, and as such to present a more balanced portrait of federal sentencing. At a minimum such reporting would make current use of community service sentences more 79. See generally id. HeinOnline -- 66 S. Cal. L. Rev. 205 1992-1993 SOUTHERN CALIFORNIA LAW REVIEW [Vol. 66:155 visible. And, we suspect, more generally it would lead to greater consideration of this important option. Alternative sanctions have been used for a long time, but they have been underappreciated, in part because they have been invisible. We believe that if they-and their punitive features--are made more visible, community service will be better understood and more frequently used. At a minimum this means that community service sentences should be described more fully and that they should be treated as "punishments" in their own right and reported as such. Perhaps this means that judges should be able to sentence offenders directly to community service. Or perhaps it means only that sentencing reports should be broken down more carefully in order to more fully describe selected conditions of probation. Whatever the case, community service is an important punishment that remains all but invisible in the Commission's major documents describing sentencing in the federal courts. We believe this oversight should be corrected. VII. CONCLUSION Parts I through IV revealed that a high portion of offenders in the Northern District of California are sentenced to community service as a condition of probation, either in conjunction with a probationary sentence or in conjunction with some period of incarceration. On the whole the number of hours served is substantial, far greater than what conventional wisdom holds is practically possible. Furthermore, almost all participants successfully fulfill the conditions of their community service. Although there is slippage in supervision of participants, only a tiny fraction of those ordered to perform community service clearly fail to do so. Thus unlike many other such programs, the CSO program in the Northern District of California appears to be a substantial sanction and well administered. The analysis in Part V compared recidivism rates for offenders subjected to the three standard sentences-straight probation, probation plus community service, and incarceration-and found that community service does not have any appreciable independent effect on the likelihood of recidivism. That is, community service does not hurt and may help-and it is certainly cheaper than incarceration. Part VI considered the current U.S. sentencing guidelines and the current distribution of offenders sentenced under them in light of these findings. In particular it explored the implications of creating a new preHeinOnline -- 66 S. Cal. L. Rev. 206 1992-1993 1992] BETWEEN TWO EXTREMES sumptive intermediate sanction between probation and imprisonment. In this thought experiment, we found that by making only marginal adjustments in the existing sentencing structure, a substantial portion of offenders could be redirected away from short periods of imprisonment into a new intermediate punishment. We emphasize the exploratory nature of this analysis. It is, after all, based upon an assessment of only one community service sentencing program, and our examination of adjustments in the GuidelinesManual was only tentative and illustrative. Still, the investigation did reveal that community service sentences can be much more punitive than is widely believed, and that there are a substantial number of low-level offenders with no or short offense histories who are strong candidates for intermediate punishments. These findings are certainly encouraging. At a minimum they suggest the value of additional assessments of the efficacy of community service punishments in other districts and wider experimentation with community service punishments. We hope that the U.S. Sentencing Commission chooses to encourage such an agenda. HeinOnline -- 66 S. Cal. L. Rev. 207 1992-1993 HeinOnline -- 66 S. Cal. L. Rev. 208 1992-1993