National Academies Press: OpenBook

Research on Sentencing: The Search for Reform, Volume I (1983)

Chapter: 3 Structuring Sentencing Decisions

« Previous: 2 Determinants of Sentences
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 126
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 127
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 128
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 129
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 130
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 131
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 132
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 133
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 134
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 135
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 136
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 137
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 138
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 139
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 140
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 141
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 142
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 143
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 144
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 145
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 146
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 147
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 148
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 149
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 150
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 151
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 152
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 153
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 154
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 155
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 156
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 157
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 158
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 159
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 160
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 161
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 162
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 163
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 164
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 165
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 166
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 167
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 168
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 169
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 170
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 171
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 172
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 173
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 174
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 175
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 176
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 177
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 178
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 179
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 180
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 181
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 182
Suggested Citation:"3 Structuring Sentencing Decisions." National Research Council. 1983. Research on Sentencing: The Search for Reform, Volume I. Washington, DC: The National Academies Press. doi: 10.17226/100.
×
Page 183

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Structuring Sentencing Decisions American sentencing laws and practices underwent more extensive changes in the 197~1980 period than in any other 5-year period in American history. In this chapter we review the range of sentencing innovations adopted since 1960, examine the uses of research in the development of sentencing standards, and consider problems related to the imple- mentation and enforcement of sentencing policy innovations. THE VARIETY OF INNOVATIONS THE RETREAT FROM INDETERMINACY Twentieth-century American sentencing systems before 1976 are com- monly referred to as "indeterminate." Under indeterminate sentencing judges and parole boards have wide discretion in setting prison terms within broad statutory ranges for sentence length. Usually the actual length of a prison term remains unknown to a prisoner until the parole board authorizes release. This broad discretion and uncertainty was intended to facilitate individualized treatment for purposes of rehabil- itation. In the state of Washington during much of this century, for example, judges only decided who received prison sentences: they were required by law to impose the statutory maximum sentence on all offenders to be imprisoned, and the parole board decided how long any prisoner actually remained in prison. Under the indeterminate sentencing laws 126

Structuring Sentencing Decisions 127 of other states, judges have more influence. In Pennsylvania, for ex- ample, judges set both the maximum and the minimum sentence; the minimum cannot exceed half the maximum, and the parole board's authority is confined within the range set by the judge. Although the division of authority between judges and parole boards varies from state to state, the systems are indeterminate: discretion is broad and the duration of imprisonment remains unknown until the parole authorities actually release the prisoner. What most of the sentencing changes of recent years have in common is their rejection of this pattern in which judges and parole boards make ad hoc decisions, subject to few meaningful constraints, and are effec- tively immune from review. The narrowing of discretion and the intro- duction of greater certainty into sentencing have taken many different forms. Some jurisdictions abolished parole release entirely: California established detailed statutory standards for prison sentences, Minnesota established a detailed system of presumptive sentencing guidelines, and Maine established no standards at all. Other jurisdictions made other changes: Pennsylvania adopted both mandatory minimum sentences and sentencing guidelines while retaining parole release; and in Washington the parole board established parole guidelines, the judiciary established sentencing guidelines, and the legislature later created a commission charged to develop sentencing guidelines to take effect in 1984. CLASSIFICATION OF OFFENSES The federal criminal laws and those of most states developed adventi- tiously. New offenses were created and existing sentencing laws were amended in response to particular notorious events or social changes. Sentences authorized for particular offenses varied widely, reflecting the emotions, personalities, attitudes, and political imperatives of particular times (see National Commission on Reform of Federal Criminal Laws, 1970:1246 49~. As a result, anomalies characterize the sentencing laws of many jurisdictions. Offenses of comparable seriousness are often subject to substantially different penalties: federal law, for example, recently specified a maximum 20-year prison sentence for robbery of a federally insured bank and a 10-year maximum for robbery of a post office (see Senate Report 96-553:5, 1980), and offenses of different seriousness are often subject to the same maximum penalties. Consistent and evenhanded application of sentences is unlikely to be achieved in a system in which offenses and authorized sanctions are internally inconsistent and reflect no discernible logic. To introduce greater consistency to criminal law, the Model Penal Code developed

128 RESEARCH ON SENTENCING THE SEARCH FOR REFORM by the American Law Institute (1962) classified all felonies into three classes, each bearing a specific maximum sentence of fine, probation, or imprisonment. More than 30 states have enacted new criminal codes in recent years, and these have followed the Model Penal Code's lead in classifying felonies into a small number of categories, usually three or five. Every proposed federal criminal code, from Study Draft of a New Federal Criminal Code (National Commission on Reform of Fed- eral Criminal Laws, 1970) to S. 1630, which was approved by the Senate judiciary committee on November 18, 1981, has provided for classifi- cation of felonies. REASONS REQUIREMENTS AND PRESUMPTIONS Accountability is enhanced if decision makers must justify their deci- sions. In most contexts judges must give reasons for their decisions: this allows affected parties to understand the rationale for the decision and facilitates appellate review by providing appellate judges with a basis for knowing whether the trial judge applied the appropriate rule to the case under consideration and for evaluating the persuasiveness of the reasons for the decision. Until recently, sentencing decisions were anomalous. Judges were seldom required to give reasons for the sentences they imposed, and sentence appeals were not usually available in most jurisdictions. There were for all practical purposes no bases or procedures for holding judges accountable for sentencing. Now, however, reasons requirements have been proposed and enacted in a number of forms. Criminal codes sometimes provide that judges may not impose par- ticular sentences unless they give reasons for doing so. The study draft of the National Commission on Reform of Federal Criminal Laws (1970) provides that judges may impose minimum sentences (Sec. 3201) or maximum sentences beyond specified lengths (Sec. 3202) only if the court "shall set forth in detail" the reasons for its decision. In a variation the study draft established presumptions in favor of nonincarcerative sentences and parole release at first eligibility (Sees. 3101, 3402) along with criteria for determining when the presumptions are overcome. These provisions do not expressly require that reasons be given for decisions, but the effect is the same. A defendant who contests a sentence to incarceration or retention in prison would assert that the presumption has not been overcome; the sufficiency of the decision-maker's contrary judgment and the reasons for it would be the issues under consideration on appeal.

Structuring Sentencing Decisions 129 Most major sentencing innovations and proposals that provide criteria for decisions contain reasons requirements. The California Uniform Determinate Sentencing Law, for example, requires judges to give rea- sons for imposing a sentence not specified by the applicable statutory sections. Both sentencing guidelines systems and parole guidelines sys- tems typically require that decision makers provide reasons for decisions that do not adhere to the apparently applicable guidelines. While the commonsense case for reasons requirements in sentencing is straightforward and seldom contested, implementing these require- ments raises some practical questions. For example, requiring that judges give reasons for their sentencing decisions in every case may result in the trivialization of reasons; those given may become routine and me- chanical. Consequently, most reasons requirements obligate judges to give reasons only for exceptional decisions. A related practical question concerns the form for providing reasons—whether decision makers should be provided checklists that contain possible reasons for decisions or whether they be required to write out reasons of their own devising. (See Zeisel and Diamond t1977] for discussion of some of the difficulties involved in making reasons requirements meaningful.) SENTENCING INSTITUTES American trial judges work alone. Under indeterminate sentencing laws they have broad statutory flexibility, and they are typically not required to account for their decisions. Partly to facilitate communication among judges, the U.S. Congress in 1958 authorized sentencing institutes for the federal judiciary, and similar sentencing institutes have been held by many states. At these institutes judges discuss sentencing develop- ments and often engage in simulated sentencing exercises; they then discuss their respective reactions to the simulated cases and the sentences they would have imposed. The premises of sentencing institutes are that they familiarize judges with the views of their colleagues, thus allowing them to learn whether their own attitudes and opinions are consistent with general patterns; that newly appointed judges benefit from the accumulated experience and "going rates" of their colleagues; and that all participating judges become more self-conscious in sentencing. Every major criminal law reform body in recent decades has declared its sup- port for sentencing institutes. Widespread participation by judges in institutes over the last two decades may have increased their awareness of the dilemmas sentencing poses and their receptivity to proposals for reform.

130 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM SENTENCING COUNCILS Sentencing councils are groups of judges who meet on a regular basis to discuss sentences for pending cases. Established in several federal district courts beginning in 1960, they were the first modern institutional innovation expressly aimed at reducing sentencing disparity. Each par- ticipating judge reviews presentence reports and prepares sentencing recommendations before attending the council meeting at which the recommendations are discussed. The recommendations are advisory, and the responsible judge in a case may disregard the recommendations. One of the rationales for sentencing councils is that the exchanges of views would sometimes cause judges to reconsider their initial sentencing recommendations because of incongruity with the recommendations of their colleagues. Several accounts indicate that the initial recommen- dations of judges do differ from their ultimate sentences in one-third or more of the cases that come before the councils (Levin, 1966:511; Phil- lips, 1980:36~. However, several major evaluations of sentencing coun- cils find evidence that sentencing councils do not eliminate substantial sentence disparity (Diamond and Zeisel, 1975; Phillips, 1980~. Like sentencing institutes, sentencing councils have been endorsed by every major criminal law reform body of recent years. ABOLITION OR REGULATION OF PLEA NEGOTIATION The legitimacy of the American dependence on plea negotiation as a primary method of case disposition has long been questioned. The Pres- ident's Crime Commission (1967) and the American Bar Association Task Force on Sentencing Alternatives and Procedures (1980) both re- viewed objections to plea bargaining and proposed methods to bring it into the open and to subject it to regulation. The National Advisory Commission on Criminal Justice Standards and Goals (1973:46) rec- ommended abolition of plea bargaining. At least nine major objections to plea bargaining have been asserted. First, until the 1970s, most plea negotiation was hypocritical: a majority of convictions resulted from guilty pleas, but defendants were required to deny in court that they had been offered inducements to plead guilty (President's Crime Commission, 1967:9~. Second, serious principled ob- jections can be made to the propriety of offering defendants inducements to waive their constitutional right to trial: in effect, prosecutors threaten to punish the assertion of trial rights by withholding from defendants benefits they would receive if they pleaded guilty. Third, there is always a risk that an innocent defendant will plead guilty from fear of being

Structuring Sentencing Decisions 131 sentenced more harshly if convicted after trial: this risk is especially great when the defendant is offered a probationary sentence for pleading guilty and is threatened with going to prison if convicted. Fourth, plea- bargaining conflicts with the public interest in soundly administered Justice: presentence reports and other investigations of the defendant are generally only available after conviction and are not available at the plea-bargaining stage. Thus the judge must often approve or disapprove a proposed bargain without adequate information about the defendant. Fifth, plea bargaining can result in excessive leniency for professional criminals who are familiar with the courts' operations and are repre- sented by courtroom regulars: unsophisticated minor offenders who are unaware of the manipulative benefits of plea bargaining may simply plead guilty to the original charges and be treated relatively harshly. Sixth, institutionalized plea bargaining undermines the substantive crim- inal law: defendants plead guilty not to the offense they committed but to some lesser offense that has been negotiated. One often cannot know from the offense of conviction what offense was actually committed. Seventh, plea bargaining effectively shifts power to set sentences from judges to prosecutors. Eighth, plea bargaining reduces judges' aware- ness of investigations and arrests and thereby lessens their knowledge of police practices and their influence on them. Ninth, by merging the conviction and sanctioning decisions, plea bargaining increases the risk that each decision will not receive the separate attention that it should. Although the moral and practical case against plea bargaining is quite strong, several arguments have been made in its defense. First, it is sometimes said that the criminal courts would be grossly overburdened if plea bargains did not induce most defendants to plead guilty. ~ Supreme Court Chief Justice Warren Burger has observed (Santobello v. New York 404 U.S. 257, 260 t19713~: [Plea bargaining] is an essential component of the administration of justice. Properly administered, it is to be encouraged. If every criminal charge were subjected to a full-scale trial, the States and the Federal Government would need to multiply by many times the number of judges and court facilities. Second, plea bargaining relieves defendants and prosecutors of the un- certainties of trial. Third, it can be used to mitigate the harshness of mandatory sentencing laws that prescribe punishments more severe than ~ For the contrary argument and supporting evidence that the majority of defendants would continue to plead guilty without plea bargains, see Feeley (1979:Ch. 8), Heumann (1978), and Rubinstein et al. (1980~.

experience.) 132 RESEARCH ON SENTENCING THE SEARCH FOR REFORM a prosecutor believes are warranted. Fourth, plea bargaining in which leniency is exchanged for information, assistance, and testimony in other prosecutions can serve important law enforcement interests. Fifth, plea bargaining can be viewed as a form of dispute resolution in which the parties compromise their differences and thereby achieve a more mu- tually satisfactory resolution than if the parties were unfailingly adver- sarial. Sixth, plea bargaining permits prosecutors to achieve convictions in cases in which evidentiary or procedural problems might otherwise result in acquittals. Efforts have been made to "abolish" plea bargaining in full, or in part, in several jurisdictions. In Wayne County, Michigan, for example, the prosecutor forbade plea bargaining by his assistants in cases in which a firearm was used in the course of a felony (see Heumann and Loftin, 1979~. The most dramatic plea-bargaining ban occurred in Alaska. Ef- fective August 15, 1975, the attorney general banned plea bargaining in all its forms (Rubinstein et al., 19804.2 For much the same reasons that some prosecutors have attempted to ban plea bargaining, others have attempted to regulate it by establishing internal office policies governing charge and sentence bargains and sen- tence recommendations (e.g., Kuh, 1975a,b). Plea bargaining has also been regulated as part of more general efforts to establish and enforce office policies and systems of managerial controls (see Eisenstein and Jacob, 1977; Jacoby, 1980~. Efforts have been made in a number of jurisdictions to institutionalize plea bargaining. In one series of projects supported by the National Institute of Justice, scheduled plea confer- ences included the judge, the lawyers, and—if they wished to partici- pate—the defendant, the victim, and the involved police officer. (See Kerstetter and Heinz t1979] for a report on the Dade County, Florida, MANDATORY MINIMUM SENTENCES Between 1977 and 1980, mandatory minimum sentencing laws were adopted in 27 states and were under consideration in at least 14 others (U.S. Department of Justice, 1980a). Mandatory minimum sentencing laws take several forms but have as their common characteristic the statutory directive that convicted defendants whose offenses and prior 2 Because public prosecution is generally organized at local levels, most state attorneys general lack authority to promulgate such a ban and the means to enforce it. In Alaska, however, public prosecution is organized on a statewide basis.

Structuring Sentencing Decisions 133 record fall within specified categories be sentenced to a prison term of not less than a specified period of years. Under some mandatory sen- tencing laws, judges retain the option to impose a nonincarcerative sentence but must impose a sentence of at least the mandatory minimum term on those whom they send to prison. Other laws expressly preclude nonincarcerative sentencing options and direct that all persons convicted of the designated offense receive a term of imprisonment of not less than the mandatory minimum term. Massachusetts's Bartley-Fox law, for example, provides that all persons convicted of unlawfully carrying a firearm be imprisoned for a term not less than 1 year. Other variants are more complicated. A Michigan law enacted in 1977 requires that persons convicted of the use of a firearm in a felony receive a prison sentence of not less than 1 year; both the firearms charge and the underlying felony charge have to be either pled or proved, and the minimum sentence law does not apply if either charge is not proved. STATUTORY DETERMINATE SENTENCING LAWS (ABOLITION OF PAROLE) Determinate sentencing laws take two forms. In the first, discussed in this section, a legislature specifies the presumptive sentences or sentence ranges. In the second, discussed in the next section, a legislature sets the general outlines of the sentencing system and the standards for sentences but delegates the responsibility for developing guidelines to . . . a sentencing commission. Determinate sentencing exists in those jurisdictions in which the lengths of prison sentences can be determined, assuming the prisoner's good conduct in prison,3 at the time the judge announces the sentence; the release date is not determined later, by a parole-type agency.4 By that criterion, at least nine states have enacted determinate sentencing laws: 3 Most state prisons operate good-time systems under which the length of any prison sentence can be reduced as a reward for good behavior while in prison. Throughout this report, discussion of the lengths of prison sentences should include the qualification "assuming good behavior in prison and that good-time credits are not administratively reduced or increased." 4 A somewhat different concept of determinacy includes parole systems under which release dates are set in the early months of confinement but excludes those jurisdictions that have not established relatively detailed standards for sentencing and parole decisions. (See van Hirsch and Hanrahan t1979:25-35] on the desirability and practicality of parole systems that set release dates early; on definitions of determinacy, see von Hirsch and Hanrahan t1981:29~2964.)

134 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM Alaska, California, Colorado, Indiana, Illinois, Maine, Minnesota, New Mexico, and North Carolina. The statutory determinate sentencing sys- tems in those states vary widely. At one extreme, Maine abolished its parole board in 1975 when it enacted a comprehensive criminal code based on the Model Penal Code. Except for the maximum sanctions specified for each class of felonies, no criteria were provided to guide judicial sentencing decisions. Thus Maine judges retain the substantial unregulated discretion that judges typically have had under indeterminate sentencing systems. Because of the abolition of parole, however, prisoners can predict at sentencing when they will be released (see Zarr, 19764. California's Uniform De- terminate Sentencing Law, at the other extreme, abolished parole re- lease for most prisoners and enacted detailed statutory sentencing stan- dards. That law provides that, when sentencing offenders to prison, judges choose one of three specified sentences as the "base term" for persons convicted of a particular offense (for example, 2, 3, or 5 years for robbery). The middle term is to be imposed in an ordinary case. The higher or lower term may be imposed in cases with aggravating or mitigating circumstances. In addition, the California law provides for a variety of increments (called "enhancements") that can be added to the base term if various prior-record factors (primarily prior incarcerations) or aggravating offense circumstances are alleged and proved. Among the specified aggravating circumstances are use of a firearm or other dangerous weapon, serious bodily injury, and major property loss or damage. The other existing determinate sentencing laws range between those of Maine and California. They provide more guidance to judges than the former but less than the latter. Several criticisms of such laws have been offered. Zimring (1976) notes that sentencing is especially vulnerable to being politicized when detailed sentence criteria are placed before a legislature: introduction of amendments to increase sentences is politically effective evidence of a legislator's devotion to law and order. However carefully developed proposed statutory sentencing criteria may be, they can be altered simply by changing a number in a committee room or even on the floor of the legislature. The many bills introduced in the California legislature to increase sentence severity since adoption of the original determinate sentencing law provide some support for Zimring's observation. A related objection is that legislatures are not institutionally suited to the development and review of detailed sentencing policy. Sentencing is but one among many subjects competing for legislators' time and attention, and they lack special expertise in the subject. The legislative

Structuring Sentencing Decisions 135 floor is a poor place for consideration of any subject that requires de- tailed, sustained attention. While legislatures are well suited to the adop- tion and enunciation of broad normative choices, they are institutionally less capable of the detailed policy making and gradual refinement of policy over time. For these reasons, legislatures have frequently dele- gated rule-making authority over technical subjects, such as regulation of securities and public utilities and, recently, sentencing, to adminis- trative agencies. A third objection is that statutory sentence criteria tend to shift dis- cretion from judges to prosecutors. When, as under California law, the offense of conviction and any pled and proved enhancements determine the applicable sentence, some sentencing power may be shifted from judges and placed in the hands of the lawyers participating in the plea- . . bargaining process. PRESUMPTIVE/PRESCRIPTIVE SENTENCING GUIDELINES AND SENTENCING COMMISSIONS Existing sentencing guideline systems differ on two important dimen- sions: their legal authority and the influence and role of empirical in- formation on past sentencing practices in generating the guideline sen- tences. Guidelines are presumptive or voluntary, depending on their legal force. Guidelines are also labeled "descriptive" or "prescriptive," depending on whether they are designed largely to articulate and codify past sentencing practices (descriptive) or are focused primarily on de- veloping new sentencing policies (prescriptive). Although there are four possible combinations of these dimensions, two principal combinations are actually found: presumptive/prescriptive guidelines, illustrated by those in Minnesota, Pennsylvania, and Washington; and voluntary/de- scriptive guidelines, illustrated by those in Denver and Massachusetts.5 In three jurisdictions, Minnesota, Pennsylvania, and Washington, the legislatures have delegated authority for developing detailed sentencing criteria to sentencing commissions. The sentencing commissions are charged to develop presumptive sentencing guidelines. Judges are ex- pected to impose sentences recommended by the guidelines in ordinary 5 Depending on the dimension of particular interest in each context, we may refer to sentencing guidelines simply as presumptive or voluntary or as "descriptive" or "pre- scriptive." A1SO7 "descriptive" and "prescriptive" are in quotation marks because these terms are widely used but not literally accurate: regardless of their origins, all guidelines are statements prescribing policy and most have used data describing past practice in some way in their development.

136 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM cases and to provide reasons for sentences that do not follow the guide- lines. Minnesota's is a determinate sentencing system. Parole release has been abolished, and the sentencing guidelines specify "real-time" sentences. Pennsylvania's guidelines are indeterminate and affect only the minimum sentence to be served before parole eligibility: the parole board continues to decide when prisoners are released. The Washington guidelines will establish determinate sentences for offenders convicted after July 1, 1984, while the parole board will establish release dates for offenders convicted prior to that date until it ceases to exist on July 1, 1988. These presumptive guidelines have substantial legal authority, deriv- ing from the sentencing commission's legislative mandate. The guideline sentence must be imposed or an explanation provided as to why some other sentence was imposed. Both defendants and prosecutors have the right to have the sufficiency of that explanation reviewed by an appellate court. (By contrast, voluntary sentencing guidelines are hortatory and create no defendants' rights; noncompliance by judges does not give rise to a right of appeal. Voluntary guidelines have thus far been initiated by judges and not legislatures.) The operation of a presumptive guidelines system can be illustrated by the Minnesota guidelines. Table 3-1 shows the sentencing grid of the Minnesota Sentencing Guidelines Commission. The left-hand column lists the 10 categories of criminal offenses, and the top row lists the 7 categories of criminal offenders. The offense categories include all com- monly occurring criminal offenses, ranked in order of their seriousness. The offender categories are based on a defendant's "criminal history" (prior record), using a weighted scoring system developed by the com- . . mlsslon. The sentence for any defendant is found by first determining the offense severity and criminal history ranking and then consulting the cell of the sentencing grid in the applicable row and column. The cells above the bold block line call for sentences other than state imprison- ment: the numbers in these cells represent the prescribed lengths of stayed (i.e., unexecuted) sentences.6 Each cell below the bold line con- 6 In general, a stayed sentence is one that is not carried out. Two types of stayed sentences are permitted under Minnesota law: a stay of imposition and a stay of execution. A stay of imposition means the defendant is convicted of a felony, is given a probationary sentence (that may include up to 12 months in jail), and upon completion of the sentence the felony conviction is reduced to a misdemeanor on the record. If the offender violates probation, a felony prison term may be imposed. A stay of execution means that the defendant is convicted, a felony conviction is placed on the offender's record, the offender is placed on probation, and the felony prison term that is pronounced but not carried out may be executed if probation is violated.

Structuring Sentencing Decisions Structuring Sentencing Decisions 1 f - Dot J—1 lVllllllC;bUL~ OO11~11~1118 Dilly. OO11~11~111` ~ ~~ ~ ~— .: : ~ ~ 1 ~ T: ~~ ~ SEVERITY LEVELS OF CC~'VICTION OFFENSE to 1 2 1 3 4 5 6 or more 2l 24 r 20 1 23 Lo 22 27 32 21-23 25-29 30-34 Unauthorized Use of Motor Vehicle Possession of Marijuana Theft-related Crimes ($150-$2500) Sale of Marijuana T 1 12* 1 12* 12* 12* 1 15 18 II 12* 14 17 Theft Crimes ($150-$2500) III 12* 12* 13 16 19 Burglary—Felony Intent Receiving Stolen Goods IV ($150-$2500) 15 18 21 25 24-26 32 30-34 41 37 45 Simple Robbery V 18 23 26 24 32 23-25 30-34 43 4105 27 11 293-°31 1 1 ~L~ 38 36~0 44 42 46 65 6~70 46 43-49 54 50-58 81 75-87 54 50-58 65 60-70 98 90-104 Assault, 2nd Degree VI 21 Aggravated Robbery VII 41 38-44 49 45-53 Assault, 1st Degree Criminal Sexual Conduct? VIII 1st Degree 54 50-58 65 60-70 76 71-81 95 89-101 113 106-120 132 12~140 Murder, 3rd Degree IX 97 90100 116 . 111-121 119 116-122 140 133-147 127 124-130 162 153-171 149 143-155 203 192-214 176 168-184 243 231-255 205 195-215 284 270-298

138 RESEARCH ON SENTENCING THE SEARCH FOR REFORM The guidelines of the Pennsylvania Commission on Sentencing are similar in concept but provide much broader ranges. The Washington state sentencing commission statute was passed in April 1981; the com- mission is in the preliminary stages of developing its guidelines. VOLUNTARY/DESCRIPTIVE SENTENCING GUIDELINES The first sentencing guidelines system was established in Denver in 1976 (see Wilkins et al., 1978~. The Denver guidelines resulted from an effort to apply to sentencing the research experience and technology developed in establishing the U.S. Parole Commission's parole guidelines (see the next section). The premises of the Denver guidelines feasibility project stated that "the gradual build-up of case-by-case decisions results in the incremental development of a sentencing policy" and that an empirically based guidelines system "takes advantage of, and incorporates, the col- lective wisdom of experienced and capable sentencing judges by devel- oping representations of underlying court policies" (Wilkins et al., 1978:xlii, 10~. The researchers attempted to develop a mathematical model of the determinants of sentence outcomes in Denver (and also in Vermont, where the effort was abandoned before guidelines were implemented) as the basis for guideline formulation. Various models of sentences in Denver were developed based on the results of multiple regression analy- ses applied to data on already sentenced cases. The models were then tested on a validation sample, and voluntary/descriptive sentencing guidelines were developed that ostensibly embodied the existing latent sentencing policies of the court (see Wilkins et al., 1978~. The initial Denver guidelines were expressed as a matrix. Separate matrices were developed for three felony classes and for each misde- meanor class. The offense severity score resulted from efforts to scale the severity of offenses within each statutory offense class, and the offender scores were based on offender variables found to explain sig- nificant amounts of variation in sentences. Judicial compliance with the guidelines was voluntary, and noncompliance did not give rise to rights of appeal by either defendants or prosecutors. The Law Enforcement Assistance Administration, which had pro- vided support for the Denver feasibility study, supported the develop- ment of a second generation of judicial sentencing guidelines in Cook County (Chicago), Illinois; Essex County (Newark), New Jersey; and Maricopa County (Phoenix), Arizona (see Kress, 1980~. Similar guide- lines were developed with local funds in Philadelphia. A third generation of federally funded "descriptive" sentencing guide- lines has been developed in demographically diverse counties of Mary-

Structuring Sentencing Decisions 139 land and Florida in order to assess the impact of uniform guidelines in different regions of a single state. The National Institute of Justice has supported an ongoing evaluation (by Abt Associates) of the construction and impact of those guidelines. Voluntary guidelines need not necessarily be based on statistical ef- forts to model past sentencing practices. Judges in a particular court could agree to certain normative propositions about the purposes of sentencing and adopt voluntary/"prescriptive" guidelines reflecting that agreement. To date, however, most sentencing guidelines projects have followed the general pattern of the Denver guidelines: collect data on a sample of disposed cases; perform multivariate analyses to develop a model of the independent variables, with their respective weights, that influence sentences; and design a guideline format by which sentences in pending cases can be calculated on the basis of the model. Such processes to generate voluntary/"descriptive" sentencing guidelines have been undertaken by the judiciary at the state level in Michigan, Mas- sachusetts, and New Jersey. State-level sentencing guidelines for se- lected offenses have been established in several states, including Alaska and Washington. Voluntary/"descriptive" sentencing guidelines projects at the local level have been initiated in at least 11 states (see Criminal Courts Technical Assistance Project, 1980~. PAROLE GUIDELINES Parole guidelines have been adopted by the U.S. Parole Commission and by the parole boards of several states, including Florida, Georgia, Maryland, Minnesota, New York, Oklahoma, Oregon, and Washington. The first parole guidelines system was initiated by the (then) U.S. Board of Parole in October 1972 as part of a pilot project to test the feasibility of regionalization of the board's work. The matrix-type parole guidelines developed for use in the northeast region were subsequently modified and in March 1976 were mandated by the Parole Commission and Reorganization Act for use throughout the federal system. Most of the state parole guidelines systems are patterned after the federal guide- lines. The U.S. parole guidelines emanated from the Parole Decision-Mak- ing Project of the National Council on Crime and Delinquency Research Center, in collaboration with the U.S. Board of Parole. One phase of that project was an attempt to identify the weights given by decision makers to various criteria in the parole decision. Research showed that decision makers' primary concerns were the severity of the offense, the prisoner's parole prognosis, and the prisoner's institutional behavior and

140 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM "that a parole board's decisions could be predicted fairly accurately by knowledge of its ratings on these three factors" (Gottfredson et al., 1978~. There have been spirited political conflicts in various jurisdictions between proponents of parole guidelines and proponents of parole ab- olition. At the federal level, for example, bills to abolish parole and establish a sentencing commission and presumptive sentencing guide- lines have twice been approved by the Senate judiciary committee. The bills developed by the House judiciary committee, in contrast, have consistently contained provisions calling for the retention of the Parole Commission and parole guidelines. Minnesota operated a parole guide- lines system for almost 5 years, only to abolish it when the state's sen- tencing guidelines system took effect on May 1, 1980. Washington first established parole guidelines in 1976, but, as we noted above, recent legislation created a sentencing commission charged to develop pre- sumptive sentencing guidelines and provides for abolition of parole. Supporters of parole guidelines assert that well-organized, well-managed parole boards can achieve greater policy consistency than judges because parole boards are small, continuing collegial bodies; that parole boards applying consistent policies can reduce sentencing disparity by compen- sating for the disparate prison sentences imposed by dozens of judges throughout a jurisdiction, thereby in effect performing an appellate sentence review function; that parole boards can act as prison population control mechanisms by speeding releases when necessary to relieve pop- ulation pressures; that parole boards, by being less visible, are less subject to public pressures and are freer to take risks in releasing in- mates; and that parole guidelines are likelier to be followed than criteria for judicial sentencing because the status of hearing examiners as insti- tutional employees makes them more amenable to discipline and man- agerial controls and their decisions more easily subject to review than are those of judges. Critics of parole guidelines note that they cannot supplant promul- gation of standards for judicial decisions because they have no relevance to the "in/out" decisions (whether or not to imprison); that they per- petuate a "Rube Goldberg" system of sentencing in which parole boards reduce the lengths of sentences that judges have increased in expectation that parole boards will reduce them; that sentencing is a decision of immense symbolic importance and is more appropriately a judicial than an administrative decision; and that because defendants are entitled to greater procedural protections in court than at parole hearings, the factual quality of the evidence considered in making the sentencing decision is likely to be of higher quality.

Structuring Sentencing Decisions APPELLATE SENTENCE REVIEW 141 In virtually all common law jurisdictions except the United States, ap- pellate review of sentences has been the principal method used to de- velop principles for and achieve consistency in sentencing. Approxi- mately 40 percent of all criminal appeals heard by the English Court of Appeal between 1956 and 1976 resulted in affirmance of the defendant's conviction but variation in the sentence imposed (Advisory Council on the Penal System, 1978, App. H:202; also see Cross, 1975; Thomas, 19794. Appellate review of sentences is also extensively used in the Scandinavian countries (Andenaes, 1983) and West Germany (Weigend, 1983~. In all of these countries the standards for sentencing that are the bases for review derive from the accumulation over time of a body of case law from which a national sentencing tariff has been extracted and gradually refined. Numerous U.S. law reform bodies have proposed the adoption and invigoration of sentence appeal in this country.7 A sizable minority of American states have long had systems of appellate sentence review, established by statute, by constitutional provisions, or by judicial inter- pretation of state law, but in most cases they appear to be ineffectual. The report of the Criminal Justice Standards Project of the American Bar Association (1980: 18.193-18.197) notes: A number of careful studies have examined the operation of state appellate review systems under which an offender can appeal a sentence as excessively severe.... Without serious exception, these studies found that appellate review had little more than a negligible impact, generally providing a remedy only in egregious cases but not capable of developing clearly articulated criteria or standards by which to guide future sentencing decisions. Several organizational factors appear to contribute to the limited im- portance of appellate sentence review in the United States. In Massa- chusetts and Connecticut, the review divisions are composed of trial court judges sitting 2(~25 days per year. Their infrequent sittings and limited organizational resources do not facilitate the development of policy. Moreover, since the division members are trial court judges, it is not clear that their colleagues would expect them to develop policies. 7 A partial list includes the American Bar Association's Criminal Justice Standards Project (1980), the National Advisory Commission on Criminal Justice Standards and Goals (1973:116-118), the National Commission on Reform of Federal Criminal Laws (1970, Sec. 1291), and the President's Commission on Law Enforcement and the Admin- istration of Justice (1967:145-146~.

142 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM Even if the review divisions constructed sentencing standards, it is un- clear how they would be communicated to other judges. In Massachu- setts the review division gives no reasons for its decisions, and its de- cisions are communicated only to the trial court judge who sentenced the offender. In Connecticut reasons are supplied by the division and may be published, but in practice the reasons are very general and rarely . . appear in print. This ineffectiveness should not be surprising. Under the indeterminate sentencing systems that characterized U.S. state systems for most of this century, the prison sentences that judges imposed were often nominal: the parole boards had primary responsibility for deciding how long peo- ple remained in prison. There was little reason for appellate judges to interfere with sentences that would be adjusted by a parole board in any event, especially when the effect of doing so would be to increase the appellate work load without benefit of correspondingly increased resources or personnel. Plea bargaining may also have impeded the development of appellate sentence review. In sentence bargains, the defendant expressly agrees to the sentence received and is not well placed to later object to that sentence. Judges may believe that de- fendants who have had charge bargains, similarly, are not entitled to object to any sentence that can be imposed for the offense to which the defendant pled guilty. Finally, sentencing was not subject to established criteria, except for the maximum sentences authorized by statute (or in some cases the common law), and there were, accordingly, no standards that an appellate judge could invoke to determine whether a particular sentence was excessive in length or otherwise inappropriate. The prospects for meaningful appellate sentence review may be greater under some determinate sentencing systems than they were under in- determinate sentencing. Determinate systems have as one of their aims increased accountability for the sentences judges impose. Appellate re- view may be facilitated by the combination of published detailed stan- dards for sentences and the requirements that judges give written reasons for imposing sentences that deviate from the apparently applicable standards. Together these new rules may provide substantial bases for assessing the appropriateness of appealed sentences. Under the Min- nesota guidelines system, for example, judges are authorized to depart from the guidelines only when "substantial and compelling circumstan- ces" are present and must provide a written statement of the reasons for doing so. Minnesota law (1978 Laws, Ch. 723, Sec. 11) enjoins the Minnesota supreme court to . . . review the sentence imposed or stayed to determine whether the sentence is inconsistent with statutory requirements, unreasonable, inappropriate, ex-

Structuring Sentencing Decisions 143 cessive, unjustifiably disparate, or not warranted by the findings of fact issued by the district court. There are a number of controversial questions associated with ap- pellate sentence review. Should both prosecutors and defendants be entitled to appeal, or only defendants? Should sentence appeals be available in cases in which a sentence was explicitly or implicitly ne- gotiated? Should an appellate court's power be limited to granting or denying the relief sought by the appealing party or should the court be able to increase the sentences of defendants who appeal (or decrease sentences appealed by the prosecutor)? Should appeals be heard by regular appellate courts, by specially constituted sentence appeals courts, or by panels of trial judges (as is now done in Massachusetts and Mary- land)? In summary, the changes in sentencing described in this section vary substantially in form and ambition. Some, like sentencing councils and institutes, seem primarily aimed at making decision making more self- conscious. Others, like classification of offenses, reasons requirements, presumptions, and voluntary/"descriptive" sentencing guidelines, are modest efforts to encourage greater consistency. Still others—bans on plea bargaining and abolition of parole are radical breaks with tradition that signal a fundamental reappraisal of the criminal justice process. The other changes—statutory determinate sentencing, presumptive/ "prescriptive" sentencing guidelines, parole guidelines, mandatory min- imum sentences, and appellate sentence review—can be seen as efforts to establish general sentencing criteria and to work toward sentencing that is evenhanded, accountable, and reasonably consistent. Despite apparent agreement on the need to increase consistency and accountability in sentencing, however, the contemporary sentencing re- form movement, in contrast to earlier movements in U.S. history, is characterized by sharp disagreements over the normative goals of sen- tencing (or the justifying aims of sanctions) and the amount of punish- ment appropriate for various situations (see Rothman, 1981~. In addi- tion, there are tensions between attainment of the normative goals of a reform and the interests and functional goals of people and organi- zations that constitute the criminal justice system. These conflicts and tensions have shaped contemporary reform efforts. DEVELOPING GUIDELINES: MODELING AND DATA ISSUES Social science research methods have been extensively used in the de- velopment of sentencing and parole guidelines. The three principal bod-

144 RESEARCH ON SENTENCING THE SEARCH FOR REFORM ies of research concern: modeling past practices to guide or inform formulation of future standards; developing devices to predict recidi- vism; and projecting the impact of proposed changes. In principle, the methods involved are equally pertinent to legislative consideration of statutory presumptive sentencing laws, prosecutorial consideration of charging and plea-bargaining policies, and judicial sentencing commis- sion and parole board consideration of guidelines. In practice, however, the methods have been most commonly used in parole and sentencing guidelines projects. The rest of this chapter reviews a series of research and policy issues that must be addressed when empirical analyses of past practices are used in establishing standards. Sentencing guidelines are treated as the archetypal case for illustrative purposes, but the discussion is generally applicable to other forms of sentencing standards. MODELING PAST PRACTICE The first empirically based sentencing standards were the U.S. parole guidelines. Although social scientists have long been developing "ex- perience tables" that set out base-expectancy recidivism rates, the novel aspect of the Parole Decision-Making Project was that it used estimates of recidivism risk from experience tables as one of the factors in making parole release decisions (Gottfredson et al., 1975; see also Hoffman, 1975; Hoffman and Gottfredson, 1973~. Building on that parole guidelines experience, a sentencing guidelines project was established in Denver (Wilkins et al., 1978) to test the feasibility of developing similar, empirically based guidelines for sen- tencing. The researchers recognized that sentencing is a more complex process than parole decision making: parole boards decide whether to release prisoners, while judges decide both whom to imprison and for what maximum period. Nonetheless, the researchers (Wilkins et al., 1978:7) twere] confident that there did exist an implicit policy formulation which acted as an underpinning for judicial decision-making in the sentencing area. Through careful analysis of present practice, [they] believed it possible to discover that implicit policy and make it explicit. It was expected that this process would inform judges of the elements of that "implicit policy" so that some of those elements could be in- corporated into sentencing guidelines. The notion of descriptive guidelines implies an intent to do little more than create a statistical restatement of what a court has been doing. For this reason, judges have been expected to be less likely to resist guide-

Structuring Sentencing Decisions 145 lines than other sentencing reform innovations (Wilkins et al., 1978:3~31~: When comparing sentencing guidelines to legislatively mandated sentencing proposals, the most striking positive practical aspect of the guideline system is that it is judicially implemented and judicially controlled.... The use of sen- tencing guidelines should lead to less circumvention because it is the existing policies of the court itself that are initially being made explicit.... Guideline use . . . will significantly reduce unjustified variation from the established norm by making the established policy of the court explicit. The first generation of sentencing guidelines led to several challenges to "descriptive" guidelines: first, whether such a thing as an "implicit policy" of a court exists; second, the degree to which the statistical models intended to capture the "implicit policy" reflected actual sen- tencing practice and, more generally, whether any simple linear model can adequately reflect the complex considerations that influence sen- tencing decisions (and thereby serve as the basis for establishing a sen- tence for any particular case); and third, whether the availability of models of past practice obviates the need to make normative choices when developing guidelines. There remains considerable disagreement over whether any implicit policy exists in any court. Some argue that different judges invoke dif- ferent considerations in any given case—some will focus on retributive principles while others are primarily concerned with incapacitation and that shared considerations are weighed differently by different judges, so that talk of a common implicit policy can be misleading. Any model of past practice based on pooling cases of different judges may lead not to the discovery of a single implicit policy but to an average of several different implicit policies. This is most clearly illustrated when views on sentencing within a court are polarized and the averaged implicit policy constitutes a position that no individual judge would endorse. Others hold that there are observable consistencies in the sentencing practices of different judges in a court. Judges all take offense serious- ness into account, whether because a more serious offense deserves a more severe punishment or because a person who committed a more severe offense may be perceived as a greater threat to society if set free. Thus, analysis of past decisions of a court allows one to describe the practice of that court in terms of the case attributes that are taken into account, the weights associated with each attribute, and the particular mathematical form in which those weighted attributes can be combined to predict the expected sentence that would be handed down by a judge of that court in any particular case. The degree to which any model represents actual court practices de- pends on the skills of the modeler in incorporating the complexity of

146 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM the considerations that enter sentencing decisions. If a model represents a court's practices poorly, or if it incorporates discriminatory practices of a court, then its use would be undesirable and could be misleading as a guide to future decisions. A model that is generally representative of a court's decisions on sentence outcome can have several policy uses. First, the articulation of past sentencing practices can serve as a standard for judges in their individual sentencing decisions, and also as a basis for possible recon- sideration of prevailing practices in an iterative process of description, evaluation, and modification of sentencing policies. Second, information about past practices can provide a basis for assessing how much proposed standards diverge from traditional patterns. Because radical changes may engender resistance by those officials whose decisions are at issue, identification of sharp divergences may help policy makers anticipate likely sources of resistance. Third, knowledge of past practices can pro- vide benchmark information that educates policy makers to the actual rather than the presumed operation of the criminal justice system. For example, legislators generally consider prison sentences in terms of stat- utory maximums while judges deal with maximum sentences to be served; knowledge of the much shorter prison terms actually served may facil- itate informed consideration of realistic policy choices. Fourth, reliable data on past practices can be used to project the impact of alternative proposed policies and thereby enable decision makers to assess the costs and feasibility of various policy options. A statistical model of past practice should not be used merely as a "sentencing machine" or as the sole or primary basis for formulating sentencing policy. Sentencing policy and practice are both dynamic; blindly incorporating past practice into "descriptive" sentencing guide- lines may institutionalize a behavior that no longer reflects current prac- tice or policy (if it ever did). Furthermore, the mathematical form of the model of past practice that has generally been used has been a simple linear model, but actual decision behavior almost inevitably uses much more complex logic; that logic should be explored in discussions with court personnel and should inform modeling efforts. And even if a model can be formulated that is a credible statistical characterization of past practice, such a model cannot be transformed into "descriptive" sentencing guidelines without making ethical judg- ments. The Denver report (Wilkins et al., 1978:31-32) notes that the research which undergirds the guidelines development, and the guidelines themselves, are essentially descriptive, not prescriptive.... They do not tell what the sentences or the criteria ought to be.

Structuring Sentencing Decisions However, some ethical judgments are inevitably required. 147 In the case of the simplest hypothetical descriptive sentencing guide- lines (in which no ethically objectionable variables, such as race, directly influence sentences), if there is unacceptably large variation in average judicial behavior, there must be a decision on how great a divergence from the mean will be accepted. The range of accepted variation will depend on the extent to which one wishes to reduce disparity, and it thus requires a judgment about the optimum amount of variation to be encouraged or permitted. Even in this simple case, however, omitted or mismeasured variables can introduce errors in the estimates of effects and result in misleading guidelines, particularly if the omitted variables are correlated with in- cluded variables. If the omitted variables are not correlated with any of the included ones, the estimates of the parameters of the model will not be biased, but they will be inefficient. This could lead to errors in assessing the statistical significance of individual variables. The problem becomes particularly important when an omitted vari- able is ethically objectionable. Suppose, for example, that the race var- iable is omitted from the model- presumably to avoid perpetuating any racial discrimination in sentences in the resulting guidelines and that race is correlated with one of the included variables, such as prior record. Then, if race had actually mattered in the past, the estimation of the equation without the race variable will lead to an erroneously estimated equation and to guidelines that build in the ethically unacceptable effect of race by giving an excessive weight to prior record, which incorporates the role of race as well as that of prior record. Under such guidelines defendants who have poor prior records would receive severe sentences not simply because of the effect of prior record on judicial decisions but also because in the past those with poor prior records tended to be black and blacks were sentenced more severely. Past racism would thereby be incorporated in the guidelines through the prior-record variable. Although one cannot eliminate the effect of ethically objectionable variables by eliminating them from the equation being estimated, it is possible to purge the models of past decisions of the contaminating effect of objectionable variables and use the purged equations in guidelines construction. However, this task also necessarily involves an ethical choice. The model must be formulated initially with the race variable (for example) included, and then any estimation of discrimination-free sentencing would use that model with the race variable suppressed. But in using the model with the purged estimates to inform future decisions, one must decide how all offenders should be treated: one can decide they should be treated as whites have previously been treated, as blacks

148 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM have been treated, as the average of the prior treatment of blacks and whites, or as some other combination. (For a more technical discussion of this and related issues, see Fisher and Kadane [Volume II].) Thus, "descriptive" guidelines do not provide a way to achieve a value-free depiction of past sentencing practices. Efforts to model past practices are useful tools in the development of sentencing guidelines, but the guidelines developers must inevitably make normative choices. USING EMPIRICAL DATA All sentencing guidelines are prescriptive in the sense that they involve normative choices and are hortatory in recommending future sentences; they vary with respect to the role of empirical data in their formulation. At one extreme are empirically derived guidelines purported to be based only on statistical description; these we regard as inappropriate and illusory. Efforts to develop such guidelines place researchers in the position of making policy choices, sometimes by default (see Coffee, 1978~. At the other extreme are guidelines uninformed by considerations of empirical data on past practices or the likely impact of policy choices. While feasible, we view this approach (most closely approximated in past experiences of legislatures in adopting mandatory minimum sen- tencing laws) as undesirable. In the middle is a policy-development process in which empirical data on past practices and projections of future impact are considered and inform policy choices. Such a middle ground appears to have characterized the development of the U.S. Pa- role Board's guidelines and the Minnesota and Pennsylvania sentencing guidelines. There have been many efforts to develop empirically informed sen- tencing guidelines. In each case, data have been collected on a sample of previously sentenced cases, generally from official court records. The efforts to model past practices have typically involved multivariate analy- ses aimed at identifying the combination of variables that explains the greatest proportion of variation in sentence outcomes. The resulting multivariate models have several important limitations. First, they can only describe what judges have done on the average: individual deviations are lost, and polar opposites are represented only poorly by their midpoint. The variables reflected in the model may not be the ones that influenced the judge's decision (the judge may not have seen them or may have ignored them), and the data contained in court records, presentence reports, and similar official and agency records are often unreliable. Also, the records may fail to include influential vari-

Structuring Sentencing Decisions 149 ables (e.g., the defendant's demeanor). These factors contribute to the error in any model that is estimated. Furthermore, even if the models reflect the variables and their weights, they may not adequately reflect the logic with which the variables are combined. As a result, sentencing models seldom explain more than a third of the variance in sentences, often less, and consequently provide at best a blurred picture of past patterns. These limitations require that modeling and data collection efforts be undertaken together. Both must start with careful observation of all the relevant participants, and especially prosecutors, probation officers, and judges. Representatives of each of these groups must be interviewed in systematic ways to elicit the considerations that they believe enter their own decisions and recommendations. This will give rise to the formu- lation of models that are potentially richer than those based solely on a simple linear enumeration of the variables available from court rec- ords. In developing data collection forms, researchers must be sensitive to the variables identified in interviews. If those variables are not avail- able in the records, they may have to be collected independently or prospectively as part of the research. Despite their flaws, multivariate analyses can be useful. They can provide crude but otherwise unavailable information on the relative weight apparently given by judges to important variables like offense seriousness and prior record. Multivariate analyses can be used to test for interactions among variables, e.g., the influence of offense serious- ness on sentence type and the influence on that relationship of a third factor, such as race. They can suggest how the various relevant variables have been treated in the past; they can warn of the potential role of inappropriate variables; they can permit comparison of the treatment of typical cases in the past with normative judgments as to how they should be treated. They should not, however, be viewed as dictating sentencing standards. Rather they should represent a starting point for the application of judgment and expertise. In many jurisdictions, these models, however crude, would represent the first attempt at articulating existing sentencing practices, and, if used with discretion, such efforts are likely to lead to more rational policy discussion and development. Many of the methodological and modeling issues raised by multivar- iate analyses in guidelines development are not generically different from those raised by sentencing research generally. The problems of omitted variables, variable measurement and scaling, measurement er- ror biases, and selection error biases discussed in Chapter 2 are equally troubling in this applied context. Sparks (Volume II) reviews the efforts

150 RESEARCH ON SENTENCING THE SEARCH FOR REFORM of a number of sentencing guidelines projects to gather data, model past practices, and translate the findings into empirically informed guidelines, and he discusses specific difficulties of those endeavors. Assuming that guidelines development will continue to include efforts to model past practices, three aspects of data analysis warrant consid- eration: data definition, data collection, and sample design. DATA DEFINITION There is no simple answer to the question of what data on past practices should be collected. At one extreme, the New Jersey guidelines project codebook contained 847 variables because the project staff "decided that every bit of data could possibly affect sentences, and that therefore no assumptions should be made at the onset to dismiss any data" (McCarthy, 1978:10~. At the other extreme, information might be col- lected only on variables that are both theoretically reasonable and be- lieved to be correlated with sentencing practices—namely the variables that will be used in later analyses and in the formulation of guidelines. Neither polar strategy is satisfactory. Collection of data on all con- ceivable variables is likely to prove unwieldy and to be highly vulnerable to problems of missing items and data unreliability. Data even on such relatively concrete offender variables as prior record are often incom- plete and inaccurate. Data on offender variables like past employment, education, and social stability are less reliable and are more often miss- ing. Data on variables such as offenders' parents, income, employment, and place of birth are often unavailable in the records of operating criminal justice agencies and are likely to be of exceedingly low relia- bility. In the New Jersey effort, for example, data on "education of offender's parent/guardian" were recorded in only 7 percent of the cases, and the reliability of these data is unknown (McCarthy, 1978:16, fnl2; Sparks and Stecher, 1979~. Perhaps the best prescription is that data should be collected on all variables that are reliably available and that can reasonably be believed to be associated with outcomes in a nontrivial number of cases. It would also be important to attempt to determine the kinds of information that are available to judges at the time sentences are imposed. Data Collection All of the empirically informed sentencing guidelines projects to date have collected data retrospectively, usually from court records and pro- bation office presentence reports. Several serious problems arise with

Structuring Sentencing Decisions 151 such data: some official data may not have been available to the judge at sentencing; the judge may not have used some of the information that was available; and environmental and subjective factors (e.g., de- fendant demeanor) that may have been influential in the decision are not contained in readily available court records. Models based on data not available to or used by a judge and missing data on the factors that actually influenced a sentencing decision will be subject to potentially large errors in estimation. One alternative strategy that has been sug- gested (Sparks, 1981) is prospective data collection, i.e., obtaining data on cases sentenced after beginning the research, using observation and interviews as well as official records. This would usually require a longer period for guidelines development to ensure a rich sample of sentenced cases. Sample Design Sample design choices depend on the intended uses of the data. For example, quite large samples may be required if separate models are to be-developed for subcategories of defendants. At least several hundred cases would be required for each subset if separate models were to be developed for (1) different offense classes, (2) guilty-plea and trial dis- positions, (3) male and female defendants, (4) white, black, Hispanic, native-American, and other racial or ethnic groups, or (5) for separate judicial districts, counties, or urban-suburban-rural areas. In addition, concern for missing data and for variables that have skewed distributions would argue for large or stratified samples. On the other hand, all these models need not be formulated on independent samples, and more efficient sampling designs are possible. However, the sample must be large enough to validate the statistical model of sentencing on a sample of cases that were not used in model construction. To control for prob- lems of changing case mix or of changing judicial practice, the validation sample ideally should be contemporaneous with the construction sam- ple. Whatever the method used to develop the data base, the resulting data and estimated models are but raw material for informing the de- velopment of sentencing standards. Guidelines developers must still confront a large number of policy and technical choices. DEVELOPING GUIDELINES: POLICY AND TECHNICAL CHOICES Reformers seeking to change official behavior and generate some de- sired distribution of sentences are subject to a recurring tension between

152 RESEARCH ON SENTENCING THE SEARCH FOR REFORM the normative and policy goals they wish to achieve and the need to obtain cooperation from decision makers. They must figure out how to make officials do what they want them to do. The tensions between reformers' desires to specify criteria that exactly express their normative goals and the need to promulgate standards in forms that will gain compliance from decision makers are present in a number of critical policy and technical choices. Some of these choices can be illustrated by reviewing options examined and selected by the Minnesota Sen- tencing Guidelines Commission. UNAVOIDABLE POLICY CHOICES The Guidelines Offense Guilty plea rates vary, but typically between 75 and 95 percent of con- victions in a jurisdiction result from guilty pleas (Sourcebook, 1980:Table 5.19), many of which follow plea negotiations. Where charge bargaining is prevalent, the conviction offense is the offense to which the defendant pled guilty, not necessarily the offense that was originally charged. Policy makers must decide whether sentencing standards should be applied to the conviction offense, which may be an artifact of plea bargaining, or to some other offense measure. No solution is ideal. Using conviction offenses rewards defendants who have the most effective lawyers and punishes those who, for whatever reason (sometimes naivete or contri- tion), decide to plead guilty to the offense originally charged. Other options are available. For example, a critical provision of the Model Sentencing and Corrections Act (National Conference of Commissioners on Uniform State Laws, 1979) prescribes: "In determining the appro- priate guideline to follow the court shall consider the nature and char- acteristics of the criminal conduct involved without regard to the offense charged" [emphasis added]. Such provisions are common in parole guideline systems. Basing guidelines on defendant's "actual offense be- havior" arguably deprives defendants of the benefits of the bargains in return for which they waived their trial rights and can result in the punishment of defendants for alleged but unproven behavior. The choice is between a principled approach that may be impractical and a realistic choice that is unprincipled. The Minnesota Sentencing Guidelines Com- mission opted for principle and elected to base guidelines on conviction offenses because "serious legal and ethical questions would be raised if punishment were to be determined on the basis of alleged, but unproven, behavior" (Minnesota Sentencing Guidelines Commission, 1981:2~.

Structuring Sentencing Decisions Guilty-Plea Discounts 153 "Plea or trial" is a commonly used variable in sentencing research. Substantial evidence suggests that defendants who plead guilty receive sentencing leniency or a discount in exchange for a guilty plea (see Chapter 2~. If such discounts are common, and if counsel and judges believe they are necessary to induce the majority of defendants to plead guilty, sentencing policy makers are presented with a dilemma. If they provide discounts for guilty pleas, as suggested by Schulhofer (1979, 1980), they can be accused of encouraging an unattractive aspect of the criminal justice process and of placing unwarranted pressure on de- fendants to plead guilty especially when the guilty plea means the difference between probation and prison. If they do not provide dis- counts, they can be accused of inviting future circumvention of guidelines because they have deliberately defied courtroom conventions in which defendants, lawyers, and judges have an interest. The Minnesota commission "determined that the severity of of- fenders' sanctions should not vary depending on whether or not they exercise constitutional rights during the adjudication process" (Min- nesota Sentencing Guidelines Commission, 1981:13~. In other words, willingness or refusal to plead guilty and waive one's constitutional right to a trial and related rights may not be used to justify departures from applicable guidelines based on conviction offenses. But the informal practice of discounting through a charge-bargaining arrangement in which the defendant agrees to plead guilty to a lesser charge has not been regulated by the Minnesota guidelines. Social Variables Most empirical analyses of sentencing practices and outcomes have con- cluded that the seriousness of the offense and the offender's prior record (e.g., prior convictions or incarcerations, custody status at the time of the offense) are the best predictors of sentences. Various status variables (e.g., education, employment, marital status, and residential stability) are also commonly thought to be germane to sentencing outcomes. But such variables are significantly correlated with race, class, income level, and sex. While direct use of race as a sentencing criterion would be unconstitutional, the use of other criteria correlated with race is unlikely to be declared unconstitutional (see Coffee, 1976), and the use of var- iables that are correlated with class or sex is even less likely to present constitutional problems.

154 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM The Minnesota commission identified status variables '`as factors that should not be used as reasons for departure from the presumptive sen- tence because these factors are highly correlated with sex, race, or income levels" (Minnesota Sentencing Guidelines Commussion, 1981:13~. The other option, of course, is to use such variables, especially when they have been shown to be significantly predictive of sentence out- comes. Early versions of the U.S. Parole Commission's guidelines used employment, educational, and residential pattern variables in a "salient factor score" used to place prisoners in parole prognosis categories. Over the years, however, for policy reasons the Parole Commission has eliminated those variables. The version that became effective September 1, 1981, contained none of them. UrbanlRural Differences Much anecdotal and some empirical evidence (see Martin, Volume II) suggests that, in many states, there are significant local and regional differences in sentencing severity. In Pennsylvania, for example, offen- ses against the person are punished substantially more severely in sub- urban and rural counties than in Philadelphia. The Minnesota Sentenc- ing Guidelines Commission also found evidence of some regional differences in sentencing, especially for offenses against the person. There is a tension between the generally perceived need to establish uniform sentencing criteria throughout the jurisdiction contributing to the state prison population (it is in prison that disparities become most apparent, when prisoners compare their sentences) and the competing consideration of deference to local experiences, culture, and attitudes. It is difficult in principle to justify sentencing variations that are attrib- utable solely to local experiences and attitudes, but as a practical matter, uniform state standards that depart substantially from local practices are especially likely to be resisted. In Minnesota the guidelines commission's decision to promulgate standards that do not authorize local differences does not appear to have been especially controversial. While no jurisdiction with statewide guidelines has explicitly taken account of local differences or authorized their invocation as the justification for departures from otherwise ap- plicable standards, in some states this issue has been sidestepped by adopting guidelines with wide ranges that implicitly allow regional dif- ferences in sentencing to continue. In Pennsylvania the initially proposed guidelines would have resulted in substantial increases in sentence se- verity in Philadelphia and Pittsburgh and decreases of severity in sub- urban and rural areas. This aroused strong opposition to "uniformity,"

Structuring Sentencing Decisions 155 which contributed to the rejection of the guidelines and their subsequent revision to permit broad judicial discretion, indirectly allowing for local variation. Comment The preceding list of unavoidable policy choices does not exhaust the variety of major issues that must be resolved in formulating sentencing standards. Some others include: whether prior arrests not resulting in conviction may be considered in sentencing standards; whether juvenile court records may be considered, and to what extent; and whether some sort of "decay" device should be developed so that prior criminality ceases to be considered or is given less significance after some period of time at liberty without offending. The policy questions discussed in this section are subject to a common tension: they are often perceived and discussed in terms of principle, yet practical concerns urge pragmatic compromises. To the extent that policy makers resolve issues in terms of principle, they risk irrelevance: implementation of policy requires the cooperation of the practitioners who operate the process. The more that sentencing standards are viewed as unrealistic and unresponsive to real needs, the less likely is compliance with them. Minnesota's decision not to acknowledge a guilty-plea dis- count in its sentencing guidelines, for example, presents prosecutors and defense lawyers with a choice: either to stop negotiating pleas and sen- tencing concessions or to devise ways to reward guilty pleas (by charge reductions) even though the guidelines would appear to ban that prac- tice. Of course responses by practitioners need not be uniform. Pros- ecutors could manipulate guidelines for some kinds of defendants but not others. For example, persons charged with less serious offenses could be offered charge concessions that would make a prison sentence un- likely, while serious offenders might not be offered any concessions. Preliminary analysis of changing practices in Minnesota suggests that this is what is happening in aggravated robbery cases (Minnesota Sen- tencing Guidelines Commission, 1982~. PRACTICAL ISSUES IN STRUCTURING SENTENCING DISCRETION In establishing criteria for sentencing that will be followed, the devel- opers of guidelines face a number of practical questions with respect to approach, specificity, and methods for scaling offense and offender var- iables. Mechanisms to structure sentencing vary in the range of the factors to be considered and in the specificity of the criteria to be used

156 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM for decision making. At one extreme are traditional indeterminate sen- tencing systems, in which statutes specify only maximum (and occa- sionally minimum) lawful sentences for each offense. At the other ex- treme, one could design a detailed sentencing schedule that specifies precisely the in/out and sentence-length decisions for each offense type. Most recently adopted statutory presumptive sentencing and guidelines systems fall between these extremes. Approaches for Formatting and Presenting Sentencing Standards Sentencing reform initiatives have taken various approaches in format- ting and presenting sentencing criteria. Some are quite complex and involve many variables; others are simple. Theoretically the differences among them are purely formal, since each approach could be designed to use the same information in making sentencing decisions. As is in- dicated below, however, the differences may have important effects in practice. Statutory Formulation of Standards Statutory presumptive sentence laws are one form of sentencing standard. Typically they specify presumptive sentences or ranges. (See Lagoy et al. [1978] for descriptions of deter- minate sentence laws in California, Illinois, Indiana, and Maine.) Most such statutes limit their specifications to the dimension of offense se- verity; other salient sentencing information is left for consideration by judges as aggravating or mitigating circumstances. In contrast, Califor- nia's Uniform Determinate Sentencing Law provides for sentence "en- hancements" based on various prior-record and offense circumstances. Although the California law is expressed in conventional statutory prose form, in substance California's sentencing criteria could be expressed as a matrix. For robbery, the matrix would look like Table 3-2. The rows divide robberies into three categories (ordinary, aggravated, and mitigated) in order to specify base terms. The columns show the incre- TABLE 3-2 California Robbery Matrix Years Offense Base Term Enhanced Term 2 3 4 Aggravated robbery 5 6 7 8 9 Ordinary robbery 3 4 5 6 7 Mitigated robbery 2 3 4 S 6

Structuring Sentencing Decisions 157 mental consequences of having various prior incarcerations and offense circumstances (weapon use, serious injury, large property loss or dam- age) pled and proved. Although no other state has adopted statutory sentencing criteria that are as detailed as California's, a number of bills prescribing detailed statutory sentence criteria were introduced in the U.S. Congress in the mid-1970s. The proposed Fair and Certain Punishment Act (S. 3752, 1976), for example, subdivided offense definitions on the basis of the offender's intent and the resulting harm and specified a presumptive sentence for each subcategory that could be increased or decreased by no more then 40 percent for aggravating or mitigating circumstances. Matrix Format The original U.S. Parole Commission guidelines were expressed as a matrix. The earliest state parole guidelines (in Minnesota, Oregon, and Washington), and the initial "descriptive" sentencing guidelines (in Denver, Chicago, and Philadelphia) also used a matrix format, as do most of the parole and sentencing guidelines systems now in effect. The widespread adoption of the matrix approach results largely from its practicality. It is compact and efficient and can convey information much more efficiently than can statutory prose. This makes it easy to understand and apply and thereby fosters consistency. Offense severity is ordinally scaled, and offender scores are scaled and uniformly applied. Some variables are explicitly included, and this implicitly diminishes the significance of others that might otherwise be used. Some variables are explicitly excluded. Consequently, there is less likelihood of arbitrary choices or policy inconsistencies that arise from oversight. Sequential Guidelines Another approach, illustrated by the Washing- ton State parole guidelines, involves a sequential series of calculations. There are guidelines for eight different offense categories (e.g., robbery, property offenses, assault, drugs). For each offense category the guide- lines specify variables that, if present, prescribe addition of a specified increment (or range) of months of imprisonment. For each offender a term of months is calculated for the present offense. A similar pri-or- record guideline sets out prior conviction variables, each specifying in- crements in months. The sum of these increments is then added to the offender's base sentence. Finally, a "public safety" guideline containing variables characterizing criminal history, social stability, and institu- tional behavior is used to predict recidivism probabilities and on the basis of that calculation to reduce the prison term by a specified per- centage. The guideline release date is determined by combining the

158 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM results of the present offense, prior-record, and "public safety" calcu- lations. The range and complexity of detail that can be considered in sentencing using sequential guidelines is greater than that available through matrices but less than with computerized guidelines. Formula Guidelines In this approach, exemplified by the Massachu- setts sentencing guidelines, the applicable guideline sentence is deter- mined by use of a weighted formula. Values are assigned to the seri- ousness of the offense (A), to weapons use (B), to injury inflicted (C), and to prior convictions (D), in the formula X = 2.1A + 9B + 9C + 1.6D. The weights derive in part from regression coefficients obtained during efforts to develop a statistical characterization of sentencing prac- tices in Massachusetts. The guideline sentence range in months is given by X plus or minus 50 percent of X. Manuals Several jurisdictions have developed highly particularized guidelines manuals that provide detailed offense and offender criteria for every offense type. New Jersey's statewide guidelines may be the extreme case (see Sparks and Stecher, 1979~. Because the manuals con- tain a mass of detailed information, considerable internal cross-refer- encing is required in order to use them. The primary advantage of such manuals is that their specificity pro- vides highly detailed offense-specific information. This may be partic- ularly appropriate under a guidelines system like New Jersey's in which the manual contains the raw aggregate data derived from a guideline construction research effort and in effect provides a judge with infor- mation on past practice but then leaves to the judge the decision of how to use that information. The basic disadvantage of such a complex ap- proach is that it may be especially vulnerable to calculation errors. Computer-Assisted Guidelines One obvious solution to the complexity of a manual lies in coding its rules into a computer. The Institute for Law and Social Research, Inc., and Yankelovich, Skelly and White, Inc. (1981:xviii, hereafter cited as INSLAW) recently proposed devel- opment of computer-calculated guidelines: As the offense and offender descriptions grow rich in detail. . . the mechanics of translating all that detail into specific sentences will grow increasingly com- plex. In an era of sophisticated information processing capabilities this problem is clearly one that is not especially difficult.

Structuring Sentencing Decisions 159 In principle, there need be no difference between computerized guide- lines and guidelines that take other forms. The sentencing equations in any sentencing system are intended to transform the attributes of a case and an offender into a guidelines sentence. This can be done with the aid of a printed form that leads a court staff person through a series of calculations, including reference to some guideline matrix, or by a com- puter, probably more easily and with less chance of mechanical error. However, the INSLAW report (1981:VI-3) concludes that "computer assistance in sentencing may be an idea whose time has not yet come." Sentencing is highly symbolic, expressing as it does community de- nunciation of an offender and often an effort by the judge to impose a sentence that is commensurate with the offender's blameworthiness. As a matter of fundamental justice, individuals are entitled to a punishment process in which the circumstances of their offenses, and any aggravating or mitigating circumstances, are taken into account. Thus, any use of computer assistance in generating guideline sentences must take con- siderable care to avoid being seen as a mysterious and mechanical "de- humanization" of this process. A related proposal involves computer-assisted sentencing in which the judge (or the probation officer) can retrieve statistical information on selected aspects of past practices on a case-by-case basis (see Rhodes and Conly, 1981:Ch. 16~. If a decision maker wants to know whether the educational attainment of robbery defendants has been associated with sentence severity in the past, that information can be obtained using a computer. This scenario is not peculiar to guidelines systems; in substance it involves no more than a standard computerized infor- mation retrieval system, different only in its efficiency from conventional statistical records systems. General Observations Simple approaches, like prose guidelines and matrices, are easy to understand and so intuitively clear. Such ap- proaches, therefore, may enhance the credibility of the sentencing stan- dards that they express and thereby be more likely to elicit cooperation. This may be particularly desirable when decision makers are not legally obliged to comply with the standards. Conversely, when policy makers want to prescribe detailed, weighted, criteria involving complex combi- nations of variables for sentencing, using a computer is much less cum- bersome than using many cross-referenced matrices. When decision mak- ers' compliance can be controlled, the nonintuitive character of more complex approaches may be less important. Whether such approaches result in diminished credibility of the guidelines in the eyes of decision makers, and

160 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM accordingly in lower rates of compliance with them, is a subject on which interviews or experimental research might provide insights. Organizing Principles of Guidelines Matrices Existing guidelines systems range from those like Minnesota's and Penn- sylvania's, in which all sentences are included in one general matrix, to those like New Jersey's, which effectively establishes a separate matrix for each offense type. In between are a number of guidelines systems that use different organizing principles.8 Type of Offender The U.S. Parole Commission's guidelines (Septem- ber 1, 1981) consist of two matrices, one applicable to "youthful of- fenders" and persons imprisoned under the Narcotics Addict Rehabil- itation Act (NARA) and the other applicable to all other adult, federal prisoners. Each matrix encompasses all federal offenses and uses a com- mon method of grouping offenders on the basis of predicted group recidivism rates. Recommended prison terms in the youthful offender/ NARA matrix are shorter than those for the other matrix. Statutory Offense Classification Several guidelines systems have de- veloped separate matrices for each statutory offense class. Thus the Denver guidelines (as set forth in Kress, 1980:Appendix A) contain separate matrices for three felony classes and three misdemeanor classes. The principal argument for this organizing principle is that it defers to legislative assessments of the relative seriousness of the various statutory classes. For similar reasons, the Michigan guidelines provide matrices that distinguish among offenses on the basis of the statutory maximum sentences. In several jurisdictions, guidelines developers have concluded that individual statutory classifications can cover an extremely wide range of offense behavior under a single offense type, thereby inade- quately distinguishing among offenses, and that such classifications em- phasize the worst case in setting a maximum rather than reflecting pun- ishment for the usual case. Hence, developers have created their own scales of offense severity; the Pennsylvania sentencing guidelines and the U.S. Parole Commission's guidelines are examples. ~ Guidelines and related materials are developed and used by operating agencies and tend to be unavailable in published form. Kress (1980) discusses at length the Denver, Newark, Cook County, Philadelphia, Phoenix, and Washington guidelines and in appen- dixes reprints the guidelines and informational booklets for Denver, Philadelphia, Phoe- nix, and the state of Washington.

Structuring Sentencing Decisions 161 Generic Offense Groupings Several of the early "descriptive" guide- lines systems grouped offenses in broad general categories. For example, the Essex County (Newark) sentencing guidelines contained four grids (violent, drug, property, and miscellaneous crimes), and the Philadel- phia sentencing guidelines contained two grids (offenses against the person and all other offenses). The Washington parole guidelines con- tain eight generic groupings (murder II, manslaughter, sex offenses, robbery, assault, property offenses, drugs, and escape). One advantage of the generic approach is that it permits greater specificity of criteria than statutory offense classes. Linking Offender Score to Current Offense Guidelines may be designed to link an offender's prior-record score to the class and seriousness of the current offense in various ways. The Washington parole guidelines, for example, accord different weights to various kinds of prior felony convictions, depending on the present offense. Thus, a prior assault conviction adds 48 months to a base sen- tence when the current offense is murder; 24 months for assault; 12 months for robbery; and 6 months for drugs. By contrast, although statutory offense classes are typically heterogeneous, guidelines like the original ones in Denver use the same criminal history criteria for all persons sentenced under a single matrix and cannot weight prior-record items differently in accordance with their relevance to the nature of the current offense. The same sentencing criteria can be expressed under various grouping systems. Even a single comprehensive matrix system like Minnesota's could be particularized by adopting different criminal history scoring systems for each offense type: the consequences of having a particular prior-record score could be uniform even though the factors contributing to the scores might vary among offense types. For example, separate scoring systems could be devised for each offense so that a single variable (e.g., a prior rape conviction) might represent three points in connection with a current rape conviction, two points for a current assault convic- tion, one for theft, and zero for tax evasion. Specificity of Sentencing Standards An important goal of sentencing policy changes is achieving substantial consistency in sentencing patterns while permitting special treatment for special cases. For example, under the presumptive Minnesota sentencing guidelines, the judge is directed to impose sentences from within narrow ranges (plus or minus 5-8 percent from the midpoint) "unless the case

162 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM involves substantial and compelling circumstances." When the sentence departs from the guidelines the judge must provide a written explanation as to why the sentence imposed is more appropriate than that provided in the guidelines. The adequacy of this explanation is subject to review by the state supreme court. The aim of "descriptive" sentencing guidelines was similar. The de- velopers sought to establish guideline ranges that covered 8(~85 percent of the sentences imposed in the construction and validation samples. Some of the remaining "outliers," the other 15-20 percent, may have been extraordinary cases, but others may have been ordinary cases that received aberrant sentences. The developers of those guidelines ex- pected that extraordinary cases would continue to receive extraordinary treatment (see, e.g., Wilkins, 1981~. The ordinary outliers were a pri- mary target of descriptive sentencing guidelines: the judge, considering imposing a 5-year sentence in a case for which, according to the guide- lines, 85 percent of convicted persons receive a sentence of 1-1~/: years, may reconsider and impose a sentence from within the guideline range. To achieve greater consistency in sentencing and at the same time allow sufficient flexibility to accommodate cases presenting special cir- cumstances, guidelines developers have had to address various other technical issues. Points and Ranges Developers of sentencing guidelines distinguish be- tween points and ranges. A point guideline specifies a single punishment for a particular combination of offense and offender circumstances: for example, persons convicted of burglary who have two previous felony convictions shall be imprisoned for 17 months. A range guideline spec- ifies outer limits on permissible sentences: for example, persons con- victed of burglary who have two previous felony convictions shall be imprisoned for a specified term from within the range of 14 to 20 months. Point guidelines have been discussed in the literature (e.g., Sparks et al., 1982) but have not been adopted in any jurisdiction. The practical distinction is between ranges and point/range combi- nations. The U.S. Parole Commission's guidelines provide a range in months from which examiners are to set the release date in ordinary cases. Most statutory determinate sentence laws (see, e.g., Lagoy et al., 1978:Table 5) and various sentencing guideline systems (e.g., those in Massachusetts and Michigan) also establish range guidelines. A few jurisdictions, including Minnesota (see Table ~1), have adopted point/range sentencing criteria that designate a single term of months for ordinary cases and also a range within which a sentence can vary to reflect aggravating and mitigating circumstances that warrant some mod-

Structuring Sentencing Decisions 163 ification but do not justify a major deviation from the ordinary sentence. As noted above, California established three prison terms for every felony (for robbery, the terms are 2, 3, and 5 years). Section 1170(b) of the California Penal Code provides "the court shall order imposition of the middle term unless there are circumstances in aggravation or mitigation of the crime." Thus for robbery, 3 years is the point and 2- 5 years is the range. The enhancements have the effect of raising the upper limit of the range. No research findings have been published comparing the effects of range and point/range guidelines on compliance rates. Proponents of the Minnesota guidelines suggest that point/range guidelines are opti- mally structured to foster compliance and allow flexibility. By setting a point, they provide a benchmark. By setting a narrow range within which judges may set sentences in ordinary cases to reflect special circum- stances, they encourage judges to adjust sentence lengths within the range rather than to depart from it. And by permitting departures from the range when there are substantial and compelling circumstances, subject to a reasons requirement and review, they also accommodate highly unusual cases without sacrificing the integrity of the guidelines. The Widths of Ranges Guidelines systems and statutory presumptive sentence laws differ substantially in the widths of the sentence ranges from which decision makers may choose. At one extreme, the Massa- chusetts guideline range (for date of first parole eligibility) is the number of months calculated from the Massachusetts guideline formula, plus or minus 50 percent. Thus, if the formula yielded 60 months, the guideline range would be 30 to 90 months. At the other extreme, the Minnesota sentencing guidelines range is the point guideline term plus or minus only 5-8 percent, and the Washington parole guideline range is the term of months determined in accordance with the state's sequential calcu- lation plus or minus 12.5 percent. Table 3-3 sets out sample ranges for the Massachusetts, Minnesota, and Washington guidelines. The Minnesota and Washington guideline ranges are narrow. The Massachusetts ranges are wide. If Minnesota judges and Washington parole hearing examiners comply with their guidelines in a substantial majority of cases, one would expect those jurisdictions to achieve sub- stantial consistency in the lengths of prison sentences served. However, because the ranges are narrow, one might expect that decision makers will depart from the narrow guidelines more often than under broad ranges. Conversely, one might predict relatively high apparent compli- ance rates with the Massachusetts sentencing guidelines, especially for longer sentences, although critics of sentencing disparities might not be

164 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM TABLE 3-3 Ranges of Presumptive Sentences Under Massachusetts, Minnesota, and Washington Guidelines Presumptive Sentence State lea Ma 30a 6oa Massachusettsb 5-15 1~30 15~5 3~90 MinnesotaC 9-11 19-21 2~32 57-63 Washington 9-11 1~22 2~34 5~67 a Midpoint of interval (in months). b Actual Massachusetts range. c Ranges derived from Minnesota sentencing guidelines grid. Estimated and rounded. impressed by a claim of 85 percent compliance rates with guidelines that specify such broad ranges for permissible sentences. Questions about ranges cannot be answered without discussion of normative premises and without reference to the context in which the guidelines will be implemented. An adherent of the goal of just deserts, who places high value on equality in sentencing and the reduction of disparities, would favor narrow ranges. One with utilitarian goals, in contrast, might urge broad ranges that permit lengthy incarceration when incapacitative, deterrent, or rehabilitative considerations appear germane and relatively short incarceration in other cases. One's view of the width of ranges may also depend on predictions about official reactions to guidelines. If one predicts that decision makers will reject the precise guidance of narrow guidelines, a plausible reform tactic would involve setting ranges wider than would ideally be preferred. If narrow guidelines were often rejected, there would be no residual guidance, and the result might be wide disparities in sentences. Broader guidelines might channel more decisions into the guideline range and thereby achieve less overall disparity in sentences. In a context in which administrative or other controls, such as credible appellate review sys- tems, can be brought to bear on decision makers, the prospects for compliance with narrow ranges may be greater than when such controls are absent. Thus narrow guidelines may be more practicable in Wash- ington, where parole examiners are subject to administrative controls, and in Minnesota, where appellate courts review sentences and a sen- tencing commission monitors sentences, than in Massachusetts, where judicial compliance with the guidelines is voluntary and judges can sim- ply ignore the guidelines if they find them too confining.

Structuring Sentencing Decisions 165 InlOut Guidelines Systems of sentencing criteria, including those that have adopted sophisticated graduated standards for determining sen- tence lengths for persons to be imprisoned, face a more difficult dilemma in finding a satisfactory approach for deciding who should be imprisoned. Even California's detailed statutory standards do not tell judges whom to imprison. Most convicted California felons are eligible for probation or suspended sentences, and the law is silent on that choice (except for a few offenses that are not "probationable"~. The Minnesota sentencing guidelines embody a clear policy on the question, but they can also result in outcomes that may be viewed as unfair. Under the Minnesota sentencing matrix, cells below the "in/out line" (see Table 3-1) specify a state prison sentence; those above that line specify a sentence other than state prison. The magnitude of the difference in sentences is highlighted by the cells adjoining the in/out line. A prior felony conviction adds one point to a defendant's criminal history score. Thus the difference between two persons convicted of second-degree assault but whose sentences are governed by adjacent cells could be one felony conviction. Yet one defendant (Row VI, Col- umn 3) should receive a nonimprisonment sentence and the other (Row VI, Column 2) should receive a 34-month prison sentence a dramatic difference. It is possible, and ironic, that the Minnesota guidelines, generally designed to reduce disparities and treat similar cases similarly, may result in increasing the differences in sentences received by similarly situated offenders whose cases fall close to the in/out line, particularly in light of the continuation of charge negotiation that may affect the offender's location in the guideline grid. The Minnesota commission carefully considered the philosophical and policy implications of the placement of the line on the grid. Yet any rigid single line (or mandatory decision rule) treats adjacent cells on different sides of the line differently. And in Minnesota the impact of that line on sentences is very substantial, although the difference be- tween cases just above and below it is not necessarily greater than the difference found between cases in adjacent cells on the same side of the line. Thus both the policy embodied in the line and its impact in cases is likely to be particularly troubling to judges and others concerned about the fairness and appropriateness of the punishment in cases in- volving close calls. While someone who supports retributive goals may be comfortable with the placement of the line on the Minnesota grid and the commission's effort to treat similarly situated offenders the same and to eliminate certain factors defined as illegal from decision making in individual cases, a utilitarian who is more concerned with individu-

166 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM alized justice may be more troubled by a clear policy that is rigidly or arbitrarily applied or overlooks subtle distinctions in individual cases. The same dilemma also arises in "descriptive" sentencing guidelines. Suppose statistical efforts to characterize past practices indicate that persons falling within a particular cell have been imprisoned in 55 per- cent of cases. The choices presented to the guidelines developers are to make that 55/45 cell an "in" cell, which would change sentencing practice if followed and probably undermine the guidelines in the eyes of judges who know that this is an ambiguous category of cases, or to adopt guidelines that have probation as one end of the guideline range. This latter option acknowledges the ambiguity of cases in that cell but pro- vides no guidance to judges. The alternatives a clear but possibly arbitrary policy or no policy guidance pose a difficult dilemma because in many jurisdictions a sub- stantial proportion of cases in which imprisonment is a realistic possi- bility fall in cells abutting the in/out line. In developing its guidelines, the Minnesota Sentencing Guidelines Commission conducted a study of case dispositions for a weighted sample (weighted N = 4,369) of cases decided in fiscal 1978.9 After designing the grid, the commission ex- amined what percentage of persons would have been imprisoned, by guideline cell, had the guidelines been in effect (Knapp, 1982; Minnesota Sentencing Guidelines Commission, 1980, 1982~. Table 3~ shows that 575 of the preguideline cases analyzed in the study would have fallen within sentencing guideline cells that specify impnsonment. Slightly more than half of those defendants (302) would have fallen within cells abut- ting the in/out line. Another 542 defendants would have fallen within nonimprisonment cells immediately above the in/out line. Thus only 273 persons would have fallen clearly within the imprisonment cells, and 844 would have fallen in the cells adjacent to the in/out line. In other words, 83 percent of the convicted offenders for whom prison was a realistic option fell in guideline cells abutting the in/out line. We do not pretend to have an answer to this problem. It may lie in having more refined and detailed criteria for categories of cases that abut art in/out line than are required for those cells that pertain to low- probability or high-probability imprisonment cases. 9 The Minnesota Sentencing Guidelines Commission's dispositional study consisted of data on 2,332 cases sentenced before guidelines: a 42 percent random sample of male offenders convicted and receiving a felony or gross misdemeanor sentence in fiscal 1978 and all females similarly convicted in that year. Counties with large Indian populations were oversampled. The cases were then weighted to reflect the distribution of all felony . . convictions.

Structuring Sentencing Decisions TABLE 3 - Hypothetical Application of Minnesota Guidelines to Preguideline Cases Sentenced in Fiscal 1978, Classified by Guideline Categories 167 Offense History/Risk Score Severity 0 1 2 3 4 5 6 + TOTALS 4 13 44 59 56 26 74 15 (474) (126) (69) (32) (23) (8) (16) a (748) II 6 24 27 56 85 53 (477) (90) (82) (24) (14) (18) III 6 16 40 57 (534) (171) (100) (79) IV 6 19 42 44 (563) (185) (139) (34) 100 16 (11) (716) 79 72 58 20 (35) (16) (21) (956) 86 62 77 19 (30) (23) (11) (986) V 17 37 78 83 80 100 50 33 (119) (34) (14) (13) (10) (2) (4) (197) VIb 12 22 45 86 66 61 100 25 (231) (78) (58) (15) (13) (14) (2) (412) VII 39 68 86 85 100 100 100 62 (97) (57) (28) (15) (11) (4) (7) (219) VIII 42 38 87 100 100 100 58 (46) (26) (16) (6) (10) (2) (0) (106) IX 35 100 100 68 (6) (0) (0) (4) (2) (0) (0) (13) X 100 100 100 - 100 (13) (5) (4) (0) (0) (0) (0) (17) TOTALS 9 24 45 63 80 62 77 23 (2571) (774) (511) (222) (149) (88) (72) (4387) NOTE: The numbers of cases shown in the figure are estimates of the cases represented by the sample, i.e., 4,369 cases estimated from the 2,332 cases in sample. The weighting procedure used to estimate the cases contains a rounding procedure that will occasionally cause the product of the percent and number of cases to yield fractions. a In/out line: the presumptive guideline sentence below the line is state prison; the pre- sumptive sentence above the line is something other than state prison. b The significant number of cases at level VI are presumptive imprisonment cases because of the application of mandatory minimum laws. SOURCE: Minnesota Sentencing Guidelines Commission (1982:18, Fig. 3~.

168 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM Scaling Offenses and Offenders The scaling problems that confound sentencing research also arise in guidelines development. For guidelines developers there are at least two principal problems: first, how to scale offenses in terms of their seriousness; and second, how to weight in- dividual attributes in offense and prior-record scoring systems. The offense scaling problem is generally posed as a choice between adoption of statutory offense classifications (either directly in terms of formal offense classes or indirectly in terms of statutory maximum sen- tences), and independent development of an offense severity scale. As a practical matter, statutory sentence criteria are necessarily governed by statutory classes, but most administrative or judicial guidelines proj- ects have developed their own severity rankings. Thus the U.S. Parole Commission (see Gottfredson et al., 1978) and the Minnesota and Penn- sylvania sentencing commissions developed their own offense severity rankings (see Martin, Volume II). Those jurisdictions that base their guidelines on statutory categories have typically developed intraclass variables for scaling offenses (see Kress, 1980:Appendix, for examples). The weighting question reduces to whether scoring systems require simple addition of zero or one point for each attribute or call for the calculation of scores using a differentially weighted sum of those attri- butes to express offense and offender scores. Most guidelines systems have used simple, zero/one scoring systems for attributes (e.g., one point for each prior conviction, one point if on parole at the time of the offense, etch. Although in theory the use of different weights for the variables in scoring systems (like Massachusetts's) could make more subtle distinctions between cases, as a practical matter there are several arguments against using a differentially weighted scoring system. First, since even sophisticated statistical analyses seldom explain more than one-third of the variation in sentences, the weights derived in such analyses may fail to accurately reflect the relative importance of variables in empirically derived guidelines. Second, the need for the potential precision of a differentially weighted system is greatly diminished in developing empirically informed guidelines. Third, simple zero/one scor- ing systems have been found to perform about as well as more statis- tically sophisticated procedures in tests of the predictive powers of var- ious scoring systems (see, e.g., Gottfredson and Gottfredson, 1979~. How Many Models? Researchers attempting to characterize past sen- tencing practices frequently find that the variables that explain the in/ out decision are different from those that explain the sentence-length decision. In developing "descriptive" sentencing guidelines, the most accurate model is a bifurcated one in which one set of variables guides

Structuring Sentencing Decisions 169 the incarceration decision, and another set of variables (or the same variables with different weights) guides the sentence-length decision, and possibly yet another set of variables guides the choice among non- incarcerative sentence options. So far as the panel is aware, no such multistage sentencing guidelines models have been developed, although some bifurcated models of past sentence outcomes have been estimated (see, for example, Rhodes, 1981~.1° Philosophical Implications The growing interest in the philosophy of punishment in the 1970s has influenced sentencing policy. For example, several recent innovations have expressly embodied retributive premises and rejected the legiti- macy of rehabilitative goals of punishment (see Chapter 1~. The panel takes no position in the philosophical debates on punishment but seeks to call attention to the fact that adoption of particular premises has important implications for the development of sentencing guidelines. In the interest of clarity the issues discussed in this section are assessed from the perspective of stereotyped punishment models: "thorough- going retribution" and "modern orthodoxy" (see White, 1978:7~. A retributivist believes that the moral quality of offenders' acts defines the amount of punishment they deserve, and the achievement of equality and proportionality in the distribution of punishment are given high priority. In Hart's (1968:231) words, a retributive theory asserts: first, that a person may be punished if, and only if, he has voluntarily done something morally wrong; secondly, that his punishment must in some way match, or be the equivalent of, the wickedness of his offense; and, thirdly, that the justification of punishing men under such conditions is that the return of suffering for moral evil voluntarily done, is itself just or morally good. A modernist "allows some place, though a subordinate one, to ideas of equality and proportion in the gradation of the severity of punishment" (Hart, 1968:233), while placing greater emphasis on the utilitarian goals of deterrence, rehabilitation, and incapacitation. Scaling of Offenses Retributivists are especially concerned with the coherence of offense severity scales. A theory of equality necessarily entails a theory of significant differences. Criminal codes often do not it Kress (1980:132) reports that, in developing the early "descriptive" guidelines sys- tems, bifurcated guideline models were considered but rejected in several cities.

170 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM provide bases for drawing sufficiently detailed distinctions between of- fenses of different moral consequence. In the federal system, for ex- ample, offenses are not classified, and the maximum sentences author- ized for various offenses are arbitrary (see Low, 1970a,b). In states having offense classification systems, each of the three or four felony classes encompasses acts of diverse characters. Consequently, most guidelines projects have developed their own systems of offense clas- sification based on the decision makers' determination of the relative seriousness of various offenses. A related problem is that criminal code sections are often drawn in general terms that do not make distinctions that sentencing policy makers consider relevant. From a single statutory definition of extortion (18 Pa. C.S. 3923), Pennsylvania's sentencing commission crafted five different extortion offenses (on the basis of the amount of money involved and the circumstances) and gave each of them a different severity ranking (Pennsylvania Bulletin 12:431, 1982~. When policy makers adopt retributive premises, offense scaling will be a matter of substantial importance, and efforts like those in Penn- sylvania will likely be required. To a modernist, scaling is less important (though not irrelevant) because retributive concerns are but one among many sets of punishment goals that should influence sentencing deci- sions. The Range of Discretion We noted earlier that ranges for sentences vary substantially in the amount of discretion that they allow decision makers, from Minnesota's plus or minus 5-8 percent to Massachusetts's plus or minus 50 percent. Retributivists, given the high value they attach to the achievement of equality and proportionality, would insist on narrow sentencing ranges. The enabling statute in Minnesota, for ex- ample, permitted the commission to establish ranges of plus or minus 15 percent, but the commission chose the narrower 5-8 percent range because "the Commission felt that broad ranges would increase the disparate treatment of similar cases and, in a sense, would allow disparity to continue in practice while defining it away in theory" (Minnesota Sentencing Guidelines Commission, 1980:12~. A modernist, by contrast, would prefer broader ranges because they permit a decision maker to give substantial weight to such considerations as deterrent and incapa- citative effects when they appear appropriate. Thus, Morris (1974:75) argues that equality of suffering should not be a primary goal of sen- tencing: To say that a punishment is deserved . . . is not to say that it ought to be imposed. The concept of desert . . . is one of a retributive maximum; a license to punish the criminal up to that point but by no means an obligation to do so.

Structuring Sentencing Decisions 171 Considerations other than retribution determine how much punishment should be imposed short of that maximum. The Punitive Content of Guidelines All sentencing and parole guide- lines systems include some measure of the seriousness of the present offense and some measure of the offender's prior record. Philosophical premises affect the relative significance accorded the present offense and the past record. This can be illustrated by reference to the U.S. Parole Commission's adult parole guideline matrix, which is set out in Table 3-5. A retributivist, who believes in looking only at the severity of the current offense, would find much to fault in the sentences specified in this matrix. If an offender's punishment should exactly "match the wickedness of his offense," the past criminality or risk of recidivism of the offender is irrelevant, and all persons convicted of a particular of- fense should receive the same punishment (but see von Hirsch t1981] for an argument that retributive sentencing schemes can appropriately take prior record into account). The U.S. parole guidelines take a dra- matically different position. A comparison of the first and last numbers in the three middle offense severity rows indicates that an offender's criminal history can increase the severity of the prescribed sentence by a factor of three (1~32, 14 44, 2~72~. A "very good" offender con- victed of a "high" severity offense (1~20 months) could serve the same sentence as a "poor" offender convicted of a "low" severity offense (12-16 months). In the U.S. Parole Commission's guidelines, the "of- TABLE 3-5 U.S. Parole Guidelines: Recommended Months of Incarceration Before Release on Parole for Adults Offender Characteristics Very Offense Severity Good Good Fair Poor Low ~6 6-9 9-12 12-16 Low moderate 0-8 ~12 12-16 16-22 Moderate 1~14 14-18 18-24 24-32 High 14-20 2~26 26-34 3W4 Very high 2i36 36-48 48-60 60-72 Greatest I 4(~52 52-64 6i78 78-100 Greatest II 52+ 64+ 78+ 100+ SOURCE: U.S. Parole Commission Rules Sec. 2-20 (effective September 1, 1981~; 28 Code of Federal Regulations 2-20.

172 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM fender characteristics" score is based on a recidivism prediction table. and they effectively embody an incapacitative premise. The Minnesota Sentencing Guidelines Commission selected its final guidelines grid from among several that were variously characterized as "just deserts," "modified just deserts," "incapacitation," and "modified incapacitation" (see Knapp, 1980~. Figure 3-1 sets out the four sample grids. The commission discussed the philosophical implications of the sentence patterns in each grid and ultimately settled on a slightly revised version of the "modified just deserts" grid (see Table ~1~. The Minnesota commission's choice is slightly ironic. In selecting the "modified just deserts" model, the commission devoted most of its attention to the location of the in/out line. In the guidelines that were promulgated, the left-to-right differentials are more pronounced than in the U.S. parole guidelines. In one row (Row VI, Table ~1), the most severe sentence (104 months) is more than four times more severe than the least severe (23 months). Thus the guidelines adopted by the Minnesota commission do not fully reflect the just deserts premises embraced by the commission. In summary, to a retributivist, an offender's prior record should play little if any role in determining the nature of the punishment and so sentencing guidelines should be weighted to give little significance to prior criminal history. To a modernist, conviction makes an offender subject to a wide range of possible sanctions, and incapacitative, de- terrent, and rehabilitative considerations then enter in setting sentences. Philosophical premises not limited to the retributivist/modernist con- trast can also influence decisions about the legitimacy of various sen- tencing variables. General moral considerations may lead decision mak- ers to eliminate certain possible sentencing criteria like race, sex, and class that are overtly invidious and such social variables as employment history, education, and residential stability that are neutral on their face but correlated with invidious variables. General moral or ideological views may also lead to rejection of sentencing concessions to reward guilty pleas. The Minnesota commission, for example, decided that none of these criteria should be given weight in sentencing. Questions of the practical ramifications of philosophical views of pun- ishment are not new, but such questions have become more widely debated in recent years. When parole and sentencing decisions were mostly invisible and unreviewable, as they were under indeterminate sentencing systems, there was little need to ponder such matters as the systematic role of prior record or social variables in sentencing. With the development and promulgation of detailed sentencing and parole

Structuring Sentencing Decisions a. Just Deserts (low) 1 111 > IN ~ v IV v Vl V111 IX (high) X (low) 1 11 IV - > an V I I V Vl V111 IX (high) X _ _ ! 1 1 1 — I Imprisonment (low) 1 11 111 iV V Vl V11 V111 IX l I I I (high) X 0 1 2 3 4 5 6 (low) H ISTORY (high) high emphasis—severity very low emphasis—criminal history c. Incapacitation J Imprisonmen ~ (low) 1 11 111 lo: Al Oh V IV V Vl V111 IX I I I I I I J (high) X 0 1 2 3 4 5 6 (low) HISTORY (high) very low emphasis severity high emphasis—criminal history 173 b. Modified Just Deserts _ . — Imprisonment 0 1 2 3 4 5 6 (low) HISTORY (high) moderate emphasis—severity low emph - is—criminal history d. Modified Incapacitation _ I Imprisonment _ 1 1 1 1 1 1,~1 0 1 2 3 4 5 6 (low) H ISTORY (high) low emphasis~severity moderate emphasis—criminal history FIGURE ~1 Dispositional models considered by the Minnesota Sentencing Guidelines Commission. SOURCE: Knapp (1980:13~.

174 RESEARCH ON SENTENCING THE SEARCH FOR REFORM criteria, many of these questions have become salient and are receiving more attention from policy makers. THE PROCESSES OF DEVELOPING, IMPLEMENTING, AND ENFORCING NEW SENTENCING STANDARDS This section is concerned with political choices and planning questions that confront policy makers as they develop, implement, and enforce new sentencing standards and try to maximize compliance with them. DEVELOPMENT PROCESSES AND IMPLEMENTATION Sentencing is a complex process involving discretionary decisions by many people. Attempts to promulgate new sentencing standards that result in institutional changes have varied markedly in the processes by which reforms have been designed and the consideration given to po- litical repercussions of policy choices. Those jurisdictions that have made extensive efforts to obtain the understanding and support of all affected interest groups appear to have been more successful in gaining legislative approval when needed and fuller compliance when implemented than those that have not made such efforts. Martin's (Volume II) case study of the politics of sentencing reform in Minnesota and Pennsylvania indicates that: the complexities of developing sentencing guidelines involve not only the tech- nical issues related to the development of statistical models of past sentencing practices and projections of future prison populations, but also the political aspects of the policy-making process. In both states the legislature created a sentencing guidelines commission to promulgate guidelines that would go into effect unless rejected by the legislature. The differences in mandates, development processes, products, and outcomes illustrate the political problems of attempting to redistribute discretionary authority and change local sentencing prac- tices. In Minnesota the legislature determined that disparities in the pun- ishment of offenders convicted of felonies should be reduced through presumptive sentencing guidelines and the abolition of the parole board's discretionary releasing authority. The commission defined its task pri- marily as the development of public policy rather than as a technical activity. To gain support for its guidelines, the commission conducted a broad campaign to influence individuals and interest groups that would be affected by the guidelines and to involve them in the development process. The commission held a series of public meetings to publicize

Structuring Sentencing Decisions 175 its activities and solicit policy recommendations; cultivated close rela- tions with the media; had its members establish and maintain good relations with their respective constituencies; and held open meetings and arranged final hearings on the draft guidelines at meetings of as- sociations of trial judges, county prosecutors, and public defenders prior to submission to the legislature. The final guidelines embodied com- promises negotiated within the commission to reflect the views of interest groups and the extensive public debate the commission's work fostered. Consequently, all of the concerned groups perceived themselves as hav- ing a stake in the guidelines and viewed the resulting standards as pref- erable to the status quo; hearing no criticism, the legislature allowed the guidelines to go into effect without change. In Pennsylvania the commission was created to design guidelines for both felony and misdemeanor offenses (a broader mandate than in Min- nesota) and without clear guidance concerning the prison population ramifications of its possible decisions. When the commission designed guidelines similar to those in Minnesota, they met with criticism from virtually every interest group. This resulted in part from the commis- sion's limited effort to build a constituency and from the likely effect of the proposed guidelines. To reduce disparity in a state characterized by wide variation in local sentencing patterns, the commission proposed to average sentences statewide and restrict judicial discretion. Judges were angered at this limitation of their authority, and suburban and rural judges, prosecutors, and legislators opposed the reduction of sen- tence severity in their jurisdictions. These groups joined forces to press for legislative adoption of a resolution directing the commission to revise its guidelines to widen ranges and increase sentence severity. The revised guidelines now in effect aroused little opposition because they maintain symbolic acceptance of statewide standards but are broad enough to accommodate traditional local sentencing practices. The voluntary sentencing guidelines that have been adopted in several jurisdictions had neither legislative authorization nor broad support from or involvement of the local judiciary. Nor were defense bars and pros- ecutors' offices involved in the development process or, in several in- stances, even informed about the guidelines after they had been pro- mulgated. In Denver, Rich et al. (1981:165) note: the criminal division decided en bane that the probation department . . . should distribute the guideline work sheets to the judges but not to the attorneys. Sentencing guidelines . . . were to be downplayed as much as possible. In Chicago (Rich et al., 1981 :180~: . .

176 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM with a few exceptions, Chicago prosecutors and defense attorneys were unaware of the existence of sentencing guidelines. Given these limited efforts to obtain cooperation and support from lawyers, who jointly with judges influence the sentences that convicted persons receive, it should come as no surprise that Rich et al. (1981) found little evidence of commitment to or impact of voluntary sentencing guidelines in the cities they studied. Sparks et al. (1982) point to the likely importance of efforts to inform affected interested groups of the nature and intent of sentencing guidelines. In contrast to the Minnesota commission's efforts to build bridges to affected interests during the development process, in Massachusetts only minimal information about the guidelines was disseminated before the guidelines were introduced. Equally important as political bridge-building during the development stage is attention to the mechanical operation of a new sentencing system to facilitate compliance. Proposed sentencing and parole guidelines or determinate sentencing schemes are often complex. Applicable sen- tences can be identified only after numerical and statistical calculations have been made, based on the availability of necessary information. All such systems, but especially the more complex ones, are vulnerable to missing data and administrative errors, which can be reduced by careful planning, attention to detail, and the development of quality control procedures to verify adherence to the new policies. Officials cannot be expected to comply with a complex system without adequate familiarity with their responsibilities, information about its operation, and a dis- incentive for continuing "business as usual." Furthermore, efforts must be made to ensure that all the information necessary for calculating guideline sentences is routinely available. The effect of careful attention to implementation issues, including training the people who must implement new sentencing standards, developing mechanisms that facilitate implementation, and creating pro- cedures for monitoring compliance, is illustrated by the contrast in ap- proach and results in Minnesota and several jurisdictions with voluntary sentencing guidelines. To facilitate successful implementation of its guidelines, the Minnesota Sentencing Guidelines Commission under- took the following activities (among others): · Prepared commentary, which was included in training materials on the guidelines, to clarify the commission's intent, the relevant statutory provisions, and the applicable rules of criminal procedure to aid court personnel. · Worked with the corrections department to supplement the state- wide criminal records information system to ensure the availability of necessary data.

Structuring Sentencing Decisions 177 · In conjunction with the Minnesota Corrections Association, devised a new presentence investigation form that includes information neces- sary for the calculation of guideline sentences. · Recommended several legislative changes necessary to facilitate transition to the new system. · Conducted extensive training sessions for all judges, prosecutors, probation personnel, and defense attorneys to familiarize them with the guidelines. In addition, smooth implementation in Minnesota was facilitated by the following activities by affected participants: 1. Establishment of procedures for sentencing hearings by an ad hoc judicial committee. 2. Development of rules and standards for appellate review of sen- tences under the guidelines by the state supreme court. 3. Revision of prerelease and furlough policies consistent with the guidelines policy by the Department of Corrections. In contrast to Minnesota's approach, in Chicago and Newark training was limited and implementation mechanisms were largely ignored. The results were chronic problems with missing data and errors in computing guideline sentences, disputes about the accuracy of the facts on which computations were made, and disagreements over definitions of guide- line variables, e.g., what constitutes a weapon or injury (Rich et al., 1981~. And Sparks et al. (1982) observe that the Massachusetts guide- lines were not effectively presented even to the Massachusetts judiciary. Their presentation was very brief (2 hours), written materials were lim- ited to the guidelines themselves, and the description of the guidelines by one of the judges on the committee that produced them was inadequate. ENFORCEMENT The legal authority of sentencing standards and the existence, nature, and credibility of enforcement mechanisms also appear to affect the likelihood that the standards will be followed. Because there is little i~ Judge Douglas Amdahl, a member of the Minnesota Sentencing Guidelines Com- mission and chief judge of the Hennepin County (Minneapolis) District Court (and, subsequently, chief justice of the Minnesota supreme court), took the initiative by estab- lishing an ad hoc committee of judges and lawyers to propose procedures for sentence hearings. The committee's proposed rules were presented to a statewide meeting of trial court judges, modified in light of comments received at the meeting, and submitted to the chief judges in each judicial district for voluntary administrative adoption. The rules have been adopted in most judicial districts.

178 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM research on these issues, this section simply suggests several distinctions that may usefully inform efforts to implement sentencing innovations. Legal Authority Rules are characterized by various degrees of legal authority, and de- cision makers can be expected to give greater weight to rules of greater authority. Three levels of legal authority can be distinguished in the sentencing initiatives implemented to date, and a fourth level may soon exist. Voluntary Voluntary standards such as voluntary sentencing guidelines typically possess only collegial authority, so their surface credibility may be critical. Part of the rationale of the early "descriptive" guidelines was that, by articulating the prevailing norms and embodying the implicit policies of a court, voluntary compliance would be achieved. One might hypothesize, however, that compliance with voluntary sentencing stan- dards will be low: (1) when the development process is not understood and respected by the judges who must use or ignore them; (2) when the standards take a form that is alien to judges, such as Massachusetts's weighted formula guidelines; and (3) when the ranges for sentencing are narrow and deviate from prevailing practices in a large percentage of cases. Since these guidelines by definition are voluntary, a primary aim of developers must be to persuade judges that compliance with the guidelines is a good idea and that compliance will achieve important public aims.~3 Presumptive Presumptive sentencing and parole guidelines and stat- utory sentencing standards have presumptive authority; they are to be observed in the ordinary case and are to be disregarded only under i2 The original Denver guidelines were intended to encompass 8~85 percent of the sentencing decisions in the construction and validation samples (Wilkins et al., 1978~. Consequently, assuming that individual judges did not drastically change their respective sentencing standards, one would expect 8~85 percent of the sentences to be consistent with the guidelines. Such consistency with guidelines, however, does not mean that judges were compliant; judges could sentence consistently with the guidelines even if unaware of them. ~ ~ ~ A i3 The developers of the original guidelines recognized the importance of judicial sup- port. They made strenuous efforts to induce the judges to perceive guidelines development as a collaborative process in which the judges make the critical policy decisions (Wilkins, 1981~.

structuring Sentencing Decisions 179 special circumstances. Decision makers are generally required to give reasons for their decisions when they depart from the presumptive sen- tence. Thus under the federal parole guidelines, the Minnesota sen- tencing guidelines, and California's Uniform Determinate Sentencing Law, decision makers are required to provide written reasons for de- cisions contrary to the applicable presumptive standards. However, merely establishing presumptive standards does not necessarily lead to high compliance rates. Without effective appellate or other review, the rea- sons requirement may be meaningless. Or if the range of discretion permitted by the presumptive standards is large, substantial disparity may exist even though most sentences are in formal compliance with the applicable standards.~4 Mandatory Mandatory sentencing laws formally require that decision makers make particular dispositions. The mandatory sentence generally establishes a fixed minimum penalty for a broad class of cases that may vary widely in their individual circumstances. Such laws are vulnerable to circumvention because they are inflexible and require imposition of penalties that judges and prosecutors may believe to be inappropriate in individual cases. Their rigid and often severe penalties provide a powerful plea-bargaining weapon to a prosecutor who can promise to dismiss the crucial charge if a defendant pleads guilty to other charges. Their inflexibility can thus alter the balance of power relations in plea . . aargalnlng. Judicial Rules Judicial rules for sentencing, which do not yet exist in any jurisdiction, are an intermediate case. In several jurisdictions, no- tably Massachusetts and Michigan, statewide sentencing guideline proj- ects may eventually result in promulgation of court rules that give guide- lines presumptive foice. Trial judges are subject to numerous court rules on case processing and procedure; they are accustomed to adhering to such rules. Other things being equal, one might expect that sentencing guidelines that are promulgated as court rules are likely to possess greater authority than are voluntary guidelines and would be likely to result in greater levels of judicial compliance. i4 Under the original Illinois and Indiana presumptive sentencing laws, the statutory ranges applicable to persons convicted of forcible rape and to those who had two prior nonviolent felony convictions were ~50 and 6 60 years, respectively (Lagoy et al., 1978:399~. Under such presumptive standards, 100 percent compliance rates could easily exist along with gross unwarranted disparity.

180 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM Review Mechanisms Legal authority by itself affords little basis for predictions about sub- stantive compliance with sentencing standards. Judges, parole exam- iners, and lawyers can ignore or willfully circumvent even presumptive and mandatory sentencing standards. Legal authority becomes mean- ingful in the presence of credible review mechanisms that pose a realistic threat that failure to comply will lead to appeal to and overrule by a higher authority. Appellate Sentence Review Obtaining judicial compliance with sen- tencing standards may present some difficulties. Appellate sentence re- view appears to provide reasonably searching scrutiny of sentencing decisions in other countries. In the United States, however, there is no tradition of rigorous appellate review of sentences. Indeterminate sen- tencing laws gave immense discretion to sentencing judges, and there were no obvious criteria that appellate judges could invoke in order to assess the appropriateness of particular sentences. Recent presumptive sentencing guidelines and statutes may provide meaningful standards for appellate review in the United States. The Minnesota supreme court, for example, is carefully reviewing appeals arising from departures from the Minnesota guidelines. As a result, case law is now developing in that state articulating certain basic principles governing the choice of appropriate sentences. Two principles in par- ticular have been affirmed in various Minnesota supreme court rulings: (1) that the sentence be based on the conviction offense and not on alleged but unproved offenses and (2) that the severity of the sentence should be proportional to the seriousness of the offense when compared with other offenses (see Minnesota Sentencing Guidelines Commision, 1982~. One should not, however, be sanguine about the prospects of appel- late sentence review as a policing mechanism. Its greatest drawback is that it is dependent on appeals by the parties. If no one appeals, appellate courts will have no opportunity to review sentences, and quite substantial departures from guidelines or from statutory presumptive or mandatory sentencing laws will be beyond the ken of the courts. Most convictions result from guilty pleas, often pursuant to plea negotiations, and neither party has an interest in appealing such negotiated sentences. If a pros- ecutor has agreed to accept a plea conditioned on the defendant's re- ceiving a below-guideline sentence, later appeal of the defendant's sen- tence is unlikely. Thus the only cases in which appeals are likely are

Structuring Sentencing Decisions 181 those in which there was no agreement about a sentence and in which the judge failed to impose a sentence within the prescribed range, or there was an agreement the judge did not honor. Even if the number of appeals is small, however, appellate sentence review may have a powerful indirect effect on the application of pre- sumptive sentencing standards by providing an evolving frame of ref- erence within which plea bargaining occurs. Both prosecutors and de- fense counsel must negotiate in the shadow of the threat that if they are uncompromising, the case may be appealed. A possible result, therefore, is the gradual development of "going rates" for negotiated sentences, shaped and limited by the formal sentencing standards. Administrative Review The U.S. Parole Commission is in effect a sen- tence review agency for the federal district courts. Its parole guidelines are based on offense seriousness rankings and a parole prognosis (the "salient factor" score). The judge's sentence is not taken into account except when a minimum sentence is longer than the maximum guideline sentence or when a maximum sentence is shorter than the minimum guideline sentence: in these relatively rare cases (10 to 20 percent) the sentence prevails and the guidelines are overridden. Thus the Parole Commission in effect applies its own sentencing standards post hoc in reviewing sentence lengths to set release dates. Administrative sentence review of this sort has some advantages over appellate sentence review. Because parole review is not dependent on initiation by a party, it is less subject to collusive evasion of applicable sentencing standards. (However, it is not immune from collusive ma- nipulation: sentence bargaining with judicial acquiescence can ensure a sentence shorter than the earliest applicable parole guideline release date.) To avoid the effects of charge bargaining, the Parole Commission applies its guidelines on the basis of actual offense behavior, not the conviction offense, thereby adjusting for the effects of varying charging and bargaining patterns in different parts of the country. Moreover, because the Parole Commission has its own internal system of admin- istrative review, the quality of the reviewing decision can be assessed and revised when appropriate. Review mechanisms also are more likely to affect behavior in admin- istrative sentencing systems like parole than in judicially dominated systems. Administrative review procedures that have both formal and informal authority can be established. The formal authority resides in the review body's capacity to decide that the reasons provided for a noncompliant decision are unconvincing and to unilaterally change that

182 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM decision. The informal authority results from the bureaucratic nature of parole organizations: as civil servants, examiners have career ad- vancement concerns that encourage compliance with agency policy. There has been some discussion in Congress of various ways to en- hance the Parole Commission's sentence review function. For example, it has been proposed that, in connection with a presumptive sentencing guideline system resembling Minnesota's, parole release be available only in those cases in which a judge has imposed a sentence longer than is provided in the applicable guideline. No comprehensive system of administrative review of sentences has, to the panel's knowledge, been established. California has established a partial administrative review system, but it does not yet appear to have had significant effect. California Penal Code Section 1170(f) ini- tially directed the Board of Prison Terms: not later than one year after the commencement of the term of imprisonment tto] review the sentence [in all cases] and . . . by motion recommend that the court recall the sentence and commitment previously ordered and resentence the defendant in the same manner as if he had not previously been sentenced if the board determines that the sentence is disparate [emphasis added]. Whether this injunction will be effective in the future remains to be seen. Informal Review Mechanisms Several informal processes can be used for review of compliance with sentencing standards. Efforts can be made to attract media attention to sentencing. On numerous occasions citizens groups have organized court observation systems in order to monitor judicial sentencing behavior. The Minnesota commission has established an internal monitoring system. Sentencing information forms must be completed by the trial court for every case and must include reasons for departures in cases where the sentence departs from the guidelines. One copy of the sentencing report form is sent by the court to the commission for review. The commission has established a review and follow-up procedure by its staff that involves initially calling the pro- bation officer to obtain missing information or correct errors in sentence calculation prior to sentencing; contacting the judge if written reasons for a departure are missing following sentencing; and, if these are not forthcoming, contacting the chief judge in the judicial district. This monitoring system provides the sentencing commission with records on compliance and departures and serves as a constant reminder to judges that sentences are reviewed by the commission in every case, which may encourage judges to follow the guidelines.

Structuring Sentencing Decisions Such monitoring devices, if well designed, may also be essential in the process of amending sentencing standards. They can provide feed- back on the effects of the new standards on the system as a whole and on the congruence between actual and desired sentencing practices. This information can serve as the basis for modifying guidelines to overcome unanticipated problems and to accommodate changes in community standards and values affecting sentencing. 183

Next: 4 Sentencing Reforms and Their Effects »
Research on Sentencing: The Search for Reform, Volume I Get This Book
×
Buy Paperback | $85.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF
  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!