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

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

5 Sentencing Policies and Their Impact on Prison Populations One of the important consequences of changes in sentencing policies is their impact on prison populations. This issue is especially important at a time when prisons are increasingly crowded. However, both short- term and longer-term perspectives on prison populations must be con- sidered in policy making, since population projections suggest a decrease in the number of prisoners by the end of the 1980s in a number of states. The size of the prison population is shaped by the number of offenders committed, the length of their sentences, and the time they actually serve. These, in turn, may be affected by demographic changes in the population, changes in demographic-specific crime rates, legislatively established sentencing policies, police and prosecutorial policies, judicial decision-making practices, the exercise of authority by prison officials in awarding and revoking good time, and parole boards' release and revocation policies. The panel examined the relationship between sentencing policies and prison populations because anticipation of the impact on prison of ex- isting and alternative sentencing policies makes explicit the choices among levels of punitiveness and their costs and is an important aid to respon- sible policy making. In this chapter we examine recent changes in prison populations and the implications of these changes for the health and safety of inmates and correctional staff; the methods, uses, and limitations of projections of future prison populations for policy making; and alternative strategies for coping with prison populations that have grown beyond the capacities of existing prison facilities. 225

226 RESEARCH ON SENTENCING THE SEARCH FOR REFORM CHANGES IN PRISON POPULATIONS AND THEIR IMPLICATIONS INCREASES IN PRISON POPULATIONS Except during World War II, American prison populations increased at about the same rate as the civilian population from 1930 until the early 1970s, when a dramatic increase began in the number of prisoners in- carcerated in federal and state prisons (see Figure 5-1~. The rate of incarceration in state and federal prisons rose 62 percent between 1972 and 1981: from 95 per 100,000 population to 154 per 100,000 popula- tion.i Between the end of 1972 and the end of 1978, the federal and state prisoner population sentenced for more than 1 year increased by about 50 percent, from 196,183 to 294,580 (U.S. Department of Justice, 1975, 1976, 1977, 1978b, 1979~. Since then, sentenced state and federal prisoner populations have continued to rise to 352,476 at the end of 1981, another 19 percent in 3 years (U.S. Department of Justice, 1982a). The sharpest inmate increases occurred in state prisons, which hold about 60 percent of all offenders. Between 1972 and 1981, net state prison populations increased by 89 percent, from 174,470 to 330,307 inmates. Between 1939 and 1970 the median state prison incarceration rate was 98.8 per 100,000 civilian population; in 1972 this rate had fallen to 84, but by 1978 it had risen to 124, an increase of 48 percent in 6 years; by December 31, 1981, it had climbed to 144, a further increase of 16 percent in 3 years (U.S. Department of Justice, 1982a). Increases in state prison population were most pronounced in the South. For decades the rate of incarceration in the South was higher than in other regions, and the gap grew during the 1970s. Between 1970 and 1978 state prison populations grew by 84.1 percent in the South, while they increased by 41 and 44 percent, respectively, in the North and North Central states, and by only 8.6 percent in the West (see Table 5-14. In 1970, the South accounted for 39 percent of all state prisoners; by 1978, that number was 48 percent. The greater increases in the South far outpaced population increases there: the incarceration rate increased 63 percent in the South and only 43 percent for the nation as a whole. Though federal prisoners make up only 6 percent of the national prisoner total, recent changes in federal prison populations illustrate the ~ Rates of incarceration are computed as the ratio of inmates in a jurisdiction for every 100,000 civilian population in that jurisdiction as estimated by the Bureau of the Census. State rates vary due both to different policies regarding incarceration and differences in accounting practices; for details, see Mullen and Smith (1980:11~.

227 Go Ct of m o ._ Ct i_ o ._ Ct - o o Cal ._ Pa Ct V) 1 Lo m ~ ~ O^ ~ ~ "0 =3 =~= o o o Ct ~ Go ~ ~ Co ~ Coo o Y o ~ Ct ~ ~ ~ AL al G C ~ =0 to . _ O Ct . _ P" o o ._ ._ O o o ._ ._ O CL O .0 on C) + + + 1 + Go O of Cal C~ ~t ~ ~ t— ~ ~ ~ o oo ~ ~4 ~ ~ ~ ~ o . . . . . ~ c~ ~ oo ~ ~ ~ oo ~ + + + + + oo O cr~ ~ ~ o ~ oo ~ ~ ~ ~ - ~ o o oo ~ oo ~ v u~ ~ o ~ ~ ~ ~ oN ~ o u~ c~ ~ ~ ~ oo ~ c~ ~ ~ ~ - ~o ~ o ' ~ ~ & ~ ~ ~ ,= ~ _ C<, ~ ~ ~ ~ ~ Ct ~ o Z Z cn 3 ~ Ct - o Ct - s~ Ct =.~ o ~ 3 z ~ o ~ ^= ,= ~3 _ ~ Ct C) o ;^ U' ~ Ct o s :<s =- o ~ o ~ ~ ~ 3 =^Z ~ z ~ ~ ~ =^ Z o °G ;^Z'~ O-~ Ct > Ct ~ O ~ ~ ~ '~ O z ^~3 o~ ed ~<: ~ ~ ~ ~ e~ ~ 3 ^~. ~ ~ Z,`-~ ct 0 '- ~ ~ ~ 0 O w C: ~) ~ ^ 3 ^ ~ ce _ ~ ~ ~ ,~ ·0 ..D ¢ ~4 ._ ;` 3 ~: o oo ._ U7 Ct 3 g oc s~ o ~_ o . . o oo o~ _' - Ct o - ~s . . ,~ o C~

228 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM is 160 o ~ 1 40 cat 8 Of a in a: 1 00 z u' 80 60 1 \~/ 01 1920 1940 1950 1960 1970 1980 YEAR FIGURE 5-1 Annual imprisonment rate in the United States: 193~1981. SOURCES: Adapted from Blumstein and Cohen (1973:203), Carlson et al. (1980:114), and U.S. Department of Justice 2~1982a). impact of changes in prosecutorial policy on prison populations. Be- tween 1977 and 1980, federal prison populations dropped, principally due to a change in emphasis in the Justice Department that sharply reduced prosecution of auto theft and bank robbery cases and increased resources for prosecution of white-collar crime, major narcotics viola- tions, organized crime, and political corruption cases, all of which take longer to convict and result in shorter sentences. PRISON CAPACITY AND CONDITIONS OF CONFINEMENT The dramatic increases in prison population have not been accompanied by corresponding increases in prison capacity, resulting in overcrowding and a decline in living standards in prisons. The decline in prison pop-

Impact on Prison Populations 229 ulations through most of the 1960s led to selective closing of facilities and a virtual halt in the construction of new prison facilities. When the downward trend was reversed and prison populations increased by more than one-third between 1972 and 1976, crowding became common, and the conditions in many older facilities became a cause for many legal actions. The greater activism of courts in addressing conditions of con- finement in prison as a constitutional issue and emerging professional standards and accreditation procedures2 have limited legally acceptable options in dealing with population pressures and created a critical prob- lem in many corrections systems. By March 1982, prison authorities in 28 states and the District of Columbia were operating under court orders arising from violations of the constitutional rights of prisoners related to the conditions of confinement or overcrowding, including sweeping orders covering entire correctional systems in 10 states. In addition, according to the American Civil Liberties Union.Foundation, legal chal- lenges to major prisons were pending in 9 other states (Criminal Justice Newsletter, 1982~. A congressionally mandated national study of prison inmates and facilities found that 61 percent of federal prisoners, 65 percent of state prisoners, and 68 percent of prisoners in local jails had less than 60 square feet of floor space, the minimum standard promulgated by the Commission on Accreditation for Corrections (Mullen et al., 1980:61, 75~. Because definitions of capacity vary widely, there is no single or clear estimate of the number of inmates that can be held in existing state and federal prisons, and the study used three different measures of capacity. Using the reported capacity measure, the study found that state and federal prisons had slightly more than sufficient capacity to hold all inmates confined in 1978 in state and federal prisons together (i.e., 96 percent of all space was used). By the measured capacity stand- ard- one inmate per cell of any size or, for dormitories, the smaller of (1) the number of square feet of floor space/60 or (2) the jurisdictionally reported capacity state and federal prisons were operating at 115 per- cent of capacity. By the most stringent physical capacity standard, de- 2 In 1974 the American Correctional Association established the Commission on Ac- creditation for Corrections to develop a set of uniform standards that would provide measurable criteria for assessing the safety and well-being of correctional system inmates and staff. It has published 10 volumes of standards, including Adult Correctional Insti- tutions and Adult Local Detention Facilities (Commission on Accreditation for Corrections, 1977a,b). Other standard-setting efforts include the American Bar Association's Tentative Draft of Standards Relating to the Legal Status of Prisoners (1977) and the Federal Standards for Corrections drafted by the U.S. Department of Justice (1978a).

230 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM fined as a minimum of 60 square feet of floor space per inmate, prisons were operating at 171 percent of capacity (Mullen et al., 1980:65~. An additional indicator of crowded conditions in state prison facilities is the extent to which state prisoners are housed in local facilities. According to the Bureau of Justice Statistics, 6,497 inmates (2.1 percent of the total) were housed in local jails at the end of 1979, including 39.4 percent of Mississippi's state prisoners and 24.6 percent of Alabama's (U.S. Department of Justice, 1981c: 154. Relief from population pressures through expansion of prison and jail facilities is not anticipated. In its survey of facility construction, reno- vation, acquisition, and closing plans between March 31, 1978, and December 31, 1982, an Abt survey found jurisdictions planning a total increase of 52,843 beds or an overall increase of about 24 percent in net additional capacity (Mullen and Smith, 1980:80~. But these increases are in rated capacity, not in the measured capacity standard of 60 square feet (suggesting the possibility of capacity increases achieved by changing definitions, greater use of double-celling, and a decline in living stan- dards). They also include intended expansions for which appropriations had not yet been authorized by state legislatures and so they may be inflated.3 The increase of 48,651 state and federal prisoners between year-end 1977 and mid-year 1981 (U.S. Department of Justice, 1981b) and the time lag between planning, appropriation, and construction of a facility suggest that facility expansion is not keeping pace with ex- panding populations, that prison crowding is increasing, and that short- term approaches to correctional population pressures are badly needed now. The dramatic increases in prison populations, as well as changes in sentencing policies, raise two major questions for institutional manage- ment: What is the general effect of crowded prison conditions on inmate health and behavior and on institutional management? What is the effect of determinate sentences on institutional programs, offender miscon- duct, and disciplinary procedures? Inmate Health and Behavior It is widely believed among prison administrators and researchers that crowding has adverse affects on inmate health and behavior and, by 3 Another survey of prison construction plans completed in May 1981 (National Mor- atorium on Prison Construction, 1981) found that, although states "planned" to expand by a total of 102,350 beds, if "planned construction" is defined as a facility that has been approved by a legislative body or been included in a governor's budget, only 25,316 additional beds were authorized or under construction.

Impact on Prison Populations 231 increasing tension and aggression, may contribute to inmate violence and prison riots. But systematic research on these subjects is recent and limited. Cumulation of knowledge has been hindered by variation in the definitions and measures of crowding and density and their behav- ioral, physical, and attitudinal outcomes, resulting in noncomparable findings. The initial research on crowding involved studies that documented its deleterious effects on animal health and behavior.4 Studies of the re- lationship between population density and human behavior have yielded more ambiguous results than the animal research. Correlational studies of the incidence of social pathology in communities and households differing in population density have been a primary source of empirical conclusions about the effects of crowding (see Bordua, 1958; Galle et al., 1972; Schmid, 1960; Schmitt, 1957; Shaw and McKay, 19424. But these ecological approaches must be viewed with great caution because of the danger of making inferences about individuals from aggregate data, particularly where findings may be confounded by high correlations of deleterious factors like economic and educational deprivation with density. Observational studies of hospital patients and of children in groups of varying size have yielded inconsistent results (see Hutt and Valzey, 1966; Ittelson et al., 1972; Loo, 1972; McGrew, 1970~. Individuals con- sistently interacted less in larger groups, but Loo reported less aggression in the denser situation, while Hutt and Vaizey found more. Sample surveys have found positive correlations between persons per room or perceptions of the household as crowded and disrupted inter- personal relationships within the home (Baldassare, 1978; Mitchell, 1971), stress diseases (Booth and Cowell, 1976), and poor mental health (Gove et al., 1979~. Again the correlation of density with other potentially deleterious factors may confound the findings. In addition to these general studies of crowding, several studies have examined the effects of crowding and high density in prisons. (Although the terms "density" and "crowding" sometimes are used interchange- ably, "density" refers to a physical condition and "crowding" to a sub- jective reaction to that condition. Further, there is a distinction between social density, the number of occupants per living unit, and physical 4 Conditions of high density were found to impair fertility and reproduction in mice (Christian, 1960~; to disrupt maternal ties, lead to homosexuality, and produce social withdrawal in rats (Calhoun, 1966a,b); and to cause increased aggression and emotionality, prostration, convulsions, and death in several mammalian species (Ader et al., 1963; Barnett et al., 1960; Bullough, 1952; Calhoun, 1956, 1962; Christian, 1960; Keeley, 1962; Rosen, 1961; Turner, 1961~.

232 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM density, the number of square feet per individual.) Studies of the effects of social density have found that inmates living in open dormitories feel more crowded, rate their environments more negatively, have higher rates of illness complaints, and feel greater psychological stress than inmates living in single or double occupant rooms (Cox et al., 1979, 1980; Paulus et al., 1975~. Some of the negative effects were reduced through use of dividers in dormitories, permitting an increase in privacy without an increase in space (McCain et al., 1980~. McCain et al. (1976) also found that increased physical density led to progressive and mea- surable increases in negative effects on inmates, including higher rates of illness complaints, perceived crowding, mood states, rating of the environment, perception of choice and control, and nonaggressive dis- ciplinary infractions. Increased social density was a more important contributor to the physical and psychological effects than increased phys- ical density. The disciplinary data were obtained in only one institution, however, and only nonaggressive infractions were considered. Some studies of the effects of crowding-related stress in prison, as measured by blood pressure, find that inmates in open dormitories have higher blood pressure than those in single cells, that increased social density in dormitories increases blood pressure (D'Atri, 1975; Paulus et al., 1978; Ray, 1978), and that transfers from single occupancy cells to multiple occupancy dormitories cause increases in blood pressure (D'Atri et al., 1981~. However, McCain et al. (1980) found no density- related blood pressure effects. Nacci et al. (1977) report that the federal correctional institutions that were most overcrowded relative to rated capacity, particularly those that had a large number of young offenders, had the highest disciplinary infraction rates. Similar results are reported by Megargee (1977) where the amount of living space available and the density were significantly associated with both the number and rate of disciplinary violations. However, Megargee found that the rate at which disciplinary reports (not distinguished by level of seriousness) were given over a 36-month period was not significantly related to the overall population in the institution, but he looked only at medium-sized institutions that varied within a narrow range (45~550 inmates). A third study (McCain et al., 1980) found both sheer population size and increased density in prisons associated with negative effects, including disproportionate increases in rates of disciplinary infractions, violent death, suicide, and death of inmates more than 50 years old. In sum, research is just beginning to sort out the complex and often overlapping effects of social and physical density, crowding, and insti- tutional size on inmate perceptions, morale, health, and behavior. A

Impact on Prison Populations variety ot negative effects have been hypothesized to result from sus- tained high-density living conditions, particularly in large institutions. The evidence, however, is still fragmentary due to methodological short- comings, divergent measures, and correlations that may result from uncontrolled confounding variables like prison age and condition, se- curity level of the institution, and the amount of time prisoners are confined to cells. It is very difficult to separate the individual physical, psychological, and behavioral effects of crowding from the effects of the many other undesirable aspects and deprivations of a prison^environ- ment. In order to provide a comprehensive interpretation of the variety of crowding effects on inmates, future research on crowding effects should look across institutions; introduce richer controls for other attributes of these institutions; examine variations in the effects of crowding over time, including long-term effects (more than 3 years); compare housing types explore individual and group differences in reaction to high-density living arrangements; and develop a wider variety of measures of be- havioral effects to supplement attitudinal measures. 233 Determinate Sentencing and Institutional Programs Under indeterminate sentencing policies, corrections institutions de- veloped a broad range of rehabilitation programs, including vocational, educational, and social skills development; individual and group ther- apy; and partial physical custody (i.e., work release and placement in halfway houses). Some expected that the shift to determinate sentencing policies would result in a reallocation of institutional resources from rehabilitation programs to custodial uses, a drop in program partici- pation as inmates no longer felt coerced in such programs, and greater motivation from those inmates who continued to participate voluntarily. To date, only two preliminary studies have been completed on the effects of determinate sentencing on institutional programs in three ju- risdictions; both studies have serious limitations. Brady (1981) examined the impact of determinate sentencing in Oregon and California on prison programs and inmate attitudes toward them. His interviews with ad- ministrators, custodial and program staff, and inmates indicated that participation in programs in both states continued at about the same level as before determinate sentencing, but that many staff members sensed some change in inmate behavior and attitudes toward partici- pation: many inmates were more negative; a few were more motivated. The contribution of California's Uniform Determinate Sentencing Law (DSL) to these attitudinal changes is unclear; what is clear to prison

234 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM staff is that "rehabilitation and programs have less general appeal ito inmates] than perhaps five years ago" (Brady, 1981:9~. California's in- stitutional programs have continued, but a clear shift in goals and policies has occurred due to both disciplinary problems in institutions and DSL. Prisoners are being reclassified according to their background and "prior incarceration behavior" rather than their amenability to rehabilitation. While the programs themselves have not been altered, internal security dominates the prison atmosphere and staff view prisoners as less co- operative than they were prior to implementation of DSL. Brady's findings must be viewed as preliminary. There are no com- parative data on pre-DSL and post-DSL levels of participation, and other changes occurred simultaneously with implementation of the new law, including more crowding and an increasingly youthful and violent inmate population. Stone-Meierhoefer and Hoffman (1980) examined the effects on pro- gram participation of setting presumptive parole release dates within 120 days of admission to prison. Comparing program participation by an experimental group (randomly assigned prisoners given presumptive parole release dates) and a control group (prisoners considered for pa- role after serving one-third of their sentence), they found that experi- mental group members enrolled in somewhat fewer programs, partic- ularly education programs, than control group members and appeared to drop out of a slightly higher percentage of the programs in which they enrolled. No significant difference was found between the groups in the number of persons enrolling in at least one type of educational or work program, but experimental group members joined significantly fewer programs than did those in the control group. The authors at- tribute this difference to the fixed parole date that obviates the need for program participation in order to impress the parole board. The generalizability of these findings is limited. Staff and prisoners were aware that the experiment was part of a pilot study of the U.S. Parole Commission's new parole guidelines. The guidelines had already been implemented, so all inmates already could predict their parole dates, thereby reducing parole-related incentives for program partici- pation for both experimental and control groups. Inmate Misconduct, Disciplinary Procedures, and Determinate Sentencing Determinate sentencing was also expected to affect disciplinary prob- lems and procedures in prisons (Goodstein, 1980; Morris, 1974; von Hirsch, 1976~. Supporters of determinacy reasoned that uncertainty about

Impact on Prison Populations 235 the length of time to be served caused anxieties, tensions, and frustra- tions among inmates that contributed to more institutional misbehavior and interpersonal violence (Bennett, 1976; Park, 1976~. Conversely, critics of this viewpoint suggested that determinacy would increase dis- ciplinary problems by removing the threat of longer imprisonment. Three studies have produced data on the effects of determinacy on prison discipline (Forst, 1981; Goodstein, 1981; Stone-Meierhoefer and Hoffman, 1980~. Their methodologies, research questions, and findings vary and provide at best preliminary and suggestive obervations rather than reliable conclusions. Stone-Meierhoefer and Hoffman (1980) found that granting pre- sumptive parole dates did not appear to adversely affect discipline within federal prisons. Comparisons of experimental and control groups (as described above) indicated no significant differences in the proportion of inmates committing disciplinary infractions, the total number of in- fractions, or the number of inmates committing major infractions (after controlling for months of exposure). However, because federal prisoners differ in both offense type and background from state prisoners and the sample pool underrepresented long-term prisoners arguably those with little incentive to conform—the disciplinary impact of a fixed parole date may have been underestimated. ~ Goodstein (1981) used a quasi-experimental design to compare pris- oners with determinate and indeterminate sentences in three South Car- olina prisons on a number of attitudinal and behavioral measures.5 Con- trolling for differences in sentence length (because those with indeterminate sentences had committed different offenses and had longer terms), Goodstein found no difference between the prisoner groups with respect to rate of institutional misconduct. She did find, however, that inmates with determinate sentences reported experiencing significantly less stress than did those with indeterminate terms. Forst (1981) examined the impact of determinate sentencing on inmate misconduct and institutional discipline in Oregon and California. His interview and observational data indicate that determinacy has not been the answer to prison unrest that its supporters had hoped, nor has knowledge of a fixed release date led directly to increases in misconduct ~ In South Carolina, judges have the discretion of sentencing offenders to terms with fixed release dates (via a "split sentence" requiring that an offender serve a specified portion of the total sentence in prison) or to long maximum terms with the expectation of earlier parole release. Since the criminal code makes all inmates eligible for parole after serving one-third of their maximum sentences, it is possible for an offender with a split sentence to be released from prison at a fixed date prior to his parole eligibility date.

236 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM as others had feared. Indeed, it now appears that "prison violence. . . is little affected by the type of sentencing structure" (Forst, 1981:88~. In California, the change from indeterminate to determinate sen- tencing both eliminated the parole board that previously set parole release dates and reinstituted the use of good time. The law left un- touched the authority of corrections officials to refer serious misconduct to the district attorney for prosecution as a new offense and their ability to alter the quality of time served by means of a variety of sanctions, including isolation and segregation.6 Between 1970 and 1979, serious disciplinary infractions of all types rose steadily in California prisons.7 Since implementation of DSL in July 1977, however, forfeiture of good time as a disciplinary mechanism has been used modestly, but it is gradually increasing. Between July 1, and September 30, 1977, 1.7 percent of prisoners found guilty of serious disciplinary infractions (but not necessarily "good-time offenses") lost some good time; for the same quarter of 1979, good-time forfeitures had increased tenfold, to 17.2 percent of such prisoners. The median number of days of good time lost in 1978 was 10 (Forst, 1981:97-8~.8 Felony referrals to district attorneys have also increased steadily, from 931 in 1975 to 1,744 in 1978, but prison officials attribute this principally to the increase in the number of serious felonies committed in prison, rather than to a change in policy associated with DSL (Forst, 1981:5 59). In Oregon, increased use of parole release after 1972 and subsequent adoption of the parole guidelines in 1977 diminished the role of good- time forfeiture as a sanction by corrections authorities. To regain some leverage over time served as a means of controlling inmates, corrections officials in 1978 obtained parole board approval of a system for changing inmates' parole release dates under certain circumstances on prison 6 Segregation is a classification decision in California: while technically a change in an offender's placement to a more secure housing unit is not a punishment, reclassification of custody status actually functions as a qualitative sanction as well as a mechanism for protecting inmates who request it. 7 During that 9-year period, the rate of incidents per 100 average institutional population increased almost 10-fold, from 1.36 to 10.07 (Management Information Section, Policy and Planning Division, California Department of Corrections, February 26, 1979, cited in Forst, 1981:77~. Another indicator of the increase in serious misconduct is the increase in assaults on staff, which rose from 94 in 1976 to 182 in 1978 (Forst, 1981:80~. ~ Some of the prisoners found guilty of an infraction were "lifers" not subject to good- time loss. Some of the forfeiture was loss of participation credits. In 1978 a department policy directed all inmates to be assigned to a program; as jinmates failed to participate, more participation credits were forfeited.

Impact on Prison Populations 237 authorities' initiative, thereby reintroducing an element of indetermi- nacy.9 Actual use of changed release dates for handling discipline prob- lems varies among institutions and seems to be (inversely) related to crowding rather than to the rate of misconduct.~° Data on changes in inmate misconduct as a result of the introduction of determinacy in Oregon, however, are incomplete because of a court order to expunge all records of disciplinary matters between December 6, 1977, and October 22, 1979. Nevertheless, from the data that were available Forst (1981:85) observed: "We cannot distinguish any rela- tionship between inmate misconduct (as measured by disciplinary re- ports) and the change from an indeterminate to a determinate sentencing system." An apparent trend in both states is reliance on disciplinary devices that affect the quality rather than the amount of time in prison, prin- cipally through reclassification of inmates and the resulting transfers among housing units that vary in degree of security. This is viewed as having two advantages for prison administrators: it does not increase the prison population, and it has a more direct and immediate effect on the inmate, which is viewed as a more effective deterrent to misconduct. Forst also found that corrections administrators in both Oregon and California report decreased tension and anxiety over uncertainty of re- lease date but no concomitant reduction in misconduct. Though no statistical data were available to support their view, they suggest that determinacy has indirectly increased misconduct in two ways: first, it contributes to prison overcrowding, which results in heightened tension and disciplinary infractions; second, it leads to feelings of hopelessness 9 There now are four categories of prison misconduct that can result in a change in release date. The range of possible extensions to a term varies with the seriousness of the misconduct. Misconduct that is hazardous to human life can result in a change of from 50 to 100 percent of an inmate's term with a maximum extension of 5 years. Misconduct that is a hazard to security can increase a term from 25 to 50 percent to a maximum 2- year extension. Hazard to property can increase a term from 10 to 20 percent to a maximum 1-year extension, and the third in a series of rule violations within a 3-month period can increase a prison term from 5 to 10 percent to a maximum 6-month extension. it At Oregon State Prison, which was very crowded at the end of 1979, only two or three term changes were made of 4,120 disciplinary reports filed (Forst, 1981:98), as prison authorities relied on segregation (which does not increase time and thereby prison population) in preference to changes in parole release dates. At Oregon State Correctional Institution, where crowding was less critical and most inmates live in dormitories, changes in release dates were more frequent, although only 2.8 percent of all disciplinary actions filed in that institution in 1979 led to recommendations for changed release dates (Forst, 1981:98).

238 RESEARCH ON SENTENCING THE SEARCH FOR REFORM and frustration among prisoners who have long sentences that may be extended but which they can do nothing to shorten. This view is ironic in light of the rejection of indeterminacy by some because it was believed to lead to frustration and hopelessness. Prison officials are reported to favorably regard changes in disciplinary procedures stemming from the determinate sentencing laws in Oregon, since the parole board has agreed to a procedure for changing parole release dates that increases the officials' influence. In California some administrators prefer the current system that specifies acceptable be- havior, while others feel they have diminished control over inmates. In sum, all three studies of determinacy and offender misconduct, though preliminary, suggest that both critics and supporters of deter- minacy exaggerated the effect of the change. Determinate sentencing may have limited impact on prisoners' misconduct because. in relation to peer pressures and other concerns, it has little influence on the daily environment of a prison inmate. PROJECTION OF PRISON POPULATIONS: NEED, TECHNOLOGY, AND USES NEED FOR PROJECTIONS OF PRISON POPULATIONS The need to develop improved methods for estimating the impact of changes in sentencing policies on prison populations has become es- pecially important in the face of capacity constraints and increased crowding in U.S. prisons. Without consideration of the impact of policy changes on prison populations, two undesirable consequences are likely to occur: prosecutors and judges will adhere to new policies, and prisons will become severely overcrowded; prosecutors and judges will informally seek to limit prison populations through accommodations that modify mandated policies. The effects of a sentencing policy on the corrections system are gen- erally ignored by judges and often are not considered by legislatures. Some have asserted that such effects should be ignored when considering broad principles of justice or weighing individual cases. Such a per- spective, however, may be impractical during the 1980s when prisons are at or near capacity and substantial additional prison space is unlikely to be available soon. Consequently, consideration of policy changes likely to significantly increase prison population should weigh the de- sirability of the change in light of available prison capacity and the costs of increasing that capacity. Adopting a policy without providing the resources needed to imple- ment it tends to undermine respect for the law by participants in the

Impact on Prison Populations 239 system and to encourage violation through a variety of ad hoc adapta- tions. Furthermore, many jurisdictions are likely to experience increases in prison populations, even without explicit changes in sentencing policy, and these jurisdictions will need more capacity just to maintain current practices. Thus, whether considering policy changes or assessing current policies, projection of the impact on future prison populations of existing and alternative practices is a necessary component of sound public policy formulation. In making prison population projections, three factors must be kept in mind: the amount of time necessary for the full effect to be felt, the amount of compliance, and the nature and composition of the prison population. The time dimension is important in distinguishing short- term and long-term effects. A policy of incarcerating a higher proportion of a certain type of offender (e.g., a mandatory minimum sentence of 5 years for all robbers) would increase prison populations more rapidly than an increase in the average length of sentence of those categories of offenders who are already being imprisoned (e.g., an increase in the average sentence of incarcerated robbers from 4 to 7 years). The latter change will lead to a gradual population buildup over several years. The former will have an immediate, dramatic short-term effect, through increases in commitments, as well as long-term consequences. Prison population projections must also consider likely rates of com- pliance with new policies. The simplest assumptions, no compliance (i.e., a continuation of existing policy) and complete compliance, are likely to be inaccurate; actual compliance will probably lie between these extremes. Therefore, several estimates that assume different levels of compliance by justice system personnel are desirable. Policy changes may alter the composition of prison populations and the length of the inmates' sentences. These effects, in turn, may have repercussions for programs and levels of control in institutions. For example, an increase in the number of violent youthful offenders serving long terms may suggest a need for increased custodial staffs, since such offenders have lower rates of participation in institutional programs and worse disciplinary records than other offenders. PROJECTION METHODS: THEIR USES AND SHORTCOMINGS Naive Projections Using the Existing Situation as a Baseline The simplest projection method rests on the assumption that next year's prison population in the absence of a policy change will be the same as the current population. (For a more detailed discussion of projection methods, see Blumstein tVolume II].) This method, while offering the

240 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM considerable advantage of simplicity, assumes stability of prison popu- lations over time, absent any policy changes, despite evidence of changes in crime rates, sentencing practices, and the demographic characteristics of offenders. The further one projects into the future, the less accurate the baseline data are likely to be as projections. Extrapolations of Time-Serzes Data Simple linear extrapolations of future prison population based on recent trends have often been used by researchers and corrections planners. Over the last decade, however, forecasting procedures have become more sophisticated than trend analyses (see Box and Jenkins, 1976; Granger and Newbold, 1977; Nerlove et al., 1979~. To the degree that the future is like the past, it is possible to accurately forecast the future of a wide variety of historical patterns using techniques that include linear trends, shifts in level, shifts in slope, seasonal cycles, and other temporal regularities. Once these historical patterns are captured in a small set of parameters whose values are estimated from the observed time series, optimal forecasts are available (i.e., with minimum mean square forecasting errors). These forecasts do not invoke any causal structure, which is both a strength and a weakness of the technique. The strength is that they do not rely on current social science theory, which may not be able to explain incarceration rates. There is no need to collect data on causal variables and forecast their values (which would be required for forecasts of the outcome variable of interest): pure induction from the outcome variable alone will suffice. The weakness is that, if the underlying causal relationships produce new temporal patterns, time-series forecasts will be inaccurate. Thus, such forecasts tend to be more useful in making short-term projections than long-term ones. Furthermore, because ex- trapolations from time-series data only consider the time variable and assume a constant rate of change in other factors that influence prison populations, they are of limited utility to a legislature considering the effect of a policy change. There is still lively debate about the accuracy of time-series extrap- olations compared to alternate approaches. A great deal depends on specific applications, and experience with forecasts of prison populations is limited. Nevertheless, time-series forecasts are valuable procedures when (1) there is no reason to believe that structural changes will occur; (2) the time-series model easily survives statistical tests of its validity; (3) the time series includes many observations (e.g., more than 100~; and (4) the time horizon of the forecasts is short.

Impact on Prison Populations Use of Predictor Variables 241 Prison populations can also be projected by relying on a variety of other variables, which are believed to be causally related to prison population, as predictor variables in an estimated regression equation. Some of the predictor variables that have been included in a prison population fore- cast model are the consumer price index (Fox, 1978), unemployment (Robinson et al., 1977), the demographic mix in the population (Crago and Hromas, 1976), and prison capacity (Abt, 1980~. There are three problems with using predictor variables. First, in projecting future prison populations, they often include variables in the model that are more difficult to project than prison population itself. Use of the unemployment rate, for example, in the absence of accurate projections of that rate, adds little to one's ability to project prison populations. Demographic variables are more easily projected because data on individuals in a demographic group, such as males aged 2~29, are available and fairly easy to project. Second, these forecast models usually do not include changes in intervening sentencing policy variables, such as the probability of imprisonment and the length of prison sen- tences. Even if sentencing policy variables are included directly as pre- dictors in models, the changes in these policy variables must then be projected. Finally, these models are at present relatively simple. They fail to consider the many interrelated political, socioeconomic, and de- mographic variables that appear to influence sentencing. But adding more variables to the model is often not feasible given the difficulty in making future projections of many of them. Projections Based on Demographic-Specific Incarceration Rates A variation of projections using general predictor variables uses de- mographic-specific incarceration rates as the predictor variable. These projections rely on marked differences in involvement with the criminal justice system among different age, race, and sex groups. In 1979, for example, the incarceration rate of males was 25 times that of females; the incarceration rate of black males was 6.7 times that of white males; and the incarceration rate of white males aged 23 (the peak age of incarceration) was 8.8 times that of white males aged 40 and older (see Blumstein, Volume II). Projections of prison populations for demographic-specific subgroups are particularly attractive when one has fairly reliable projections of demographic changes in the general population and when incarceration rates, especially for the high-rate groups (e.g., males aged 2~29), are

242 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM fairly constant over time. Examination of incarceration rates within demographic groups, however, indicates the possibility of substantial changes in these demographic-specific rates over time (see Blumstein, Volume II; U.S. Department of Justice, 1981b). In addition, this ap- proach does not include in the projection model sentencing policy var- iables, such as an increase in the proportion of burglars incarcerated as a result of new legislation or an administrative decision. While the ab- sence of sentencing policy variables could theoretically be remedied by generating demographic-specific conviction rates by offense type and then applying sentencing variables to them, data systems found in most jurisdictions do not provide sufficient information to permit estimation of conviction rates that are demographic- and offense-specific. Disaggregated Flow Models Disaggregated flow models permit detailed disaggregated examination of future prison populations. They require a data base that contains records on individual cases as they proceed through the criminal justice system. Development of such models, therefore, is feasible principally in jurisdictions with operational offender-based transaction statistics (OBTS) systems. The OBTS system includes attributes of the offender and the offense and describes the experiences of individuals as they are processed through the criminal justice system. An individual case record is created for each arrest or court filing; additional information is added to the record as the case moves through the successive stages of pro- cessing in the criminal justice system. Analyses of the records of indi- i~ The OBTS system was initiated in 1969 with funding from the Law Enforcement Assistance Administration (LEAA) to Project SEARCH in an effort to computerize reports from existing criminal justice statistical series. When the SEARCH task force found that such series did not yet exist, they turned their attention to the design of statewide statistical systems and concluded that such systems should be based on data on offenders as they passed through the system. A model system, proposed for adoption by states, emphasized selection of certain common data elements and use of a common unit of analysis (i.e., the defendant who is charged with a felony and fingerprinted). In 1973 LEAA awarded two separate grants to Project SEARCH: one to design an offender- based state corrections information system, the other to design a state judicial information system. Several states were selected to participate in the development and testing of these systems, which were intended to collect management information for daily operation, and at the same time meet the OBTS data requirements and transfer appropriate data into master OBTS files in each state. Since then a number of states have developed OBTS systems that include common court and correctional system data on individual offenders.

Impact on Prison Populations 243 vidual cases completed during a given period (e.g., a year) permits disaggregated estimates of the nature of case processing at various de- cision points in the criminal justice system. In these flow models the state or local criminal justice system is rep- resented as a series of stages processing defendants, or "units of flow." The flow through each stage of the system can be represented by a matrix of branching ratios or transition probabilities representing the percentage of cases at any stage that proceed to the next stage. These transition probabilities can be disaggregated by crime type (or any other relevant characteristic of the units of flow) to allow for differences in the way different cases flow through the system. Sentencing policy var- iables are explicitly included in this detailed characterization of case processing. This kind of model can then be applied to disaggregated projections of system inputs (e.g., crimes, arrests, or convictions) to generate projections of prison population. ~ A disaggregated flow model permits a fuller characterization of case processing than is available from incarceration rates alone. The model's flow parameters are, nevertheless, generally treated as fixed quantities because of inadequate knowledge about likely system responses to changing flow levels through the system. This is not an inherent limi- tation, however; to the extent that plausible assumptions about changes in case processing at various points can be made, the model's flow parameters can be manipulated to reflect anticipated processing or policy changes. One example of this approach is found in Blumstein et al. (1980~. Demographic- and offense-specific arrest rates are used in combination with population projections to estimate future arrests. Then, similarly estimates of the disaggregated probability of imprisonment and time served are applied to the arrests to yield projections of the size and composition of future prison populations. Using data for Pennsylvania from 1970 through 1975 and projections of demographic changes in the state's population, the model estimated future arrests and prison com- mitments for Pennsylvania to the year 2000. The projections to the year 2000, reflecting the strong effect of the postwar baby boom on the criminal justice system, suggest that arrest rates in Pennsylvania will peak about 1980, prison commitments will peak in 1985, and prison population will peak in 1990, then gradually decline. Because the pro- jections ignore possible policy changes and the likely adaptive responses in the criminal justice system to increasing population pressures on the prison system, they are likely to be increasingly inaccurate the farther they extend in time. The model is useful, nevertheless, in suggesting

244 RESEARCH ON SENTENCING THE SEARCH FOR REFORM the likely point at which additional capacity or policy alternatives will be needed to accommodate mounting population pressures, thereby helping decision makers select among alternatives. Microsimulation Models Disaggregated flow models examine the distribution of average flow rates through the criminal justice system. Microsimulation models per- mit estimation of total distributions of flow parameters by simulating the flow of individual offenders through the system, then combining their individual experiences into aggregate statistics. A microsimulation can use a group of actual case records. Such an approach was used by the projection estimates developed by the Min- nesota Sentencing Guidelines Commission (MSGC), which permitted estimation of the effect on prison populations over a S-year period of any guidelines sentencing schedule or policy option considered by the commission. In the MSGC model, the primary determinants of future prison populations are current prison population, future commitments to the prison population for 5 years, and the length of current and future prisoners' sentences. In the simulation, the movement of individual cases through the criminal justice system is governed by flow probabilities and by the length of time spent at each processing point, both of which are adjustable parameters in the model. The MSGC model was designed to permit flexibility in testing alter- native sentencing policies. When sentencing decision rules are proposed, the new sentences imposed on each of the simulated cases and the aggregate consequences for prison populations of the particular policy can be examined for a multiyear period (see Knapp, 1980; Knapp et al., 1979~. In using the microsimulation to project long-term future population, it is important that the microsimulation be augmented by projections reflecting anticipated changes in the size and composition of the cases that serve as input to the simulation. ESTIMATING THE EFFECTS OF SENTENCING POLICY CHANGES ON PRISON POPULATIONS Since sentencing policies are shaped and implemented in states (or in some instances at a local level), jurisdiction-specific projections that estimate the consequences of changes in sentencing policy for prison populations are needed in advance of a policy change. To be useful as a policy-impact estimating technique, a projection model must contain

Impact on Prison Populations 245 estimates of sentencing variables, including commitment rates and sen- tence lengths by offense type; disaggregated flow and microsimulation models are best suited for this purpose. Development of impact assess- ments involves four steps, each subject to data and methodological difficulties. The first step is identifying the subset of cases to which the policy change would apply. A mandatory sentence for use of a gun, for ex- ample, would apply to cases involving guns. Unfortunately, most data sets do not contain adequate individual or aggregate data on details of the offense and relevant attributes of the offender to permit accurate determination of which or how many cases would be affected by such a proposed policy. Approximations of missing data are thus often nec- essary, adding uncertainty to the projection. The second step is establishing the future values of policy variables. With adequate data, a proposed sentencing policy can be characterized by specifying corresponding sentencing variables for each affected of- fense/offender subset. For example, in the case of a mandatory minimum sentencing law, it is necessary to determine which offenders previously not incarcerated for a particular offense type would be subject to in- carceration under the new law and to specify the sentence lengths both for those newly incarcerated and for those who were previously incar- cerated for less than the proposed mandatory minimum sentence. (Those already receiving sentences above the mandatory minimum and those committing offenses not addressed by the mandatory law would not be affected by the law.) The third step is estimating behavioral responses to a new policy. Policies are often not implemented as planned. Actors in the criminal justice system follow a variety of adaptive strategies that may affect the number of commitments and time served under a new policy. Responses by judges to a mandatory minimum sentence law might include, for example: literal interpretations, with prison sentences, for all who satisfy the conditions of the law, for the specified mandatory minimum sen- tence; increased sentences of up to the new required minimum for all who formerly went to prison but continued sentences of probation for those previously sentenced to probation (perhaps through agreement to a plea to a lesser charge); or probation~sentences for some of those formerly sent to prison for terms shorter than the new minimum (perhaps through conviction on a lesser charge or invoking a mitigating circum- stances provision) in order to avoid the longer sentences. An assumption of literal compliance is likely to overestimate the impact of a policy; it is preferable to test several possible response patterns to establish a likely range of outcomes.

246 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM The final step is calculating the effects of a change in sentencing policies on prison populations. This step involves comparing prison pop- ulations expected under the old and new policies USiIlg various behav- ioral assumptions. The difference in projected populations reflects the effect of a new policy. Both projections of prison populations that do not include consid- eration of policy changes and those designed specifically to examine the effects of particular policy choices permit fuller appreciation of the fac- tors that affect prison populations, provide estimates of the ranges for those populations, and encourage the development of an ongoing mon- itoring system that includes data on the behavior of participants and the flow of offenders through the criminal justice system. In addition, es- timates of the impact of changes on prisons represent an important methodological device for forcing consideration of policy issues. In jurisdictions where proposed sentencing policies increase punitive- ness and further exacerbate pressures on already crowded prisons, policy makers face a dilemma: Should they increase prison capacity, which is costly and may not be needed soon after construction is completed, or look for alternative punishment strategies? Impact estimates can aid in responsible decision making by focusing attention on the explicit value trade-offs associated with a desired level of punitiveness and its costs. A note of caution is necessary in considering prison population projec- tions, however, to avoid overconfidence in projected figures and the possibility that reactions to projections will lead to self-fulfilling proph- ecies. It must be remembered that all projections are vulnerable to errors arising from data inadequacies and the uncertainty of system responses to new policies. ALTERNATIVE STRATEGIES FOR HANDLING INCREASING PRISON POPULATIONS Three general strategies are available to achieve a balance between prison capacity and prison population: expansion of capacity through changes in existing facilities and new construction; limitation on admis- sions through use of alternatives to imprisonment; and direct regulation of prison population through controls on intake and release. Most states now appear to be using at least one of these approaches to some extent. While selection among these options is primarily a policy question, policy choices can be informed by consideration of the relative short-term and long-term effects of each strategy.

Impact on Prison Populations INCREASING PRISON CAPACITY THE POPULATION-CONSTRUCTION NEXUS 247 There are no simple explanations of why some states build new prisons and others do not. The decision to construct new prison facilities appears to be influenced by a variety of demographic, social, economic, and political considerations often in combinations unique to each state. Some of the factors that tend to accelerate the decision to build are the apparent failure of alternatives to incarceration, leading to a renewed reliance on imprisonment; the need for specialized new facilities; court orders; prison disturbances or riots; state population growth; and the availability of federal funds or facilities. Prison riots and disturbances, for example, seem to contribute to construction both by the focusing of public attention on the need for more or better facilities and through the destruction of existing housing that necessitates replacement con- struction. States that are growing in overall population appear to be expanding prison capacity faster than states with stable or declining populations and also building at a higher rate than the rate of increase in the number of prisoners. The availability of existing federal facilities that require only modest renovation and involve moderate operating costs has contributed to the expansion of state prison capacities (Benson and Silberstein, 1983~. Other factors tend to retard capacity expansion. These include the existence of excess capacity in some state prison facilities; redefinition of rated "capacity" to meet population increases; political circum- stances that prevent development of a consensus on the need to build or block implementation of a decision to do so; budgetary constraints and competition for funds; site-related opposition; regulatory limitations on location and construction; and effective prison management. Increasing prison populations are costly in terms of both capital out- lays to expand capacity and increases in direct operating expenditures (cash outlays for purchase of noncapital goods and services). Such ex- penditures for adult correctional institutions both jails and prisons- for all levels of government in fiscal 1977 were about $2.46 billion. The average annual per-inmate cost for all adult inmates of state prisons was $5,461, with a range of costs across states from $2,241 to $14,946 (Mullen and Smith, 1980:115-117~. Direct current expenditures of federal, state, i2 The Supreme Court decision in Rhodes v. Chapman (452 U.S. 337 [1981~), permitting two prisoners in a single cell under certain circumstances, may have discouraged con- struction by enabling many states to legally increase prison density.

248 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM and local governments steadily increased between 1971 and 1977; the $2.46 billion spent in 1977 represented an increase of 45 percent over the 1971 figure after adjustment for inflation. It has been estimated that these expenditures would increase an additional 1~17 percent by 1982 (Mullen and Smith, 1980: 134~. Capital outlays for correctional institutions (including juvenile deten- tion facilities for state and local governments) in 1977 amounted to $415 million (only a small fraction of which was spent on equipment). Esti- mating future prison construction costs is difficult, however, due to wide variations in estimated costs depending on institutional size, region, and security level (see Singer and Wright, 1976~. The National Council on Crime and Delinquency (1977:7) estimated that construction costs per new bed range from $25,000 to $50,000; according to the Federal Bureau of Prisons, a new 500-bed facility would cost about $35,000 per new bed (U.S. General Accounting Office, 1978:13~. Financial costs are only one consideration in the complex decision regarding construction of new prison facilities, but the millions of dollars for each new prison that might be spent on other government services and facilities, particularly in a time of fiscal austerity, appear to have been a major inhibitor of prison capacity expansion. In the past few years, voters in several states, including New York and Michigan, have rejected bond issues to finance prison construction. Several studies have attempted to develop and test general hypotheses about why particular states build new prisons. Given the range of factors that might affect the construction decision, these analyses have been rather simplistic, and, thus far, the models have not fit the evidence very well. Nonetheless, a consideration of their shortcomings may be Instructive. One approach, termed the "population model," suggests that the supply of prison housing is expanded in direct response to increased demand in the form of prison population increases. However, neither recent national prisoner statistics nor a preliminary test of the correlation between measures of prison population growth between 1975 and 1981 and estimates of planned net capacity increase from 1978 through 1982 for the 50 states (Benson and Silberstein, 1983) support this model without the addition of other factors that mediate the population-con- struction relationship. An alternative "capacity model," suggested by William Nagel (1973), postulates that prison construction is itself a stimulus to prison popu- lation expansion. In this model, expanded prison capacity affects sen- tencing decisions, resulting in more prisoners to fill that capacity, re- newed population pressures, and further construction.

Impact on Prison Populations 249 Carlson et al. (1980) sought to test both the capacity and population models and to clarify the relationship between capacity and population. They found no relationship between current capacity and present pop- ulation, i.e., construction does not appear to be significantly stimulated by existing or past prison population pressure. However, they report a significant and substantial relationship between past capacity change and future populations with a 2-year lag and concluded that (Carlson et al., 1980:56~: On the average . . . additions are filled to rated capacity by the second year after opening additional space; within five years the occupancy of the new space averages 130 percent of rated capacity. Because the finding that prison capacity generates the population to fill it has been widely cited by the press and accorded importance by policy makers, the panel believed it important to assess the validity of the finding.~3 The independent review of the data indicates that the empirical evidence cited in the Carlson et al. study provides no valid support for the capacity model (Blumstein et al., 1983~. Errors in the study include an excessively simplistic formulation of the problem and associated statistical model; failure to test the sensitivity of the computed results to undue influence by several extreme data points; a serious computational error in calculating the univariate estimates of the coef- ficients; a highly questionable assumption that there were no changes in prison capacity in years when no new facilities were opened; inade- quate correction for errors associated with serial correlation in a model including lagged dependent variables; and failure to analyze the aggre- gate data at a state level to discern whether the conclusions were re- flected in individual states. While the results of the reanalysis do not demonstrate that there is no causal relationship between prison capacity and prison population indeed, anecdotal evidence supports such a relationshi~it is clear that the relationship is complex, that the construction decision rests on a number of factors that stimulate or discourage building, that conditions vary greatly from state to state, and that further research is needed to explain the prison construction-prison population relationship. If prison i3 The replication was made possible by Carlson, who generously made the data tape available to the panel. It was initiated by Alfred Blumstein and carried out at Carnegie- Mellon University, and the findings are available in Blumstein et al. (1983~. The data tape contained reported prison population and reported increases in prison capacity for each of the 50 states and the District of Columbia for each year from 1955 to 1976 (1,122 cases) plus one observation from 1954 and four from 1977, for a total of 1,127 cases.

250 RESEARCH ON SENTENCING THE SEARCH FOR REFORM population forecasts are correct, populations in a number of states should decline in the l990s without policy changes. This likely situation provides an opportunity to test the capacity model directly to determine whether and under what circumstances the availability of "spare" capacity affects the threshhold for selecting offenders to be incarcerated in order to fill the space. ALTERNATIVES TO INCARCERATION The effort to develop alternatives to incarceration and community-based corrections programs was stimulated by the President's Commission on Law Enforcement and the Administration of Justice (1967), which called for the integration of offenders into the community rather than their removal from it. In the early 1970s, when prisons in most states were not under population pressure, a variety of alternative programs were initiated to alter traditional case processing by prosecutors, provide alternative sanctions to prison and jail confinement, reduce the use of secure confinement facilities, and provide alternatives to continual con- finement in state prisons. The initiatives included pretrial diversion, restitution and community service programs at all stages of the criminal justice process, increased use of probation and intensive community supervision, development of halfway houses, early release programs, and statewide community corrections legislation. As prison populations mushroomed between 1972 and 1978 and per- sistent evidence indicated the unproductive effect of rehabilitation pro- grams in prisons, many groups pressed for greater use of community- based sanctions instead of incarceration for nondangerous offenders. While some advocated these programs as a way of reducing prison populations, others regarded alternative sanctions as more punitive al- ternatives to simple probation and fines and as a way of providing supervision and control of those offenders who were released into the community. From the perspective of the pressure of growing prison populations, our concern is the extent to which the proliferation of alternative sanc- tions actually displaced or reduced incarceration. Little of the existing research has been designed to answer this question. What limited evi- dence there is, however, suggests that alternative sanctions have more frequently increased the level of nonincarcerative punishment for those offenders who otherwise would not have been incarcerated than they have served as an alternative sanction for those who otherwise would have been incarcerated. Rather than reducing the use of incarceration for certain types of offenders, alternative programs have extended the

Impact on Prison Populations 251 level and scope of formal mechanisms of social control exercised by the criminal justice system (Austin and Krisberg, 19824. For example, many persons who previously would have had their cases dismissed or been given a nominal sanction are now subject to greater supervision by the state through use of pretrial diversion programs. And restitution and community service sentences often have been added to probation or incarcerative sentences, compounding the amount and duration of pun- ishment received by minor offenders. Indeed, it appears that most res- titution and community service programs were established to serve as supplements to probation and parole supervision or fines imposed on minor offenders (Austin and Krisberg, 1982~. Those programs delib- erately designed to reduce incarceration do not appear to have been effective in doing so (Flowers, 1977; Schneider and Schneider, 1979; Pease et al., 1977~. Postincarcerative release options such as work release, work furlough, halfway houses, and prerelease facilities have been designed to permit incarcerated offenders to move to lower-security facilities or to com- munity supervision several months prior to parole or conditional release. One study that examined the impact of community-based correctional programs on prison populations (Hylton, 1980) found that prison pop- ulations increased significantly between 1962 and 1979 in Saskatchewan, Canada, despite the introduction of community corrections programs. Hylton's failure to control for or examine increases in crime and police arrests and their potential effect on prison populations weakens confi- dence in his conclusion that community corrections programs had little effect. Bass's (1975) study of California's work furlough program re- ported that it experienced high rates of violations and technical escapes and, consequently, resulted in increasing the rate and length of incar- ceration for many violators. Although postrelease alternatives have re- moved incarcerated offenders from prisons earlier than they might other- wise have been released, the empty beds have been filled quickly by new admissions from the ample pool of convicted offenders eligible for incarceration. Four states (California, Colorado, Minnesota, and Oregon) have adopted community corrections acts intended to encourage communities to treat offenders locally rather than send them to state prisons by providing financial subsidies for local programs. Because the Colorado and Oregon laws are relatively recent (1976 and 1978, respectively), impact data are limited. Studies of the California probation subsidy program initiated in 1965 suggest that, although it succeeded in shifting responsibility for offenders formerly destined for state facilities to local jurisdictions, it increased the rate of incarceration at the local level and

252 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM the proportion of persons under some type of criminal justice system supervision (Lemert and Dill, 1978; Lerman, 1975; Miller, 1980~. A recent evaluation of Minnesota's Community Corrections Act (CCA) (Strathman et al., 1981) suggests that CCA-supported programs were being used to augment local sentencing options, previously limited to jail and probation but were having negligible impact on state prison populations. Unfortunately, the CCA evaluation was oriented toward assessing outcomes with little attention to illuminating the processes underlying them. Neither the question of why the financial incentive to handle offenders locally (on which the act was premised) was less ef- fective than had been expected nor the impact of other changes in sentencing policy that occurred at the same time (including the adoption of parole guidelines) was addressed in the report. The finding that alternative sanctions have not served as alternatives to incarceration is disappointing to those who sought to reduce impris- onment rates, but it is hardly surprising. Austin and Kr~sberg (1982) suggest that programs designed as alternatives to incarceration, like many other "reforms," have failed due to a combination of circum- stances. These include the interests, values, and power of key decision makers and criminal justice system agencies that oppose reductions in incarceration (including police, prosecutors, judges, and corrections ad- ministrators); the limited economic and political clout of probation and parole agencies and private reform organizations that support wider use of alternatives to incarceration; the effectiveness of powerful agencies in reshaping innovations to serve their own interests, particularly through redefinition of the client population; public concern with "lenient" (i.e., nonincarcerative) sentences given to serious offenders; and the diverse and often conflicting objectives of supporters of alternative programs. MECHANISMS TO CONTROL INFLOW AND RELEASE OF PRISONERS A third approach to maintaining an equilibrium between the population and capacity of prisons and jails is to directly regulate the inflow and release of prisoners. Some state judges have taken prison crowding into consideration by refusing to send offenders to overcrowded institutions or sending them to jail to serve prison terms.~4 Others concerned with crowding (Blumstein and Kadane, forthcoming; Manson, 1981) have i4 For example, the chief justice of the Massachusetts Superior Court would not sentence offenders to the Massachusetts Correctional Institution at Concord due to overcrowding (WCVB-TV Editorial, March 14, 1975, cited in Mullen et al. [1980:1433~.

Impact on Prison Populations 253 suggested more formal inflow control mechanisms such as allocating existing prison space to sentencing judges. The only state to have adopted an explicit limit on the inflow of offenders to state prisons (but not to jails) is Minnesota. The legislation creating the Sentencing Guidelines Commission required the commis- sion to "take into substantial consideration . . . correctional resources including but not limited to the capacities of local and state correctional facilities." The commission interpreted the law as a directive that the guidelines not generate prison populations that exceed the capacities of state institutions. The commission's prison population projection model permitted it to test various sentencing options, consider only those that did not increase populations beyond existing capacity, and finally adopt an imprisonment policy and sanction levels that would maintain prison populations at about 95 percent of existing capacity, assuming no change in crime rates or sentencing laws, for 5 years. The guidelines have been in effect since May 1980. Thus far, destabilizing policy changes have been limited: in 1980 the legislature increased the mandatory minimum sentence for possession and use of a firearm but has defeated several more drastic bills, and the commission has withstood pressures to in- crease sentence severity. As a result, Minnesota was one of the few states to reduce prison population in 1980 and in the first half of 1981 (U.S. Department of Justice, 1981b). Inflow mechanisms may be effective planning tools for allocating ex- isting space in prisons, but they cannot provide immediate relief when prison overcrowding occurs. To handle such situations, a variety of discretionary release controls are currently in use as population safety valves, the most important of which is parole. In many jurisdictions the sentencing judge maintains parole release authority over offenders sent to the local jail; when population pressures become severe, early release for jailed offenders is authorized. State parole boards have sometimes acted to control prison populations by adopting accelerated parole re- lease policies when crowding or administrative concerns require it; many continue to do so. For example, in 1961 the California legislature approved a program based on screening of inmates by base expectancy (parole prediction) scores combined with programs for more intensive parole supervision. By 1963 the Department of Corrections' research division had screened the entire prisoner population, a number of prisoners were referred for parole consideration earlier than originally scheduled, and some of these were released on parole by the Adult Authority. The Department of Corrections estimated that the program had reduced the prison popu- lation by more than 840 offenders and had saved at least $840,000. In

254 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM the early and mid-1970s the California Adult Authority again lowered prison populations dramatically by informally changing its parole release strategy. Recently, the Mississippi legislature, in response to a court order, authorized "early parole" and "supervised earned release" (Mul- len and Smith, 1980:1234. Under a similar court order, Maryland's parole board authorized extended preparole furloughs of 6~90 days for non- violent offenders (Ney, 1980:8~. In Oregon in May 1980 the parole board, working closely with the Department of Corrections, modified the history risk component of the guidelines to make the earlier release of some inmates possible in December 1980 to comply with a court order directing the Department of Corrections to reduce prison crowding (Ar- thur D. Little, 1981~. Whatever the other shortcomings of parole boards, their ability to manage prison population size is a valuable feature at a time when the number of inmates exceeds prison capacity. A parole board's insulation from political pressure enables it to make necessary but unpopular early release decisions more quickly and unobtrusively than can a governor or legislature. As part of the movement to determinate sentencing, several states have abolished their parole authorities and have adopted fixed sen- tences; others have proposed such changes. Whatever the merits of these changes, they have curtailed the ability of a centralized release authority to use early release as a population management tool. Several states that had substantial recent increases in prison populations have found it necessary to adopt alternative means of early release, including emer- gency release powers acts, executive clemency, and increases in the rate at which good time is awarded. Michigan's Prison Overcrowding Emergency Powers Act of 1980 (Public Act 519) provides that, if the state's prison system is overcrowded for 30 consecutive days, the governor shall declare a prison overcrowding state of emergency. This declaration will reduce by 90 days the minimum sentences of all prisoners who have minimum terms. The result is an enlargement of the pool of prisoners eligible for parole by making in- mates eligible for parole release earlier than they otherwise would have been. The parole board then makes case-by-case release decisions. If the 90-day sentence reduction does not result in reduction of prison population to 95 percent of rated capacity within 90 days of the decla- ration of the state of emergency, minimum sentences will be reduced by another 90 days, increasing further the pool of prisoners eligible for parole. The governor must rescind the state of emergency once the population is reduced to 95 percent of rated capacity. The act was first

Impact on Prison Populations 255 invoked in the summer of 1981, permitting a reduction in prison pop- ulation. Various forms of executive clemency date back to colonial times in the United States, but information on their historical or contemporary uses is limited.~5 Until the end of the nineteenth century, executive clemency was the only way to obtain early release from prison. With the adoption of indeterminate sentences, parole boards were created to regularize release procedures; nevertheless, states retained the executive clemency authority, as a safety valve for providing mercy and dealing with extraordinary circumstances and organizational problems (Bere- cochea, 1982~. Existing state clemency structures, eligibility criteria, and decision- making procedures vary widely. In 31 states the governor has sole au- thority to grant clemency; in 10 states authority rests entirely in the hands of a special board; and in 9 states the governor's authority is limited to granting clemency to applicants receiving affirmative rec- ommendations from a special board or advisory body (Stafford, 1977~. A recent study of the uses of sentence commutation the form of clemency that reduces a sentence and is most frequently used to grant early release for a prisoner found that regular commutations are granted very sparingly (Martin, 1982b). For example, in Illinois between 1977 and 1981, there were an average of 160 applications per year for com- mutation, with only an average of 9 granted per year. In Texas, which has very narrow grounds for commuting a sentence, there were an av- erage of 20 applications and 15 commutations annually during those years. And in New York, with a prison population of more than 25,000 inmates in 1981, a total of 102 sentences were commuted between 1977 and 1981; 50 of those were granted under a special commutation pro- cedure adopted to reduce sentences of offenders given mandatory min- imum sentences of more than 15 years under the Rockefeller drug laws after those laws were revised in 1979. Special commutations have been used by 5 states Georgia, Mary- land, Oklahoma, Utah, and Wyomin~to reduce prison populations by releasing large numbers of inmates, generally those imprisoned for is There are several types of clemency: pardons, which usually involve a recognition of guilt but the need to mitigate the penalty (or remove a civil disability); commutations, which substitute a less severe punishment for that originally imposed (often reducing a minimum sentence, thereby making the offender eligible for earlier supervised parole release); and reprieves, which postpone the execution of a sentence, particularly in capital cases.

256 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM nonviolent offenses who are within 6 months of parole release. (Most other recipients of commutations have life or very long sentences.) In Utah between November 1981 and March 31, 1982, in response to a declaration of crowding by the director of the Division of Corrections, the Board of Pardons advanced the release dates of 93 inmates within 90 days of parole release. Maryland granted 543 "seasonal" commu- tations in 1978 and 297 in 1979 to inmates who had served at least one- half their sentences for nonviolent offenses and who were nearing parole eligibility. And in Georgia between July 1, 1980, and June 30, 1981, the parole board first released 2,436 inmates on special paroles and reprieves with conditional commutations and then released 2,001 more inmates up to 6 months early on special commutations without parole supervision (Martin, 1982b). A third, also limited, population control mechanism is use of good time. Prison authorities may affect the time served on a sentence by the grant, forfeiture, and restoration of statutory and meritorious good time. If today's nominal sentences were served without good-time reductions, prisons would be far more crowded than they are. But statutory good time is of limited value as a population control mechanism in many jurisdictions because it is automatically credited to prisoners at the be- ginning of their sentence and thereafter may only be taken away for a disciplinary infraction. Furthermore, its use as a mechanism for disci- plining inmates by adding time to be served back onto their sentences conflicts with the goal of reducing prison populations. For that reason, several state agencies, including the Illinois Department of Corrections, have taken steps to limit forfeiture of good time (Jacobs, 1982) and have given prison officials more flexibility in awarding meritorious good time. In Illinois, for example,, the Director of Corrections has wide power to reward a prisoner who performs meritorious service by granting up to 90 days additional good time. Although the effect of good-time provisions may not be realized immediately, and good time poses a greater risk of arbitrary application than uniformly applied emergency- power release provisions, it can reduce prison populations. Offender classification provides a fourth tool for addressing some of the problems of overcrowding. Increases in crowding have tended to undermine existing classification procedures by increasing the frequency of assignment on the basis of available space. This situation has resulted in a vicious circle of misclassification, which can retard offenders' prog- ress through the prison system, thereby leading to longer terms, con- tinued crowding problems, and classification errors. In Alabama, for example, the team involved in court-ordered classification of the entire prison system found that at least one-half of the inmates could be clas-

Impact on Prison Populations 257 sifted for minimum or community custody although only 10 percent were so classified (Clements, 1982:75~. Solomon (1980) reports that two- thirds of prisoners classified as needing medium security in Tennessee required only minimum security. Crowding is worst in maximum security prisons. Here, due to overcrowding, jobs and other opportunities are reduced, offenders have less chance to demonstrate "progress" or ad- justment, and when they do not meet the criteria of demonstrating "improvement" their movement out of the system is slowed. Comprehensive classification criteria for management purposes, fol- lowing the principle of using the least restrictive alternative possible, should help break this vicious circle. Consistent application of such criteria should relieve crowding, particularly in maximum security in- stitutions and, by increasing opportunities for program participation and "normalization," should lead to swifter movement of inmates through the prison system. IMPACT OF ALTERNATIVE STRATEGIES In the face of crowded prisons, rising prison populations, court orders to reduce crowding and improve prison conditions, and determinate sentencing laws that limit system flexibility, policy makers in every state must develop their own strategies to maintain a balance between pop- ulation and capacity through a combination of construction to expand capacity, increased use of alternatives to incarceration, and systematic use of inflow and release control mechanisms. Research can facilitate decision making by systematically examining the implementation and effects of policies and programs in various jurisdictions and by projecting the effects of policy options under a variety of conditions. Every option has both short-term and long-term advantages and costs. Construction may be necessary to replace obsolete facilities or expand capacity in systems that have long-term expected increases in inmate populations. But prison construction is very costly, and these costs are compounded by steady increases in operating costs. Alternative policies may be more cost-effective ways of preventing crowding and avoiding the costs of new construction in jurisdictions with short-term population pressures but long-term expectations of decreased inmate populations. Because it takes about 5 years to complete construction of a new prison facility, expansion of capacity by new construction will not solve the immediate capacity needs of many jurisdictions and may have the long- term effect of increasing what is viewed as the "normal" size of the prison population.

258 RESEARCH ON SENTENCING: THE SEARCH FOR REFORM Nonincarcerative programs often are advertised as less costly and more humane than incarceration (see National Council on Crime and Delinquency, 1980; Thalheimer, 1978), but others (e.g., Greenberg, 1975; Strathman et al., 1981) have expressed doubt that alternative programs result in actual savings, more humane sanctions, and reduced recidivism, or even that they are functioning as alternatives to incar- ceration. Although alternative programs promise some relief for prison crowding and may be appropriate and less costly ways of dealing with some nonviolent offenders, institutional pressures for "success" and public resistance to community facilities suggest their continued use predominantly for offenders who are unlikely to be imprisoned, thus limiting their short-term ability to provide substantial relief for prison crowding. Prison population control mechanisms appear to offer the greatest opportunities for short-term relief from crowding. Explicit control of prisoner intake, while desirable, requires a high degree of political con- sensus, shared social attitudes toward crime control, and agreement on a decision rule or formula for determining who should be incarcerated; such consensus is not likely to prevail or be easily developed in many jurisdictions. Early release of large numbers of prisoners through ex- panded use of early parole release, executive clemency, or emergency powers acts—can be implemented quickly, is less costly than construc- tion or alternative programs (in the short run), is preferable to reliance on less visible ad hoc adaptive mechanisms that are likely to prevail otherwise, and is more flexible in emergency situations than intake controls. In a situation of sudden and severe overcrowding or an emer- gency such as a natural disaster, a prison release mechanism permits reduction of all terms or only those of certain types of offenders by a fixed amount of time to provide immediate relief to the corrections system, while intake controls cannot deal with prison populations after inmates are committed. Furthermore, if social attitudes or sentencing policies change, leading to different sentences for offenders whose of- fenses are similar but who are convicted several years apart, these dif- ferences can be addressed by a parole board or some other early release mechanism. In sum, the current state of knowledge is uncertain regarding the effects and effectiveness of various alternatives for dealing with ex- panding prison populations. What is clear, however, is that the link between sentencing policy and prison populations should be considered when developing new sentencing policies. To ensure such consideration, some formal means should be developed in each state.to provide regular projections of prison populations and assessments of the likely impact of proposed policy changes.