High rates of incarceration in the United States and the great numbers of people held in U.S. prisons and jails result substantially from decisions by policy makers to increase the use and severity of prison sentences. At various times, other factors have contributed as well. These include rising crime rates in the 1970s and 1980s; decisions by police officials to emphasize street-level arrests of drug dealers in the “war on drugs”; and changes in prevailing attitudes toward crime and criminals that led prosecutors, judges, and parole and other correctional officials to deal more harshly with individuals convicted of crimes. The increase in U.S. incarceration rates over the past 40 years is preponderantly the result of increases both in the likelihood of imprisonment and in lengths of prison sentences—with the latter having been the primary cause since 1990. These increases, in turn, are a product of the proliferation in nearly every state and in the federal system of laws and guidelines providing for lengthy prison sentences for drug and violent crimes and repeat offenses, and the enactment in more than half the states and in the federal system of three strikes and truth-in-sentencing laws.
The increase in the use of imprisonment as a response to crime reflects a clear policy choice. In the 1980s and 1990s, state and federal legislators passed and governors and presidents signed laws intended to ensure that more of those convicted would be imprisoned and that prison terms for many offenses would be longer than in earlier periods. No other inference can be drawn from the enactment of hundreds of laws mandating lengthier prison terms. In the federal Violent Crime Control and Law Enforcement
Act of 1994, for example, a state applying for a federal grant for prison construction was required to show that it:
(A) has increased the percentage of convicted violent offenders sentenced to prison; (B) has increased the average prison time which will be served in prison by convicted violent offenders sentenced to prison; (C) has increased the percentage of sentence which will be served in prison by violent offenders sentenced to prison.
Yet while individual laws clearly reflected a policy choice to increase the use and length of incarceration, it is unlikely that anyone intended, foresaw, or wanted the absolute levels of incarceration that now set the United States far apart from the rest of the world.
In this chapter, we describe and then assess the development of U.S. sentencing and punishment policies and practices since the early 1970s. The first section reviews the profound shifts in the U.S. approach to sentencing over the four decades of the incarceration rise, including the development of sentencing guidelines and determinate sentencing policies and more recent initiatives designed to increase the certainty and severity of prison sentences. The second section details principles of justice that have undergirded punishment policies in the United States and other democratic countries since the Enlightenment and demonstrates that many policies enacted over the past 40 years are inconsistent with those principles. The third section examines the disjunction in recent decades between policy-making processes and the available social science evidence on the effects of punishment policies. The fourth section surveys and analyzes disproportionate and damaging effects of recent U.S. punishment policies on members of minority groups. In the committee’s view, the nation’s policy choices that increased the incarceration rate to unprecedented levels violated traditional jurisprudential principles, disregarded research evidence that highlighted the ineffectiveness and iatrogenic effects of some of those policies, and exacerbated racial disparities in the nation’s criminal justice system.
American sentencing policies, practices, and patterns have changed dramatically during the past 40 years. In 1972, the incarceration rate had been falling since 1961 (see Figure 2-1 in Chapter 2). The federal system and every U.S. state had an “indeterminate sentencing” system premised on ideas about the need to individualize sentences in each case and on rehabilitation as the primary aim of punishment. Indeterminate sentencing had been ubiquitous in the United States since the 1930s. Statutes defined crimes and set out broad ranges of authorized sentences. Judges had discretion to
decide whether to impose prison, jail, probation, or monetary sentences. Sentence appeals were for all practical purposes unavailable. Because sentencing was to be individualized and judges had wide discretion, there were no standards for appellate judges to use in assessing a challenged sentence (Zeisel and Diamond, 1977). For the prison-bound, judges set maximum (and sometimes minimum) sentences, and parole boards decided whom to release and when. Prison systems had extensive procedures for time off for good behavior (Rothman, 1971; Reitz, 2012).
Few people questioned the desirability of indeterminate sentencing. The American Law Institute (1962) in the Model Penal Code, the National Commission on Reform of Federal Criminal Laws (1971) in its Proposed New Federal Criminal Code, and the National Council on Crime and Delinquency (1972) in the Model Sentencing Act all endorsed the approach.
Within a few years, however, the case—and support—for indeterminate sentencing collapsed. University of Chicago law professor Albert Alschuler described the sea change: “That I and many other academics adhered in large part to a reformative viewpoint only a decade or so ago seems almost incredible to most of us today” (Alschuler, 1978, p. 552).
Criticisms of indeterminate sentencing grew. Judge Marvin Frankel’s (1973) Criminal Sentences—Law without Order referred to American sentencing as “lawless” because of the absence of standards for sentencing decisions and of opportunities for appeals. Researchers argued that the system did not and could not keep its rehabilitative promises (Martinson, 1974). Unwarranted disparities were said to be common and risks of racial bias and arbitrariness to be high (e.g., American Friends Service Committee, 1971). Critics accused the system of lacking procedural fairness, transparency, and predictability (Davis, 1969; Dershowitz, 1976). Others asserted that parole release procedures were unfair and decisions inconsistent (Morris, 1974; von Hirsch and Hanrahan, 1979).
Not all objections focused primarily on consistency and procedural fairness. Conservatives objected that indeterminate sentencing allowed undue “leniency” in individual cases (van den Haag, 1975) and paid insufficient attention to punishment’s deterrent and incapacitative effects (Fleming, 1974; Wilson, 1975). Policy histories of California’s Uniform Determinate Sentencing Law of 1976 describe an alliance of liberals and conservatives favoring determinate sentencing and abolition of parole (Messinger and Johnson, 1978; Parnas and Salerno, 1978). A first set of sentencing guidelines developed by the Pennsylvania Sentencing Commission was rejected by the legislature after conservatives characterized them as being insufficiently severe (Martin, 1984).
Those criticisms sparked major changes in American sentencing and punishments, and ultimately in the scale of imprisonment. In retrospect, three distinct phases are discernible. During the first, principally from
1975 to the mid-1980s, the reform movement aimed primarily to make sentencing procedures fairer and sentencing outcomes more predictable and consistent. The problems to be solved were “racial and other unwarranted disparities,” and the mechanisms for solving it were various kinds of comprehensive sentencing and parole guidelines and statutory sentencing standards (National Research Council, 1983).
The second phase, from the mid-1980s through 1996, aimed primarily to make sentences for drug and violent crimes harsher and their imposition more certain.1 The principal mechanisms to those ends were mandatory minimum sentence, three strikes, truth-in-sentencing, and life without possibility of parole laws.2 Mandatory minimum sentence laws required minimum prison terms for people convicted of particular crimes. Three strikes laws typically required minimum 25-year sentences for people convicted of a third felony. State truth-in-sentencing laws typically required that people sentenced to imprisonment for affected crimes serve at least 85 percent of their nominal sentences.
The third phase, since the mid-1990s, has been a period of drift. The impetus to undertake comprehensive overhauls or make punishments substantially harsher has dissipated. No states have created new comprehensive sentencing systems, none has enacted new truth-in-sentencing laws, and only one has enacted a three strikes law. Mandatory minimum sentence laws have been enacted that target carjacking, human smuggling, and child pornography, but they are much more narrowly crafted than were their predecessors.3 According to annual reports issued by the National Conference of State Legislatures, several hundred state laws have been enacted since 2000 that in various ways make sentencing less rigid and less severe. Most of these laws are relatively minor and target less serious offenses. In
1A wide variety of other harsh criminal justice policies were adopted during this period, including registration, notification, and residence laws for sex offenders and a variety of “dangerous offender” and “sexual psychopath” laws. Similar initiatives affecting the juvenile justice system lowered the top age of juvenile court jurisdiction, made discretionary transfers to adult courts easier, and excluded some violent offenses from juvenile court jurisdiction regardless of the defendant’s age.
2Laws authorizing sentences without the possibility of parole were enacted for a number of reasons, including as part of a strategy by opponents of capital punishment to create a credible alternative to the death penalty.
3Summaries such as this must be hedged because no organization maintains a comprehensive database on changes in sentencing laws. The National Conference of State Legislatures for many years compiled annual summaries (of uncertain comprehensiveness) and maintains a searchable database beginning with developments in 2010 (http://www.ncsl.org/issues-research/justice/state-sentencing-and-corrections-legislation.aspx [February 28, 2014]). The Sentencing Project (e.g., Porter, 2013), the Vera Institute of Justice (e.g., Austin, 2010), and the Public Safety Performance Project of the Pew Charitable Trusts issue occasional selective summaries. None of these, however, is comprehensive or cumulative.
few cases have major punitive laws of the second period been repealed or substantially altered. High-profile changes to totemic tough-on-crime laws such as New York’s 1973 Rockefeller Drug Laws and the 1986 federal 100-to-1 law for sentencing crack and powder cocaine offenses were partial. In the first of these examples, severe mandatory penalties for many offenses continued to be required (New York State Division of Criminal Justice Services, 2012); in the second, a lower but still high—18-to-1—drug quantity differential for offenses involving pharmacologically indistinguishable crack and powder cocaine was established (Reuter, 2013).4 More typically, changes in state sentencing laws created exceptions to the coverage of mandatory minimum sentence laws or slightly narrowed their scope,5 expanded prison officials’ authority to grant time off for good behavior, made earlier release possible for narrow categories of prisoners, or reduced the probability of parole and probation revocations for technical offenses (Austin et al., 2013).
Sentencing reform initiatives proliferated in the aftermath of the rejection of indeterminate sentencing. The earliest and most incremental sought to reduce disparities through the development and use of parole guidelines and “voluntary” sentencing guidelines. These initiatives were followed by statutory determinate sentencing systems and presumptive sentencing guidelines.
Parole guidelines were the first major policy initiative of the sentencing reform movement, although one foot remained firmly in the individualization logic of indeterminate sentencing. In the 1970s, the U.S. Parole Board and boards in Minnesota, Oregon, and Washington created guideline systems for use in setting release dates. They sought to increase procedural fairness through the publication of release standards, reductions in
4Although the introduction of crack cocaine was associated with an increase in drug-related violence, subsequent reductions in violence have been consistent with the aging of the crack cocaine user and trafficker populations (U.S. Sentencing Commission, 2007, p. 83).
5Many recent changes in state mandatory minimum sentences laws authorize the imposition of some other sentence on selected offenders (Austin, 2010; Porter, 2013). Federal law long has provided such a “safety valve” for mandatory minimum sentence laws for drug crimes committed by first-time offenders who did not use violence or possess a gun and told the government all about their crime. In federal fiscal year 2012, nearly 40 percent of defendants sentenced under mandatory minimum sentence laws benefited from this provision (U.S. Sentencing Commission, 2013b, Table 44).
disparities in time served by those convicted of comparable crimes, and the linking of release decisions in part to empirical evidence on prisoners’ probabilities of subsequent offending (Gottfredson et al., 1978). The parole guidelines movement quickly lost steam, however, despite evidence of the guidelines’ effectiveness, when well implemented, in improving consistency in the setting of release dates and in time served for similar offenses (Arthur D. Little, Inc., and Goldfarb and Singer, Esqs., 1981; National Research Council, 1983, pp. 194-196). The four pioneering systems were abandoned in the 1980s, replaced in each case by presumptive sentencing guideline systems that also sought to achieve greater procedural fairness and consistency.
One advantage of parole guidelines is that they can make case-by-case decision making within a well-run administrative agency faster, less costly, and more easily reviewable than decisions made by judges. A second advantage is that, as commonly happened during the indeterminate sentencing era, parole boards can address prison overcrowding problems by adjusting release dates (e.g., Messinger et al., 1985). A major disadvantage, however, is that parole boards have authority only over those sentenced to imprisonment. Parole guidelines can reduce unwarranted sentence-length disparities among prisoners, but not between them and others sentenced to local jails or community punishments.
Voluntary Sentencing Guidelines
During the 1970s, local courts and, occasionally, state judiciaries in most states created systems of voluntary sentencing guidelines (Kress, 1980; National Research Council, 1983). Today, they would usually be referred to as “advisory” guidelines. Judges were not bound to follow them and needed to give no reasons if they did not; a defendant could not appeal the judge’s decision. Most early voluntary guideline systems were abandoned or fell into desuetude. Evaluations through the late 1980s, most notably of judicially crafted systems in Maryland and Florida, showed that they had few or no effects on sentencing decisions or disparities (Rich et al., 1982; Carrow et al., 1985; Tonry, 1996, Chapter 3).
Voluntary guidelines have attracted renewed interest because of two recent U.S. Supreme Court decisions (U.S. v. Booker, 543 U.S. 220 , and Blakely v. Washington, 542 U.S. 296 ), which created new procedural requirements for presumptive sentencing guideline systems. A small number of states now operate voluntary guideline systems, but credible research evidence on their effects on sentencing disparities is not available. However, prison population growth in two especially well-known systems using voluntary guidelines—in Delaware and Virginia—has long been below national averages.
Determinate Sentencing Laws
The most influential reform proposals during this phase called for the abolition of parole release and the creation of enforceable standards to guide judges’ decisions in individual cases and provide a basis for appellate review (e.g., Morris, 1974; Dershowitz, 1976; von Hirsch, 1976). Policy makers responded. Maine in 1975 abolished parole release and thereby became the first modern “determinate” sentencing state in the sense that the length of time to be served under a prison sentence could be known, or “determined,” when it was imposed. California came second, enacting the Uniform Determinate Sentencing Act of 1976; the act abolished parole release and set forth recommended normal, aggravated, and mitigated sentences for most offenses. Other states—including Arizona, Illinois, Indiana, and North Carolina—quickly followed California’s lead in enacting such laws. Evaluations concluded, however, that the laws had little if any effect on sentencing disparities (Cohen and Tonry, 1983; Tonry, 1996). No additional states have created comprehensive statutory determinate sentencing systems since the mid-1980s.
Presumptive Sentencing Guidelines
In 1978, Minnesota enacted legislation to create a specialized administrative agency—a sentencing commission—with authority to promulgate presumptive sentencing guidelines. Judges were required to provide reasons for sentences not indicated in the guidelines; the adequacy of those reasons could be appealed to higher courts. Minnesota’s guidelines took effect in 1980. Oregon, Pennsylvania, and Washington created similar systems in the 1980s. Evaluations showed that well-designed and -implemented presumptive guidelines made sentencing more predictable, reduced racial and other unwarranted disparities, facilitated systems planning, and controlled correctional spending (Tonry, 1996, Chapter 3). Kansas, North Carolina, and Ohio created similar systems.
The Minnesota, North Carolina, and Washington commissions operated under “population constraint” policies; the aim was to ensure that the number of inmates sentenced to prison would not exceed the capacity of state prisons to hold them. The population constraint policies worked. During the periods when they were in effect, those states experienced prison population growth well below national averages.
The primary policy goal of the early presumptive guideline systems was to reduce disparities and unfairness (Lieb and Boerner, 2001; Frase, 2005; Kramer and Ullmer, 2008). The approach was proceduralist and technocratic, focusing primarily on the development of procedures for improving consistency and predictability and of population projection models for use
in financial and facilities planning. The primary aim of North Carolina’s guidelines was to control the size of the prison population (Wright, 2002). This aim was realized: after the guidelines took effect in 1994, North Carolina’s incarceration rate through 2011 fluctuated between 340 and 370 per 100,000 population, while most other states’ rates rose substantially. Population constraint policies made obvious sense to the early sentencing commissions and the legislatures that established them.
Things quickly changed. From the mid-1980s through 1996, policy making in this area ceased to be significantly influenced by concerns about evidence, fairness, and consistency. In Minnesota, the legislature in 1989 instructed the commission to abandon its population constraint policy. In Oregon, the committee that had drafted and monitored the guidelines was disbanded, and the guidelines were trumped by a broad-based mandatory minimum sentence law enacted in 1994. The Pennsylvania Commission on Sentencing survived, but state supreme court decisions effectively converted the nominally presumptive guidelines into voluntary ones (Reitz, 1997; Kramer and Ulmer, 2008).
More generally, presumptive sentencing guidelines fell from favor. The three most recent presumptive guideline systems—those of Kansas, North Carolina, and Ohio (abandoned in 2006)—were established in the mid-1990s. A few voluntary systems have been developed since then. Sentencing commissions in Florida, Louisiana, Tennessee, and Wisconsin were abolished, and Washington’s lost its staff and budget in 2011 (Frase, 2013).
A number of studies have concluded that sentencing guidelines, especially with population constraints, help control the size of the prison population. Marvell (1995) compared prison population growth from 1976 to 1993 in nine states that had voluntary or presumptive guidelines with the national average and concluded that guidelines based on population constraints produced lower rates of population increase. Nicholson-Crotty (2004), using prison data for 1975-1998 in a 50-state analysis, concluded that guidelines based on capacity constraints tend to moderate growth in incarceration and that guidelines not based on such constraints exacerbate it. Stemen and colleagues (2006) analyzed state sentencing patterns in the period 1975-2002 and concluded that states that adopted presumptive guidelines and abolished parole release had lower incarceration and prison population growth rates than other states.
The promulgation of federal sentencing guidelines, which took effect in 1987, signaled the end of the phase of modern U.S. sentencing reform that targeted disparities and the beginning of a phase focused on increased certainty and severity. The Sentencing Reform Act of 1984 directed the U.S. Commission on Sentencing to develop guidelines for reducing disparities, to provide for nonincarcerative punishments for most nonviolent and nonserious first offenses, and to be guided by a prison population constraint policy.
The commission ignored the directives concerning first offenses and prison capacity and instead promulgated “mandatory” guidelines that greatly increased both the percentage of individuals receiving prison sentences and the length of sentences for many offenses (Stith and Cabranes, 1998). The federal guidelines were effectively converted from presumptive to voluntary by the U.S. Supreme Court in U.S. v. Booker, 543 U.S. 220 (2005).
Presumptive sentencing guidelines developed by a sentencing commission are the most promising means available to jurisdictions that want to reduce or avoid unwarranted sentencing disparities, improve budgetary and policy planning, or both. The well-documented successes of the Minnesota, Oregon, and Washington guidelines in the 1980s and of the North Carolina guidelines since their promulgation in 1994 show that both sets of goals are attainable.
Sentencing laws enacted from the mid-1980s through the mid-1990s differed substantially from most of those enacted in the preceding period. Whereas the earlier initiatives were aimed principally at making sentences more predictable and consistent and making processes fairer and more transparent, initiatives in the second phase of change in modern sentencing law typically targeted making sentences harsher and more certain and preventing crime through deterrence and incapacitation. The focus shifted from fairness to certainty, severity, crime prevention, and symbolic denunciation of criminals. The shift toward severity took place despite three generations of efforts, often with federal demonstration project funding, to develop alternatives to incarceration (sometimes synonymously called “intermediate sanctions” or “community penalties”) (Morris and Tonry, 1990).
The policy initiatives of the second phase, symbolized by the proliferation of mandatory minimum sentence laws, undermined pursuit of the aims of the first phase. Two centuries of experience has shown that mandatory punishments foster circumvention by prosecutors, juries, and judges and thereby produce inconsistencies among cases (Romilly, 1820; Reekie, 1930; Hay, 1975; Tonry, 2009b). Problems of circumvention and inconsistent application have long been documented and understood.
To illustrate this point with modern experience, we draw on the findings of the American Bar Foundation’s Survey of the Administration of Criminal Justice in the United States, which was conducted in the 1950s. According to Frank Remington, director of the project, “Legislative prescription of a high mandatory sentence for certain offenders is likely to result in a reduction in charges at the prosecution stage, or if this is not done, by a refusal of the judge to convict at the adjudication stage. The issue ….
thus is not solely whether certain offenders should be dealt with severely, but also how the criminal justice system will accommodate to the legislative charge” (Remington, 1969, p. xvii). Newman (1966, p. 179) describes how Michigan judges dealt with a lengthy mandatory minimum sentence for drug sales: “Mandatory minimums are almost universally disliked by trial judges…. The clearest illustration of routine reductions is provided by reduction of sale of narcotics to possession or addiction…. Judges … actively participated in the charge reduction process to the extent of refusing to accept guilty pleas to sale and liberally assigning counsel to work out reduced charges.” Newman (1966, p. 182) tells of efforts to avoid 15-year mandatory maximum sentences: “In Michigan conviction of armed robbery or breaking and entering in the nighttime (fifteen-year maximum compared to five years for daytime breaking) is rare. The pattern of downgrading is such that it becomes virtually routine, and the bargaining session becomes a ritual. The real issue in such negotiations is not whether the charge will be reduced but how far, that is, to what lesser offense” (Newman, 1966, p. 182). Dawson (1969, p. 201) describes “very strong” judicial resistance to a 20-year mandatory minimum sentence for the sale of narcotics: “Charge reductions to possession or use are routine. Indeed, in some cases, judges have refused to accept guilty pleas to sale of narcotics, but have continued the case and appointed counsel with instructions to negotiate a charge reduction.”
Many individuals committing offenses targeted by mandatory punishments do, of course, receive them, but others on whose behalf officials circumvent the laws do not. Mandatory punishments transfer dispositive discretion in the handling of cases from judges, who are expected to be nonpartisan and dispassionate, to prosecutors, who are comparatively more vulnerable to influence by political considerations and public emotion.6 The following subsections review sentencing policy initiatives in the second phase of change in modern sentencing law.
The term “truth-in-sentencing,” a 1980s neologism, alludes to federal “truth-in-lending” laws of the 1970s that required consumer lenders and merchants to disclose interest rates and other key financing terms. The implication is that there is something untruthful about parole release and other mechanisms that allow discretionary decisions about release dates
6The evidence suggests that changes in sentencing laws have only short-term effects on the probability of plea-bargaining versus going to trial. Once the system adjusts to new standards, usually within 1 year or 2, traditional patterns reemerge (Feeley, 1983; Tonry, 1996, Chapter 5).
to be made. Under the indeterminate sentencing systems that pervaded the United States before 1975, however, there was nothing unwarranted or untruthful about parole release. The system was meant to allow tailoring of prison terms to the rehabilitative prospects and other circumstances of individuals. Maximum sentences—for example, in the American Law Institute’s (1962) Model Penal Code—were not meant to indicate how long individuals should remain in prison but by what final date they must be released.
Policy advocates in the second phase of sentencing reform, however, defined the differences between the sentences announced by judges and the time served by prisoners as a problem that needed fixing. For example, U.S. Attorney General William Barr, writing a preface to a U.S. Department of Justice (1992) report titled The Case for More Incarceration, for example, argued that “prison works,” urged that the number of people in prison be increased, and proposed a major national program of prison construction. Barr emphasized that most prisoners were released before their maximum sentences expired, pointed out that some committed offenses after release that would not have occurred had they been locked up, and implicitly urged that discretionary parole release be abandoned as a way to achieve more incarceration.7
Proposals like Barr’s were later enacted in the Violent Crime Control and Law Enforcement Act of 1994. The act authorized $8 billion for distribution to states to pay for the construction of additional prisons, although much less was ultimately appropriated.8 To qualify for a substantial portion of these funds, states had to demonstrate that violent offenders would be required to serve at least 85 percent of the sentence imposed. Twenty-eight states and the District of Columbia satisfied this and the other federal criteria (Sabol et al., 2002, Table 1.3).
Evaluators at the Urban Institute sought to determine how truth-in-sentencing laws affected sentencing patterns and prison populations. They were unable “to draw general conclusions about the effects of truth-in-sentencing on sentencing practices throughout the nation” (Sabol et al., 2002, p. vi), but found that the laws had large projected effects in some of the seven states they examined closely. When implemented as part of a comprehensive change to the sentencing system, “truth-in-sentencing laws were associated with large changes in prison populations.” In one state, “the increase in the percentage of sentences required to be served before
7Parole abolition was also a goal of policy advocates in the first sentencing reform phase but for different reasons—because parole release disparities were unfair to prisoners and frustrated achievement of the goals of consistency and proportionality in sentencing (von Hirsch and Hanrahan, 1979). Sixteen states abolished parole for those reasons from the 1970s through the 1990s.
8The average annual state grant was $7,885,875, which U.S. Department of Justice officials estimated would pay for construction of space for 50 prisoners (Sabol et al., 2002, p. 28).
release led to larger increases in length of stay and consequently a larger effect of length of stay on the expected number of prisoners” (Sabol et al., 2002, p. vii).
In the seven case study states, the percentages of terms to be served under truth-in-sentencing were much higher than the actual percentages of sentences served by prisoners released in 1993 and the estimated percentages for those entering prison in 1991, as Table 3-1 shows. In most cases, the percentages at least doubled. The Urban Institute evaluators observed that the effects on the prison population would have been much greater had violent crime rates not fallen substantially after 1991: “Were the sentencing practices of 1996 to persist during a time when the number of violent offenses increases, the impacts on prison populations and corrections management could be dramatic” (Sabol et al., 2002, p. 31).
The RAND Corporation carried out another federally funded assessment of the effects of the federal truth-in-sentencing initiative (Turner et al., 2001). The assessment covered data only through 1997. Even so, the authors concluded, “We do know that nationwide, the imposed maximum sentence length, the average length of prison term, and the percent of term served for violent offenses have increased for TIS [truth-in-sentencing] states between 1993 and 1997. For non-TIS states, sentence lengths have been dropping, and months served have dropped slightly” (Turner et al., 2001, p. 134).
A 50-state analysis by the Vera Institute of Justice looked at the prison population effects of a wide range of sentencing policy changes (Stemen
TABLE 3-1 Actual and Estimated Percentages of Sentences Served Prior to Enactment of Truth-in-Sentencing and Percentages Expected to Be Served Under Truth-in-Sentencing, Seven Case Study States
|State||Percentage of Sentence Served by Those Released from Prison During 1993||Estimated Percentage for Those Entering Prison During 1991||Expected Percentage Under Truth-in-Sentencing|
NOTES: Percentages marked by an asterisk refer to minimum sentences; all others refer to maximum sentences.
SOURCES: Ditton and Wilson (1999); Sabol et al. (2002, Table 3.3).
et al., 2006). Truth-in-sentencing laws were included among a variety of changes that increased time-served requirements for violent crimes. The authors found that “states with separate time served requirements for violent offenders had higher incarceration rates than other states” (Stemen et al., 2006, p. iii).
Concluding one of the most comprehensive 50-state analyses of the effects of the changes in sentencing law of the past four decades, Spelman (2009, p. 59) offers the following observation:
Truth-in-sentencing laws have little immediate effect but a substantial long-run effect. This analysis makes sense: Truth-in-sentencing laws increase time served and reduce the number of offenders released in future years; the full effect would only be observed after prisoners sentenced under the old regime are replaced by those sentenced under the new law.
The authors of the Urban Institute study (Sabol et al., 2002) defined any state that had eliminated the possibility of parole release for some or all prisoners as a “truth-in-sentencing state.” Marvell and Moody (1996) examined the prison population effects of parole abolition and, using 1971-1993 state prison data, found that only 1 of 10 abolition states experienced a higher rate of increase in the prison population than the 50-state average.9 The lowest rates of growth were in Minnesota and Washington. The states included in that study, however, abolished parole release as part of the first phase of modern sentencing reform when no state had enacted a modern truth-in-sentencing law. The early parole abolition initiatives were aimed at greater transparency and in some cases at reductions in unwarranted sentencing disparities. Findings that the early abolitions of parole release operated to restrain growth in prison populations thus are not inconsistent with the findings of the Urban Institute (Sabol et al., 2002), the Vera Institute of Justice (Stemen et al., 2006), RAND (Turner et al., 2001), and Spelman (2009) that truth-in-sentencing laws operated to increase growth. Unlike the truth-in-sentencing initiatives, the earlier parole abolitions typically were not intended to increase the durations of prison sentences.
The Urban Institute, Vera, and RAND studies underestimate the effects of truth-in-sentencing laws on prison population growth because they cover periods ending, respectively, in 1996-1998 (for Ohio), 2002, and 1997. Mandatory minimum sentence, truth-in-sentencing, and three strikes laws requiring decades-long sentences inevitably have a “sleeper” effect. For many years, newly admitted prisoners accumulate; their numbers are not offset by others being released. The ultimate effects of the enactment
9Reitz (2006) concluded that parole abolition states generally had lower rates of prison population increase than parole retention states.
of truth-in-sentencing legislation in the mid-1990s thus are not yet apparent. This is true of many laws mandating decades-long sentences that were enacted during the second phase of sentencing reform. Under the three strikes laws of California and other states mandating 25-year minimum sentences, for example, most of which were enacted during 1993-1996, not a single prisoner’s 25-year term expired by 2014. Under an 85 percent rule, a prisoner serving a 25-year sentence is not eligible for release before 21 years and 3 months. Only after several more years pass will newly admitted prisoners begin to be offset by the release of others admitted decades earlier.
Mandatory Minimum Sentence and Three Strikes Laws
Mandatory minimum sentence and three strikes laws have little or no effect on crime rates, shift sentencing power from judges to prosecutors, often result in the imposition of sentences that practitioners believe to be unjustly severe, and for those reasons foster widespread circumvention.
Between 1975 and 1996, mandatory minimums were the most frequently enacted change in sentencing law in the United States. By 1983, 49 of the 50 states had adopted such laws for offenses other than murder or drunk driving (Shane-DuBow et al., 1985, Table 30). By 1994, every state had adopted mandatory minimum sentences; most had several (Austin et al., 1994). Mandatory minimum sentences apply primarily to drug offenses, murder, aggravated rape, felonies involving firearms, and felonies committed by people who have previous felony convictions.
Knowledge about mandatory minimum sentences has changed remarkably little in the past 30 years. Their ostensible primary rationale is deterrence. The overwhelming weight of the evidence, however, shows that they have few if any deterrent effects. Analyses finding deterrent effects typically observe, as we do in Chapter 5, that existing knowledge is too fragmentary or that estimated effects are so small or contingent on particular circumstances as to have no practical relevance for policy making.
Modern findings on case processing under mandatory minimum sentence laws are consistent with the findings of the American Bar Foundation Survey and the historical studies cited above. The evidence is overwhelming that practitioners frequently evade or circumvent mandatory sentences, that there are stark disparities between cases in which the laws are circumvented and cases in which they are not, and that the laws often result in the imposition of sentences in individual cases that everyone directly involved believes to be unjust. The evidence concerning case processing comes primarily from six major studies (Beha, 1977; Joint Committee on New York Drug Law Evaluation, 1978; Rossman et al., 1979; Loftin et al., 1983; McCoy and McManimon, 2004; Merritt et al., 2006). All found that prosecutors and judges (and sometimes police) in many cases changed their practices to
avoid the imposition of newly enacted mandatory minimum sentences, that prescribed harsher punishments were imposed in the remaining cases, and that overall the laws had few effects on conviction rates.10
To illustrate, New York State’s Rockefeller Drug Laws required lengthy mandatory minimum sentences for a wide range of drug offenses. With great publicity, the legislature authorized and funded 31 new courts to handle drug cases and expressly forbade some forms of plea bargaining. Practitioners made vigorous efforts to evade the mandatory sentences and often succeeded; the remaining cases were dealt with as the law dictated (National Research Council, 1983, pp. 188-189). Drug felony arrests, indictment rates, and conviction rates all declined after the law took effect. For those convicted, the likelihood of being imprisoned and the average length of prison term increased. But the likelihood that a person arrested for a drug felony would be sent to prison remained the same after the law took effect—11 percent—as before (Joint Committee on New York Drug Law Evaluation, 1978).
Massachusetts’ Bartley-Fox Amendment required imposition of a 1-year mandatory minimum prison sentence, without suspension, furlough, or parole, for anyone convicted of unlawful carrying of an unlicensed firearm. Two major evaluations of the law’s effects were conducted (Beha, 1977; Rossman et al., 1979), as well as an ambitious secondary analysis of the data produced by those two studies (Carlson, 1982). The primary findings were that police altered their behavior, becoming more selective about whom to frisk, making fewer drug offense arrests, and seizing many more weapons without making an arrest; charge dismissals and acquittals increased significantly; and the percentage of defendants who entirely avoided a conviction rose from 53.5 to 80 percent.
The Michigan Felony Firearms Statute created a new offense of possessing a firearm while engaging in a felony, and specified a 2-year mandatory prison sentence that could not be suspended or shortened by release on parole and had to be served consecutively with a sentence imposed for the underlying felony. The Wayne County prosecutor established and enforced a ban on plea bargaining and launched a major “One with a Gun Gets You Two” publicity campaign. Findings on the statute’s effects paralleled those of the above studies. Sizable increases in dismissals occurred; the probability of conviction declined; and the probability of imprisonment did not increase, but lengths of sentences increased for those sent to prison. Cases often were resolved by means of an adaptive response, the “waiver trial,”
10See also Crawford et al. (1998), Crawford (2000), Ulmer et al. (2007), and U.S. Sentencing Commission (1991) for a discussion of habitual offender laws in Florida and mandatory minimum sentences in Pennsylvania and in the federal courts, and of how prosecutors often do not file charges that trigger these sentences.
in which the judge would convict the defendant of a misdemeanor rather than the charged felony or, with the prosecutor’s acquiescence, acquit the defendant on the firearms charge. Another avoidance technique was to decrease by 2 years the sentence that otherwise would have been imposed and then add back the mandatory 2-year increment (Heumann and Loftin, 1979; Loftin et al., 1983).
Oregon’s Measure 11, adopted by referendum in 1994, required the imposition of mandatory minimum prison sentences ranging from 70 to 300 months for anyone convicted of 16 designated crimes (and eventually 5 more). RAND Corporation evaluators hypothesized that judges and lawyers would alter previous ways of doing business, especially in filing charges and negotiating plea bargains, to achieve results they deemed sensible and just. The evaluators expected that relatively fewer people would be convicted of Measure 11 offenses and more of non-Measure 11 offenses and that those convicted of Measure 11 offenses would receive harsher sentences. Their research confirmed these hypotheses. Sizable changes were observed in charging decisions (fewer Measure 11 crimes, more lesser crimes) and plea bargaining (fewer pleas to initially charged offenses, more pleas to lesser included offenses) (Merritt et al., 2006).
New Jersey’s truth-in-sentencing law required those affected to serve 85 percent of their announced sentence. This was not a mandatory minimum sentence law, but similar hypotheses apply: that charging and bargaining patterns would change to shelter some defendants and that sentences would be harsher for those not sheltered. Both hypotheses were confirmed (McCoy and McManimon, 2004).
Truth-in-sentencing and mandatory minimum sentence (including three strikes) laws are difficult to reconcile with any mainstream, or even coherent, theory of punishment, as the discussion in the next section shows. Many of the laws require sentences that are highly disproportionate to sentences received by prisoners convicted of other offenses and, as we show in Chapter 5, cannot be justified on the basis of their crime prevention effects.
We now step back from this period of policy turbulence and shifting objectives to assess the changes detailed in this section against three yardsticks—the principles of justice that underlie ideas about punishment in Western thought, the role of scientific evidence in the adoption of sentencing policies, and the unprecedented racial disparities that have resulted from the past four decades of policy changes.
Reasonable people, including members of this committee, hold differing views on the purposes and goals of sentencing and punishment. We believe it is important to discuss principles of justice in relation to criminal
punishment not to promote any particular view or set of views, but to make four points.
The first is that normative principles of justice are relevant to deciding whether a sentencing policy or a decision in an individual case is justifiable and appropriate. Criminal punishment is the paradigm instance of conflict between the interests of the state and those of the individual; criminal convictions can result in losses of property, liberty, and life. Few people want such decisions to be made casually, arbitrarily, or capriciously. That this is so can be seen by recognizing how any individual, law-abiding or not, would want criminal charges against himself or herself handled—evenhandedly, fairly, and justly. Principles of justice are inherently germane to thinking about punishments meted out for crime. The second and third points concern core ideas that recur in coherent sets of views about just punishments—that punishments should ordinarily be proportionate to the severity of crimes and that they should not be more severe, or cost more to administer, than makes sense in relation to the goals they are intended to achieve. These ideas are often (as in the guiding principles articulated in Chapter 1) referred to as the principles of “proportionality” and “parsimony.”11 The fourth point is that proportionality and parsimony have long been widely recognized as important considerations in punishment in all Western countries, including the United States.
Considerations of proportionality and parsimony have fallen into neglect in the United States. Many laws enacted in the 1980s and 1990s required less serious crimes to be punished more severely than more serious ones. Examples include mandatory minimum sentence laws requiring longer terms for people convicted of small sales of drugs than terms typically imposed for many violent offenses, and the sentencing of people to 25-year minimum terms for property misdemeanors under California’s three strikes law. Such laws violate the fundamental principle that punishments should be proportionate to the seriousness of crimes. Other laws mandating prison sentences vastly longer than can be justified by their crime prevention effects violate the principle of parsimony.
Proportionality has been a requirement of every mainstream normative theory of punishment since the Enlightenment. Retributivists, who believe that those who commit offenses deserve to be punished for moral reasons, also believe that punishments must be proportional to the seriousness of crimes. If, for example, shoplifting were punished more severely than robbery or rape, the law on its face would send the perverse moral message that
11In earlier times, as in the Model Sentencing Act of the Advisory Council of Judges of the National Council on Crime and Delinquency (1972), parsimony often was referred to as “the least restrictive alternative” principle: if several possible punishments would achieve their goals equally well, the least restrictive or costly one should be used.
shoplifting is the most serious of the three offenses. Punishing a street-level seller of a few grams of an illicit substance more harshly then someone who commits a violent offense likewise implies that an act of violence is less serious or important than a small sale of drugs.
As an idea and as a term of art, proportionality is commonly associated with retributivist views.12 Proportionality, however, is not just a retributive value. Some form of proportionality is a major component of all mainstream theories of punishment. Consequentialists, who believe that punishments can be justified by their crime prevention or other good effects, also endorse a conception of proportionality (Frase, 2009).13 They typically believe that punishment can be justified if the suffering imposed on a convicted individual prevents greater suffering by others. Thus for consequentialists, punishments should be proportional to the good effects they will produce. Punishments more severe than is necessary to achieve those effects waste public resources and impose suffering for no good purpose.
Some people, probably most, subscribe to mixed theories in which punishments can be justified by their crime prevention effects, but only if they do not exceed what would be warranted by the seriousness of the crime. That is, retributive ideas about deserved punishment set upper limits on what can justly be done to a particular individual, but anticipated crime prevention effects may be appropriate considerations in deciding what to do within those limits (e.g., Morris, 1974; Tonry, 1994).14
Restorative justice theories typically take the same position, although based on different reasoning. John Braithwaite, the most influential restorative justice theorist, offers a negative retributivist account. Proportionality per se, he argues, is not important. The important objectives are to treat offenders and victims with respect and concern and to try to repair broken or damaged relations among the victim, the offender, and the community. If restorative processes culminate in unanimous agreement among participants on substantially different consequences for offenders in comparable
12Modern retributivist theorizing dates from the nineteenth-century writings of Kant (1965) and Hegel (1991). Modern theories differ in details but agree on the core propositions that offenders deserve to be punished for moral reasons and that punishments should be proportionate to the degree of wrongdoing. Ashworth and colleagues (2009, Chapter 4) and Tonry (2011b, Part II) survey contemporary theories and theorists.
13Modern consequentialist theorizing dates from the eighteenth-and nineteenth-century writings of Beccaria (2007) and Bentham (1970, 2008). Modern theories differ in details but agree on the core propositions that punishments must be justified by their beneficial effects and should not be more severe than is required to achieve those effects. Ashworth and colleagues (2009) and Tonry (2011b) survey contemporary theories and theorists.
14Philosophers refer to this as “negative” retributivism (proportionality concerns set maximum but not minimum limits on punishment), in contrast to “positive” retributivism, in which proportionality concerns define the appropriate deserved punishment and thus set both maximums and minimums (Duff, 2001).
cases, then so be it. At the same time, Braithwaite argues, there is a human rights limit—the upper bound of proportionate sentences the justice system might impose (Braithewaite and Pettit, 1990; Braithwaite, 2001).
The ideas just summarized are consistently represented in the philosophical literature as fundamental principles of punishment, but they also reflect widely held beliefs among the general public. There are good reasons to believe that most Americans share the notions that punishments should generally be proportionate to the seriousness of crimes in the retributive sense and not be wasteful or excessive in the consequentialist sense (Roberts and Stalans, 1997). A sizable body of public opinion research, for example, shows that lay people believe punishments should be proportionate to the seriousness of crimes (e.g., Robinson, 2008, 2013), and there is widespread agreement within the United States and other countries about the relative seriousness of different crimes (e.g., Roberts et al., 2003; Darley and Pittman, 2003; Aharoni and Friedland, 2012).
The principles of justice outlined here provide a useful lens through which to evaluate sentencing changes over the past 40 years. Many sentences mandated and imposed under current laws are neither proportionate nor justifiable in terms of their preventive effects. Many street-level drug traffickers, for example, are mandated to receive minimum prison terms of 5, 10, 20, or more years—more severe than punishments received by many people convicted of robbery, rape, or aggravated assault. These laws violate retributive ideas about proportionality given that the general public typically views robberies, rapes, and aggravated assaults as more serious than most drug sales and deserving of greater punishment (Robinson, 2008). Nor can such laws be justified in consequentialist terms. Most drug policy analysts agree that, as discussed further below, imprisoning individual drug dealers seldom reduces the availability of drugs or the number of traffickers (Caulkins and Reuter, 2010; Kleiman et al., 2011).
Some three strikes laws—for example, California’s—mandate lengthier sentences for some property and drug offenses than are required for violent offenses. These laws violate retributive ideas about proportionality; few people believe property and drug crimes, even when repeated, are more serious than violence. These laws also fail consequentialist tests. If the goal is deterrence, then it makes little sense to threaten harsher penalties for theft or a small-scale drug sale than for rape; to do so implies that rape is a less serious offense. If the goal is incapacitation, then it makes little sense to protect the community by confining those convicted of drug or property offenses longer than those convicted of violent ones. If the goal is rehabilitation, then it makes little sense to use longer prison terms and incur greater expense to treat those convicted of property offenses compared with those convicted of violent offenses. If the goal is reinforcing norms, clarifying values, or reassuring the public, then it makes little sense to undermine norms
and obfuscate values by suggesting that theft is more serious than rape, or to imagine that such perverse messages will reassure the public.
We have summarized these principles to provide a normative framework for thinking about the policies that led to high rates of incarceration in the United States. In the committee’s view, many of the nation’s policy decisions that have contributed to high rates of incarceration are inconsistent with the principles of parsimony and proportionality. In Chapter 12, we argue for a reaffirmation of these fundamental and widely supported principles in setting punishment policies in the future.
Social science evidence has had strikingly little influence on deliberations about sentencing policy over the past quarter century. Many factors combined to increase sentence lengths in U.S. prisons. They include enactment of mandatory minimum sentence, truth-in-sentencing, three strikes, and life without possibility of parole laws; discretionary decisions by prosecutors to charge and bargain more aggressively and by judges to impose longer sentences; and decisions by parole boards to hold many prisoners longer, deny discretionary release altogether more often, and revoke parole more often. Some of these decisions were premised on beliefs or assumptions about deterrence, incapacitation, or both. From a crime control perspective, those beliefs and assumptions were largely mistaken (see Chapter 5).
We acknowledge that the relationship between scientific knowledge and policy making is complex, as a specialized literature on “research utilization” has long made clear (e.g., Cohen and Lindblom, 1979). A 1978 National Research Council report, Knowledge and Policy: The Uncertain Connection, notes that numerous social science studies of policy interventions had by then accumulated and that numerous efforts had been made to increase their relevance to and use for policy making. But the report observes that “we lack systematic evidence as to whether these steps are having the results their sponsors hope for …” (National Research Council, 1978b, p. 5). The committee responsible for a subsequent National Research Council report, Using Science as Evidence in Public Policy, concluded that the connection between social science knowledge and policy remained “uncertain” and that “despite their considerable value in other respects, studies of knowledge utilization have not advanced understanding of the use of evidence in the policy process much beyond the decades-old National Research Council (1978b) report” (National Research Council, 2012b, p. 51).
Scholars of policy making have long been skeptical of rational models of the relationship between research and policy, of the idea that policy
decisions do or even should flow more or less directly from scientific evidence concerning the likely effects of alternative policy choices. The 2012 National Research Council report observes that “some mixture of politics, values, and science will be present in any but the most trivial of policy choices. It follows that use of science as evidence can never be a purely ‘scientific’ matter … a dependable and defensible reason will not necessarily be used just because it is available. Re-election concerns, interest group pressure, and political or moral values may be given more weight and may draw on reasons outside the sphere of what science has to say about likely consequences” (National Research Council, 2012b, pp. 15, 17).
We do not disagree with the preceding observations, but note nonetheless that consideration of social science evidence has had little influence on legislative policy-making processes concerning sentencing and punishment in recent decades. The consequences of this disconnect have contributed substantially to contemporary patterns of imprisonment.15 Evidence on the deterrent effects of mandatory minimum sentence laws is just one such example. Two centuries of experience with laws mandating minimum sentences for particular crimes have shown that those laws have few if any effects as deterrents to crime and, as discussed above, foster patterns of circumvention and manipulation by prosecutors, judges, and juries (Hay, 1975). Three National Research Council studies have examined the literature on deterrence and concluded that insufficient evidence exists to justify predicating policy choices on the general assumption that harsher punishments yield measurable deterrent effects (National Research Council, 1978a, 1993, 2012a). Nearly every leading survey of the deterrence literature in the past three decades has reached the same conclusion (e.g., Cook, 1980; Nagin, 1998, 2013b; Doob and Webster, 2003). Despite those nearly unanimous findings, during the 1970s, 1980s, and 1990s the U.S. Congress and every state enacted laws calling for mandatory minimum sentences (Shane-Dubow et al., 1985; Austin et al., 1994; Stemen et al., 2006).
15We do not mean to imply that scholars at particular times unanimously subscribed to certain views of what the evidence showed. Wilson (1975) and others (e.g., Bennett et al., 1996) argue that scientific evidence broadly supported many of the sentencing policy changes of the 1980s and early 1990s. However, they represented a minority viewpoint. A claim by Bennett and colleagues (1996), for example, that proposed policies were justified by the existence of youthful “superpredators” was widely repudiated—including recently by a National Research Council panel (National Research Council, 2013). The weight of the evidence supporting the conclusions we offer in this section was clear during the 1980s and 1990s, as is shown by the findings of a series of National Research Council studies (e.g., on deterrence and incapacitation [National Research Council, 1978a]; on criminal careers [National Research Council, 1986]; and on sentencing reform initiatives, including mandatory penalties [National Research Council, 1983]) and elsewhere (e.g., Cohen’s  influential survey of the state of knowledge about incapacitation).
In Chapter 5, we also discuss at considerable length the evidence on the important question of the relationship between high rates of incarceration and crime. That assessment leads to the conclusion that although the growth in incarceration rates may have caused a decrease in crime, the magnitude of the reduction is highly uncertain and the results from most studies suggest that it was unlikely to have been large. The social science evidence available in the 1980s and 1990s would have predicted such a result.
Many features of U.S. criminal justice systems—including unwarranted disparities in imprisonment, invidious bias and stereotyping, police drug arrest practices, and racial profiling16—disproportionately affect blacks and Hispanics (Tonry, 2011a). Table 3-2 shows the most recent available national data on racial disparities in imprisonment, capital punishment, life sentences, and sentences of life without possibility of parole for adults and minors. The disparities are enormous. Racial disparities in imprisonment and the absolute numbers of black people, especially men, now or formerly behind bars are major impediments to the creation of an America in which race does not matter (Alexander, 2010).
Higher rates of black and Hispanic than white imprisonment were demonstrated in Chapter 2. They are partly caused and substantially exacerbated by the mandatory minimum sentence, three strikes, truth-in-sentencing, life without possibility of parole, and similar laws enacted in the 1980s and 1990s. All of these laws mandate especially severe—in recent decades unprecedentedly severe—punishments for offenses for which black and Hispanic people often are disproportionately arrested and convicted.17
16We do not discuss racial profiling by the police in this chapter because the extent to which it significantly contributes to high levels of incarceration is unclear. Police profiling results in many more arrests of black people than would otherwise occur. Research on profiling generally concludes that police stop blacks disproportionately often on sidewalks and streets, but find contraband at lower rates for blacks than for whites (e.g., Engel and Calnon, 2004; Center for Constitutional Rights, 2009; Engel and Swartz, 2013).
17In discussing data on race and ethnicity in this chapter, we sometimes refer to “blacks” and “whites.” At other times, we present data on “Hispanics,” “non-Hispanic whites,” and “non-Hispanic blacks.” The terms used depend on the data sources on which we draw. Prison and jail data published by the Bureau of Justice Statistics (BJS) through 1991 classify people as black and white, with no separate Hispanic category. Since then, national data on jail and prison populations have used a black, white, and Hispanic classification system. National arrest data compiled in the Federal Bureau of Investigation’s Uniform Crime Reports and BJS data on criminal courts and sentencing use only black and white categories, which include Hispanics.
TABLE 3-2 Black, White, and Hispanic Inmates in Prison, on Death Row, and Serving Life Sentences, Recent Years
|Imprisonment, Jaila||Mid-2011||735,601||329,400 (45%)||276,400 (38%)||113,900 (15%)||15,900 (2%)|
|Imprisonment, State and Federalb||2011||1,537,415||516,200 (34%)||581,600 (38%)||349,900 (22%)|
|Death rowc||2010||3,158||1,750 (55%)||1,316 (41%)|
|Life Without Possibility of Parole||2008||13,751||23,181||3,052|
|Life Without Possibility of Parole, Minor||2008||497||984||206|
aJail total includes Alaska Natives, American Indians, Asians, Native Hawaiians, and other Pacific Islanders. White and black subtotals exclude persons of Hispanic or Latino origin.
bImprisonment total includes Alaska Natives, American Indians, Asians, Native Hawaiians, other Pacific Islanders, and persons identifying two or more races; white and black subtotals exclude persons of Hispanic or Latino origin.
cDeath row total includes Alaska Natives, American Indians, Asians, Native Hawaiians, other Pacific Islanders, and Hispanic inmates for whom no other race was identified. White and black inmates include persons of Hispanic or Latino origin.
SOURCES: Nellis and King(2009, life without possibility of parole); Carson and Sabol(2012, prisoners); Snell(2011, death row).
We focus here primarily on disparities affecting blacks, only occasionally adverting to Hispanics, for several reasons.18 The most important is that disparities affecting blacks have long been much more acute than those for any other group. Second, the unique history of slavery, Jim Crow laws, and legally sanctioned discrimination that ended only 50 years ago gives particular salience to patterns of disparate treatment affecting blacks. Third, for the first two reasons, the literature on disparities affecting blacks is vastly larger.
Understanding extraordinary racial disparities in imprisonment is a critical challenge facing the nation. As described in Chapter 4, the political and social context in which current policies unfolded has a pronounced racial dimension. In this section, we discuss three different kinds of racial disparity.
The first concerns differences in the probability that blacks and whites are in prison on an average day. In 2011, for example, the combined federal and state incarceration rate for non-Hispanic black men (3,023 per 100,000) was more than six times higher than that for non-Hispanic white men (478). The Hispanic rate (1,238) was slightly more than two-and-one-half times the white rate (Carson and Sabol, 2012, Table 8).
The second kind of disparity concerns racial differences in rates of imprisonment relative to group differences in offending. People are sent to prison because they are convicted of crimes, so it is natural to ask whether disparities in imprisonment rates correspond to disparities in criminality. In the 1980s and early 1990s, racial differences in arrests appeared to correspond closely to racial differences in imprisonment for serious violent crimes but not for property or drug crimes (Blumstein, 1982, 1993). In the 2000s, racial differences in arrests do not correspond closely to racial differences in imprisonment for violent, property, or drug crimes (Tonry and Melewski, 2008; Baumer, 2010).
The third kind of disparity concerns racial differences in sentencing and case processing after controlling for legally relevant differences among offenses. A sizable literature has long shown and continues to show that blacks are more likely than whites to be confined awaiting trial (which increases the probability that an incarcerative sentence will be imposed), to receive incarcerative rather than community sentences, and to receive longer
18Demographic differences explain in part why imprisonment rates are higher for Hispanics than for non-Hispanic whites (Tonry, 2012). The Hispanic population is much younger, and, consistent with research on age-crime curves, proportionately more Hispanics are in their high-crime ages. In 2008, nearly 44 percent of U.S. Hispanics were under 25, compared with 30 percent of non-Hispanic whites (U.S. Department of Commerce, 2010, Table 10). In 2010, among people arrested for violent crimes, 42.8 percent were under 25 (Maguire, n.d., Table 4.7.2010).
sentences. Racial differences found at each stage are typically modest, but their cumulative effect is significant (Tonry, 2011a; Spohn, 2013).
Racial disparities in imprisonment are of long standing but worsened substantially in the 1980s and early 1990s. For a century before the 1960s, black people had been more likely to be held in prison than whites. As shown in Chapter 2, racial disparities in imprisonment began to rise in the 1960s and reached all-time highs in the 1980s and early 1990s. In recent years, differences in incarceration rates have slightly lessened. In absolute numbers, however, federal and state prisons in 2011 held more non-Hispanic black (581,000) than non-Hispanic white (516,000) inmates. In 2012, 13 percent of U.S. residents were non-Hispanic blacks, and 63 percent were non-Hispanic whites.
The critical question about imprisonment disparities is whether they result from group differences in criminality or from group differences in how cases are handled. If racial disparities in imprisonment perfectly mirrored racial patterns of criminality, then an argument could be made that the disparities in imprisonment were appropriate.19 However, if racial disparities in imprisonment resulted entirely from differences in case processing, then they would violate principles of fairness and equal treatment.
Disparities in imprisonment result from a combination of differences in offending patterns and case processing. Disentangling in detail the respective roles of each is difficult. Some insights can be gained from comparing data from victimization surveys on the characteristics of assailants whom victims can identify, but those data are limited and cover only a small category of offenses. The closest scholars have come is to compare racial patterns of arrests for particular offenses with racial patterns in imprisonment for those offenses. As Table 3-3 shows, racial disparities in imprisonment have worsened substantially since the early 1990s relative to racial patterns of involvement in serious crimes.
A classic and influential analysis of racial disparities in imprisonment in 1979 (Blumstein, 1982) concluded that racial patterns of arrests “explained” a large proportion of the disparities, especially for serious violent
19As Chapter 2 shows, however, group differences in imprisonment are strongly associated with racial and economic differences in education and employment. Important policy issues concerning the sources of those differences and their remediability would remain to be addressed.
TABLE 3-3 Racial Disparities in Imprisonment Not “Explained” by Arrests, 1979-2008
|Offense||1979 (%)||1991 (%)||2004(%)||2008 (%)|
|Murder and Non-negligent Homicide||2.8||–35||11.6||40|
NOTE: “All offenses” includes, in addition to the categories shown, “other violence,” “other property,” “public order,” and “other/unspecified” offenses.
SOURCES: For 1979: Blumstein (1982); for 1991: Blumstein (1993, Table 2); Baumer (2010); for 2004: Tonry (2011a, Table 2.4); for 2008: Baumer (2010).
crimes, and for all offenses left only 20.5 percent “unexplained.” For three serious violent crimes, small fractions of disparities in imprisonment were unexplained: murder and non-negligent homicide (2.8 percent), aggravated assault (5.2 percent), and robbery (15.6 percent). For larceny and auto theft (combined) and drug offenses, nearly half the racial disparity in imprisonment was unexplained.
Blumstein reasoned that if the percentages of black and white people held in prison for a particular offense, say, homicide, closely paralleled black and white percentages among those arrested, it would be reasonable to infer that racial patterns of involvement in crime were the primary reason for disparities in imprisonment. Blumstein’s analysis cannot prove that racial bias and stereotyping had no or little influence on sentencing patterns. He argued, though, that it was reasonable to infer that their influence was relatively small. His conclusions were confirmed by Langan (1985), who used victim data instead of arrests and prison admission data rather than population data. Blumstein’s (1982) conclusions also were confirmed by his subsequent analysis of 1991 data, which found that arrest patterns explained all but 24 percent of overall disparities in imprisonment (Blumstein, 1993).
Arrest data may be potentially misleading indicators of crime to the extent that they are distorted by bias in victims’ decisions to report alleged crimes and in police decisions to record them. Yet there are good reasons to believe that the racial patterns shown by arrest data are reasonably accurate indicators of crimes committed, at least for serious violent crimes. Victims’
descriptions of the racial characteristics of assailants and police data on victim-offender relationships in homicides have for 30 years indicated, at least for serious crimes, that racial offending patterns shown in arrest data do not deviate far from reality (Langan, 1985; Tonry, 2011a, Figure 2.7).
Other, more rigorous methods might be imagined for assessing relationships between racial patterns in crime rates and imprisonment over time at the aggregate national level, but such studies have not been carried out and published. Blumstein’s analysis was widely cited over several decades as providing convincing evidence that bias and stereotyping are not the primary cause of racial disparities in imprisonment. However, replications using data for more recent years have found that arrests explain much lower percentages of imprisonment disparities relative to Blumstein’s early studies. These findings are consistent with data reported in Chapter 2 on the increasing disjunction between racial patterns in crime and in imprisonment. Analyses for 2004 (Tonry and Melewski, 2008) and 2008 (Baumer, 2010) using the same method as that used by Blumstein show that, relative to arrest patterns, racial disparities in imprisonment became much worse in the twenty-first century compared with those found by Blumstein for 1979 and 1991. For 2004, 39 percent of overall disparities in imprisonment could not be explained by reference to arrests, and for 2008, 45 percent. Baumer (2010) concluded that for 2008, 40 percent of disparities in imprisonment for murder, 45 percent for robbery, 55 percent for aggravated assault, and 66 percent for drug offenses could not be explained by arrest patterns.
Different racial patterns of involvement in violent crime thus are part of the reason for disparities in imprisonment, but they can explain neither why disparities increased in the 1970s and 1980s nor why they remain so high in the twenty-first century. First, no significant shifts in racial patterns in arrests for violent crimes occurred in the 1970s and 1980s that could explain why black incarceration rates rose after the 1960s. Second, as discussed in Chapter 2, the relative over involvement of blacks in violent crimes has declined significantly since the 1980s.
The reason for increased racial disparities in imprisonment relative to arrests is straightforward: severe sentencing laws enacted in the 1980s and 1990s greatly increased the lengths of prison sentences mandated for violent crimes and drug offenses for which blacks are disproportionately often arrested. These two offense categories, however, raise different behavioral issues. For reasons of social disadvantage, neighborhood residence, and limited life chances that disproportionately affect them, blacks relative to whites have been more involved in violent crime and are more frequently arrested for such crimes (e.g., Sampson, 1987; Sampson and Wilson, 1995; Land et al., 1990; see Sampson and Lauritsen  for a review). Thus one reason why black Americans are disproportionately affected by tougher sentencing policies for violent crime is that they are more often arrested for
such crimes—even though the black-white difference in these arrest rates has been declining since the 1980s.
For drug crimes, the situation is different. As suggested in Chapter 2, the disproportionate numbers of arrests of black people for drug crimes bear little relationship to levels of black Americans’ drug use or involvement in drug trafficking (e.g., Western, 2006, pp. 41, 45-48; a detailed case study of racial disparity in drug arrests is provided by Beckett and colleagues ). Black people are, however, arrested for drug offenses at much higher rates than whites because of police decisions to emphasize arrests of street-level dealers (Beckett et al., 2005, 2006; Mitchell and Caudy, 2013). Legislative decisions also have specified the longest sentences for crack cocaine offenses, for which blacks are arrested much more often than whites. As the late Senator Daniel Patrick Moynihan (1993, p. 362) observed: “It is essential that we understand that by choosing prohibition [of drugs] we are choosing to have an intense crime problem concentrated among minorities.”
The committee’s review of the literature justifies the conclusion that racial bias and discrimination are not the primary causes of disparities in sentencing decisions or rates of imprisonment. There are differences, but they are relatively small. No doubt they result partly from the various forms of attribution and stereotyping discussed below. Minority defendants are, however, treated differently at several stages of the criminal justice process, and those differences influence resulting disparities. We agree with the National Research Council’s panel on sentencing research that “even a small amount of racial discrimination is a matter that needs to be taken very seriously, both on general normative grounds and because small effects in the aggregate can imply unacceptable deprivations for large numbers of people. Thus even though the effect of race in sentencing may be small compared to that of other factors, such differences are important” (National Research Council, 1983, p. 92).
The empirical literature on sentencing documents relatively small racial differences in the justice system experiences of black and white individuals with comparable criminal records and convicted of the same crime. Blacks and Hispanics are more likely than whites to be detained before trial; as noted earlier, being detained increases the probability that a prison sentence will be imposed (e.g., Demuth and Steffensmeier, 2004; Spohn, 2009). Although the evidence is not entirely consistent, the clear weight of research findings is that race and ethnicity affect charging and plea bargaining decisions in both capital and noncapital cases (Crutchfield et al., 1995; Miller and Wright, 2008; Spohn, 2013).
Black and Hispanic defendants, all else being equal, are somewhat more likely than whites to be sentenced to incarceration, and among those sentenced to incarceration in federal courts to receive somewhat longer sentences (Crutchfield et al., 2010; Spohn, 2013). Blacks are less likely than whites to be diverted to nonincarcerative punishments. In states that have sentencing guidelines, blacks are more likely than whites to receive sentences at the top rather than at the bottom of the guideline ranges (Tonry, 1996). Individual studies present divergent findings, often showing small disparities by race and ethnicity for men but not for women (or to different extents), for Hispanics but not for blacks, and for young but not for older offenders (or in each case vice versa) (e.g., Walker et al., 2006; Harrington and Spohn, 2007, pp. 40-45). Overall, when statistical controls are used to take account of offense characteristics, prior criminal records, and personal characteristics, black defendants are on average sentenced somewhat but not substantially more severely than whites. As noted above, however, small differences in this area matter. Spohn (2013, p. 168) concludes her recent exhaustive survey of disparity research thus: “Whether because of conscious bias, unconscious stereotypes linking race with crime, or colorblind application of racially tinged policies, judges’ and prosecutors’ decisions regarding bail, prosecution, and sentencing are not racially neutral.”
While there is not convincing evidence of widespread racial bias in sentencing, there is, in contrast with several decades ago, credible evidence that black defendants are treated differently. Before 1980, many studies appeared to show systematic bias in sentencing of black defendants, but subsequent analyses concluded that failure to control for legally relevant sentencing factors, such as prior criminal record, seriously undermined the persuasiveness of those findings (e.g., National Research Council, 1983; Hagan and Bumiller, 1983). Reviews of subsequent research, however, concluded that blacks were treated less favorably than whites at a number of stages—for example, in pretrial detention decisions, prosecutorial charging decisions, and decisions to impose community rather than incarcerative punishments—and that the cumulative effect of small differences at each stage was substantial (e.g., Zatz, 1987; Chiricos and Crawford, 1995; Mitchell, 2005). Research on death penalty decisions similarly shows that the race of the victim plays a role in both charging and sentencing decisions (Sorensen and Wallace, 1999; Lee, 2007); this is especially evident in cases of interracial violence (Gross and Mauro, 1989; Baldus et al., 1990).
The finding that discernible racial differences exist in sentencing and case processing is disheartening. Race should not matter when criminal sentences are imposed. Viewed differently, however, the finding is not surprising. Americans of every racial and ethnic group are influenced by stereotypes about black people’s involvement in crime. This is not to say that most Americans are bigoted or racist. Few white Americans still believe in
the racial inferiority of black people, and most believe racial discrimination is wrong. Among earlier generations of white Americans, the belief that blacks are racially inferior to whites was commonplace. Those beliefs largely disappeared after the 1960s, sometimes to be replaced by other unflattering stereotypes (Unnever, 2013). Since the 1970s, large majorities of whites have favored integrated schools, accepted having blacks as neighbors, and believed that blacks and whites are of equal intelligence (Thernstrom and Thernstrom, 1997, pp. 498-501). One typical and detailed survey of research on racial attitudes concluded that Americans’ endorsement of racial equality norms is nearly universal:
Almost all whites genuinely disavow the sentiments that have come to be most closely associated with the ideology of white supremacy—the immutable inferiority of blacks, the desirability of segregation, and the just nature of discrimination in favor of whites. In this sense, nearly every white person today has a genuine commitment to basic racial equality in the public sphere (Mendelberg, 2001, pp. 18-19).
Comprehensive recent surveys of a range of literatures on racial attitudes have reached similar conclusions (e.g., Krysan, 2012).20
Whites, and members of other groups, nonetheless are influenced by racial stereotypes (Kirschenman and Neckerman, 1991). Sociologists use the term “statistical discrimination” to describe the attribution of characteristics of groups to individuals (Wilson, 1987) as when, for example, employers’ preconception that inner-city minority men are less likely than others to be reliable workers leads them to reject reliable applicants (Pager, 2007). These issues are discussed further in Chapter 8.
Several literatures document the existence and force of racial stereotyping about crime and criminals. The media commonly portray a world of black offenders and white victims. When asked to describe typical violent criminals and drug dealers, white Americans often describe black individuals (e.g., Entman, 1992; Reeves and Campbell, 1994; Beckett and Sasson, 2004). Research on the influence of skin tone and stereotypically African American facial features shows that negative stereotypes operate to the detriment of blacks in the criminal justice system. They cause black individuals to be punished more severely than whites, and among blacks they cause dark-skinned people and people with distinctively African American facial
20This does not mean that racial anxieties and attitudes toward criminal justice have ceased to matter. Racial resentments and anxieties are major predictors of whites’ support for harsh sentencing and punishment policies and their opposition to increased public expenditure on social welfare programs (Bobo and Johnson, 2004; Bobo and Thompson, 2006; Peffley and Hurwitz, 2010; Unnever, 2013).
features to be punished more severely than light-skinned people and people with more European features.
This form of stereotyping, known as “colorism,” places darker-skinned American blacks at a comparative disadvantage in most spheres of life (Hochschild and Weaver, 2007).21 Dark skin evokes fears of criminality (Dasgupta et al., 1999) and is an easily remembered characteristic of a purportedly criminal face (Dixon and Maddox, 2005). For example, an analysis of more than 67,000 male felons incarcerated in Georgia showed that controlling for type of offense, socioeconomic characteristics, and demographic factors, dark-skinned blacks received longer sentences than light-skinned blacks: light-skinned black defendants received sentences indistinguishable from those of whites, while longer sentences were received by medium-skinned (a year longer on average) and dark-skinned (a year and a half longer on average) black defendants (Hochschild and Weaver, 2007, p. 649).
Studies of Afrocentric feature bias take the analysis one step further (Blair et al., 2004). The evidence confirms the hypothesis that stereotypically African American facial features (e.g., dark skin, wide nose, full lips) influence decision makers’ judgments (Blair et al., 2002, 2005; Eberhardt et al., 2004). Pizzi and colleagues (2005, p. 351) measured facial features of black and white defendants and concluded that practitioners treated differently not only black but also white defendants with such features:
Racial stereotyping in sentencing decisions still persists. But it is not a function of the racial category of the individual; instead, there seems to be an equally pernicious and less controllable process at work. Racial stereotyping in sentencing still occurs based on the facial appearance of the offender. Be they white or African American, those offenders who possess stronger Afrocentric features receive harsher sentences for the same crimes.
Even death penalty decisions are influenced by facial features. Looking at cases in Philadelphia in which death had been a possible sentence, Eberhardt and colleagues (2006, p. 383) “examined the extent to which perceived stereotypicality of black defendants influenced jurors’ death-sentencing decisions in cases with both white and black victims.” With stereotypicality as the only independent variable, 24.4 percent of black defendants rated below the median in having stereotypical black features
21Colorism is defined as the “tendency to perceive or behave toward members of a racial category based on the lightness or darkness of their skin tone” (Maddox and Gray, 2002, p. 250). Empirical research on the subject is comparatively new, but the phenomenon is old. Seventy years ago, Myrdal (1944, p. 697) observed in An American Dilemma: The Negro Problem and Modern Democracy, “Without a doubt a Negro with light skin and other European features has in the North an advantage with white people.”
were sentenced to death, compared with 57.5 percent of those rated above the median.
The Implicit Association Test (IAT),22 which has been taken by millions of people, was developed by psychologists to assess people’s attitudes toward members of different groups. The IAT results have consistently shown that implicit bias against blacks is “extremely widespread” (Jolls and Sunstein, 2006, p. 971) and demonstrate the existence of unconscious bias by whites against blacks (Rachlinski et al., 2009).23 It would be remarkable if criminal justice practitioners were not affected by this bias.24
A number of lessons emerge from this look back at the past four decades of changes in sentencing policy. Successive waves of change swept the nation, some affecting all or most states. During the 1970s, experiments with voluntary sentencing guidelines were undertaken in many states, and all but one state enacted mandatory minimum sentence laws typically requiring minimum 1-or 2-year sentences or increases of 1 or 2 years in the sentences that would otherwise have been imposed. During the 1980s, the federal government and nearly every state enacted mandatory minimum sentence laws for drug and violent crimes, typically requiring minimum sentences of 5, 10, and 20 years or longer. During the 1990s, the federal government and more than half the states enacted truth-in-sentencing and three strikes laws. Almost all of the states now have life without possibility of parole laws. Voluntary guidelines and statutory determinate sentencing laws proved ineffective at achieving their aims of increasing consistency and diminishing racial and other unwarranted sentencing disparities. There is
22The IAT asks individuals to categorize a series of words or pictures into groups. Two of the groups—“black” and “white”—are racial, and two are characterizations of words as “good” or “pleasant” (e.g., joy, laugh, happy) or “bad” or “unpleasant” (e.g., terrible, agony, nasty). To test for implicit bias, one version of the IAT asks respondents to press one key on the computer for either “black” or “unpleasant” words or pictures and a different key for “white” or “pleasant” words or pictures. In another version, respondents are asked to press one key for “black” or “pleasant” and another key for “white” or “unpleasant.” Implicit bias is defined as faster responses when “black” and “unpleasant” are paired relative to “black” and “pleasant.”
23People taking the IAT at the Project Implicit website are regularly warned that they may find the results of their own test disturbing: “Warning: This test has been taken more than one million times, and the results usually reveal some degree of bias” (http://www.understandingprejudice.org/iat/ [February 28, 2014]).
24Almost all demographic groups show a significant implicit preference for whites over blacks. The major exception is blacks: equal proportions show implicit preferences for blacks and for whites, but unlike whites they do not show a preference for their own group. The consensus view of the existence of implicit racial bias is based on the results of millions of tests of every imaginable group in the population.
little convincing evidence that mandatory minimum sentencing, truth-in-sentencing, or life without possibility of parole laws had significant crime reduction effects. But there is substantial evidence that they shifted sentencing power from judges to prosecutors; provoked widespread circumvention; exacerbated racial disparities in imprisonment; and made sentences much longer, prison populations much larger, and incarceration rates much higher.
The policy initiatives that swept the nation were by and large ineffective at creating just, consistent, and transparent sentencing systems. The more targeted approaches—parole and presumptive sentencing guidelines, especially when incorporating prison capacity constraints—were effective. Both parole and presumptive sentencing guidelines, when well designed and implemented, can demonstrably improve consistency, reduce disparity, and make these critical decisions more transparent. Presumptive sentencing guidelines incorporating prison capacity constraints offer a proven method for setting sentencing priorities, minimizing disparities, controlling prison population growth, and managing correctional budgets.
The evidence discussed in this chapter points to four main findings.
First, law reform initiatives aimed at achieving greater fairness, consistency, and transparency in sentencing have achieved their goals more successfully than initiatives aimed at achieving greater severity, certainty, and crime prevention.
Second, social science evidence on the effectiveness of sanctions and the operation of the justice system informed the development of parole and sentencing guidelines but had little influence on the development of initiatives aimed at achieving greater severity, certainty, and crime prevention. The evidence base on sentencing is broader and deeper now than in the 1980s and 1990s, but the primary findings have not changed significantly since they were disseminated in a series of National Research Council reports between 1978 and 1986.
Third, initiatives aimed at achieving greater severity, certainty, and crime prevention were largely incompatible with fundamental and widely shared ideas about just punishment that have characterized the United States and other Western countries since the Enlightenment. Many of the punishments imposed under the new laws have violated the principle of proportionality—that punishment should be proportionate to the individual’s culpability and the gravity of the offense. Many also have violated the principle of parsimony—that punishments should be no more severe than is required to achieve their legitimate purposes.
Fourth, racial and ethnic disparities in imprisonment reached extreme and unprecedented levels in the 1980s and 1990s and have since remained at deeply troubling levels. They are partly caused and significantly exacerbated by recent sentencing laws aimed at achieving greater severity, certainty, and crime prevention and by law enforcement strategies associated
with the war on drugs. They also result partly from small but systematic racial differences in case processing, from arrest through parole release, that have a substantial cumulative effect. And they are influenced by conscious and unconscious bias and stereotyping that remain pervasive in America despite the near disappearance of widespread beliefs about racial superiority and inferiority.