Indeterminate and Determinate Sentencing Models

Abstract

This article compares the effects of indeterminate and determinate sentencing models on recidivism using a measure of parole board discretionary release and mandatory parole release under each sentencing model. Data collected from Recidivism of Prisoners Released in 1994: United States are used to conduct a state-specific comparison of the two release programs in six mixed-sentencing states. The results indicate that the effects of different sentencing models significantly vary across the six states. Whereas mandatory parole release was more likely to have a deterrent effect on recidivism in Maryland and Virginia, parole board discretionary release was more effective in New York and North Carolina. Release programs in Oregon and Texas showed no significant differences in their effects on recidivism.

Keywords

indeterminate sentencing determinate sentencing discretionary release mandatory parole release recidivism

The United States has no single sentencing model today. Until the 1970s, indeterminate sentencing was dominant, and most states gave judges and parole boards wide discretion in sentencing offenders. Under indeterminate sentencing, at the time of conviction, an inmate did not know the length of his or her criminal sentence; rather, criminal sentences were fashioned around an inmate’s specific treatment needs. Parole boards used their discretion to release offenders when it was appropriate to return them to the community (Bureau of Justice Assistance, 1998). For decades, this rehabilitative approach was the paramount goal of corrections (Cullen & Gilbert, 1982; Frankel, 1972; Hollin, 2000).

Determinate sentencing is based on deterrence theory, which predicts that persons who receive harsh sentences will be less likely to commit future crime. Through incapacitation, determinate sentencing is designed to increase punishment’s certainty, severity, and celerity, by removing discretion from unpredictable judges and parole boards (Greenwood & Abrahamse, 1982). Under determinate sentencing, state statues determine the length of incarceration, removing discretion from parole boards (Bureau of Justice Assistance, 1998). Under this model, prisoners are paroled through mandatory release, and mandatory release dates are automatically calculated and known to all parties (Ireland & Prause, 2005; Reitz & Reitz, 1993).

Given these different sentencing models with different philosophies and assumptions about their effectiveness on recidivism, there is a pressing need to assess the effects of these two distinct sentencing models on recidivism. Although researchers have examined single treatment or punishment programs under different sentencing models using true experimental or quasi-experimental designs (see Lipsey & Cullen, 2007, for a review of these studies), few scholars have conducted a direct assessment of the effects of the two sentencing models on recidivism. The current article attempts to fill this gap in the literature by using available data from the data set Recidivism of Prisoners Released in 1994: United States published by the U.S. Department of Justice, Bureau of Justice Statistics (2002).

Research Context

Indeterminate Sentencing, Rehabilitation, and Recidivism

Since the end of the 19th century, indeterminate sentencing existed in almost every American jurisdiction (Tonry, 2000). Indeterminate sentencing was designed so correctional systems could rehabilitate offenders (Cullen & Gilbert, 1982; Lipsey & Cullen, 2007). Although indeterminate sentencing has been under attack since the 1970s and has been abolished by a number of states, indeterminate sentencing has survived in many jurisdictions with rehabilitation as a dominant correctional philosophy (Tonry, 1999). Some scholars have argued that there has been a social movement recently to “reaffirm rehabilitation” (e.g., Cullen, 2005). Even so, indeterminate sentencing and its rehabilitative ideals have been historically challenged and discredited by researchers, politicians, and the public. These criticisms were largely attributed to the work of Martinson and his colleagues.

Prior to Martinson’s 1974 article (which concluded that “nothing works”) in The Public Interest, the New York State Governor’s Special Committee on Criminal Offenders hired Martinson and his colleagues in 1966 to undertake a comprehensive survey of what was known about rehabilitation. They reviewed 231 studies and 11 different types of prison rehabilitative programs, from 1945 to 1967, that focused on educational and vocational programs, individual counseling, group counseling, programs that reduced and enhanced treatment, and medical programs. Their findings indicated that “with few and isolated experiences, the rehabilitative efforts that have been reported so far have had no appreciable effect on recidivism” (Martinson, 1974, p. 25).

For the succeeding three decades, the “nothing works” doctrine was adopted by conservative political forces, nudging policy makers away from rehabilitative approaches (Stemen, Rengifo, & Wilson, 2005). Subsequent studies throughout the 1970s (e.g., Brody, 1976; Lipton, Martinson, & Wilks, 1975) also claimed that correctional treatment did not reduce recidivism (Travis & O’Leary, 1979), creating an environment among the public and in legislatures that called for tough crime control policies. All of these forces coincided with incapacitation studies by researchers from the Rand Corporation that were instrumental in moving policy makers away from rehabilitative approaches (Greenwood & Abrahamse, 1982; Petersilia, Greenwood, & Lavin, 1977). As a result, there was a major shift in correctional philosophy and policy away from rehabilitation toward primary crime prevention, incapacitation, and deterrence.

The shift was coupled with critiques of the discretionary power of parole authorities under indeterminate sentencing (Travis & O’Leary, 1979). The main criticism centered on the justification for granting discretionary parole authority to parole boards. Two assumptions underlying the exercise of that power were subject to attack: the efficacy of treatment and the ability to predict future criminal behavior. With respect to rehabilitation, it would undermine indeterminate sentencing if paroled offenders recidivated after being released pursuant to a correctional treatment program. As for predicting future crime, if decision makers are unable to judge accurately an offender’s potential for future criminality, then the discretion secured in the name of public protection is unwarranted.

Although “nothing works” was favorably received by the public and by policy makers, many within the academic criminal justice and criminological communities challenged its conclusions. Palmer (1976, p. 179) reanalyzed Martinson’s data, pointing out that “several of the positive and partly positive results reported have given direct support to the view that the use of certain intervention methods would in fact be appropriate, at least under certain conditions.” Thus, Palmer’s critique suggested that Martinson’s analysis of program effectiveness did not capture the true impact of treatment on recidivism. Rather, Palmer (1975, 1976) contended that the offender’s characteristics, the treatment setting, and the service providers themselves must also be examined in addition to program effectiveness. Palmer showed that different methods of intervention are more likely to be associated with less recidivism in relation to some offenders compared with others.

Criticizing Martinson’s methodology, Empey (1976) argued that Martinson’s work evaluated the effects of “treatment,” which included anything from informal probation to total incarceration as well as anything from individual psychotherapy to leisure-time activities. By simply summing up the results of a host of different studies, Empey suggested that Martinson’s report should not be taken as the final word on the overall effects of correctional intervention.

With improved analytical methods (i.e., meta-analysis), post-Martinson researchers also established that the “nothing works” thesis was overblown (MacKenzie, 2001, 2006; Martinson, 1979). Using meta-analysis, Garrett (1985) reviewed 111 studies published between 1960 and 1983, concluding that treatment of adjudicated delinquents in residential settings helped to reduce delinquency. In the late 1980s, Gendreau and Ross (1987) surveyed more than 200 studies on correctional treatment from 1981 to 1987 and suggested that successful rehabilitation of offenders had been accomplished; substantial reductions in recidivism had been achieved in a considerable number of well-controlled studies.

Losel (1995, p. 89) examined 12 meta-analyses published between 1985 and 1995, positing that “the mean effect size of all assessed studies probably has a size of about .10,” which implies that rehabilitation is likely to have a modest effect on reducing recidivism. Adding to Losel’s 1995 review, McGuire (2000) identified another six meta-analyses. When combined, the 18 meta-analytic reviews, published between 1985 and 2000, produced a mean reduction in recidivism of between 5% and 10%. McGuire further indicated that many individual studies, and even meta-analyses, reported effect sizes considerably larger, suggesting that some methods of correctional intervention are consistently more effective than others in reducing recidivism. More recently, Tonry (2000) and Lipsey and Cullen (2007) reported more optimism about the effectiveness of rehabilitation programs.

Determinate Sentencing, Deterrence, and Recidivism

Although the preceding review showed that the overall effects of rehabilitation reduced recidivism, especially based on results from studies after the 1980s, disillusionment with rehabilitation became endemic within correctional circles. Disenchantment with correctional treatment, coupled with cynicism toward parole board discretionary release decisions, made the political environment ripe for change. Thus, since the mid-1970s determinate sentencing has been popular with policy makers and legislators (Brewer, Beckett, & Holt, 1981).

Most states adopted determinate-sentencing rationales with fixed prison terms, where crime seriousness, not rehabilitation, served as the basis for establishing the length of prison sentences (Brewer et al., 1981). With determinate sentencing, states experimented with different types of parole, good time, commitment criteria, and the overall severity of punishment (Fogel, 1975; Frankel, 1972; Morris, 1974). Sentence length was determined mainly by the judge, with no discretionary release input by a parole board.

Although the dominant objectives varied from state to state, the goals for determinate sentencing were primarily deterrence, incapacitation, and “just desserts” (Bureau of Justice Assistance, 1998; Marvell & Moody, 1996). By 1978, determinate-sentencing laws had been passed in six states (Brewer et al., 1981), and by 1996, 14 states had determinate sentencing (Bureau of Justice Assistance, 1998). By 2000, 16 states had abolished discretionary parole for all offenders, and 4 states had abolished discretionary parole for certain classes of violent offenders (Bureau of Justice Statistics, 2004). In 1980, more than half of prisoners released from state prisons were under a parole board’s discretion (Travis & Lawrence, 2002). Between 1990 and 2000, however, mandatory parole release increased from 29% to 39%, whereas discretionary release decreased from 39% to 24% (Bureau of Justice Statistics, 2004). Although several studies evaluated the effects of rehabilitation on reducing crime under indeterminate-sentencing schemes, few studies have examined the effects of determinate-sentencing schemes on crime control (e.g., Marvell & Moody, 1991, 1996). Studies of determinate sentencing mainly focus on the impact of individual determinate-sentencing laws on prison admissions and prison populations (e.g., Brewer et al., 1981; Frase, 1993; Marvell & Moody, 1991, 1996). These studies have found little relationship between sentencing laws and crime rates (e.g., Marvell & Moody, 1991).

In sum, although there has been a drastic shift from indeterminate to determinate sentencing since the 1970s, the effect of indeterminate sentencing with parole board discretionary release authority has not been conclusively assessed. Moreover, scholars have not determined the impact of determinate sentencing with mandatory parole release on recidivism. The present article is to conduct an empirical comparison of indeterminate- and determinate-sentencing models, measuring their effects on recidivism.

Recidivism of Prisoners Released in 1994: United States and Related Studies on Prison Releases

As previously discussed, the principal difference between indeterminate and determinate sentencing is the method of how prisoners are released from prison. Under the indeterminate-sentencing scheme, prisoners are generally released through discretionary parole board decisions, whereas under the determinate-sentencing scheme prisoners are released pursuant to mandatory release. Therefore, the current article measures the indeterminate- and determinate-sentencing models using two prisoner release types: parole board release and mandatory release.

Although collected over a decade ago, the data in Recidivism of Prisoners Released in 1994: United States (U.S. Department of Justice, Bureau of Justice Statistics, 2002) provide a valuable opportunity to conduct an empirical study with these measures. The survey contained 20 release categories, ranging from 1 = parole board decision release to 99 = unknown. Among the 20 release categories, 2 major types existed: parole board decision release and mandatory parole release. Most states in the survey had parole board decision release, mandatory parole release, or a mix of the two release types. Three previous studies have used the data to assess the effects of the release categories in one way or another for different research purposes (Rosenfeld, Wallman, & Fornango, 2005; Solomon, Kachnowski, & Bhati, 2005; Bhati & Piquero, 2008).

Rosenfeld et al. (2005) used the data to assess the contribution of ex-prisoners to crime rates. These investigators also assessed a number of influential factors, including release types on recidivism. Along with other factors such as age, race, and prior records of arrest, the investigators adopted four release types (i.e., parole board release, mandatory release, unconditional release, and other release types); they created three dummy variables using unconditional release as the reference category for their analysis. The general findings from their analysis indicate that parole board release was most likely to be negatively associated with the probability of recidivism in violent, property, and drug offenses.

Rosenfeld and colleagues’ (2005) analysis suffers from two major limitations. First, they used a dependent variable that represented the total number of rearrests after release during a 3-year follow-up and conducted negative binomial regression analyses. Such a dependent variable may be problematic. If an ex-prisoner was rearrested for a serious crime and sentenced to prison, then the individual would not be able to commit further crime and be arrested again. Only those who committed minor offenses would be able to remain incarceration free and have chances to commit further crime during the 3-year follow-up.

Second, Rosenfeld et al. (2005) did not conduct state-specific analysis given that each state might have different release procedures and parole supervision programs (see Solomon et al., 2005). North Carolina, for example, passed the Prison Population Stabilization Act in 1987, which remained in effect until January 1, 1996, influencing dramatically the parole process during these years. The Parole Commission primarily paroled misdemeanants as a class of offenders rather than consider their individual threat-to-reoffend status. Many thousands of these offenders were moved in and out of prison quickly under a system called parole and terminate (North Carolina Department of Correction, 2008). Thus, in North Carolina, those under discretionary parole release were more likely to be misdemeanants.

In Oregon, since discretionary parole release was abolished in 1989, the Board of Parole and Post-Prison Supervision imposed prison terms and made release decisions on dangerous offenders, aggravated murderers, murderers, and offenders whose crimes occurred prior to November 1, 1989 (Oregon Board of Parole and Post-Prison Supervision, 2007); therefore, in 1994, prisoners released under discretionary parole in Oregon were more likely to be violent offenders. As a consequence, expected outcomes involving different practices in different states varied widely. Also, several states in the survey had a mix of release types (mainly a mix of parole board release and mandatory release), and others only had parole board release or mandatory release; hence, it is misleading with respect to release types to classify states solely into one category or another given state-specific variation.

Using the same data set, Solomon et al. (2005) attempted to assess whether parole supervision prevented recidivism. Their assessment concentrated on parole board release, mandatory release, and unconditional release. Both parole board release and mandatory release enacted parole supervision programs, whereas unconditional release had no parole supervision at all. The investigators conducted multivariate logistic regression analysis with rearrests as a dummy dependent variable. The analysis compared the three release types for the probability of rearrests along with a number of control variables such as age, gender, and prior arrest records. Solomon et al. found that when the control variables were held constant, the rearrest rate for mandatory release and unconditional release was identical, and there was no substantial difference between rates of parole board release and rates of unconditional release. As noted in the study, “The use, duration, and intensity of post-release supervision varied significantly across states” (Solomon et al., 2005, p. 14). Therefore, the study suffered a similar analytical limitation: An overall comparison of the three release types was conducted without giving serious consideration to significant variation across states. Also, the dummy dependent variable of rearrest could not measure the time variation of recidivism, which reflected on the possible risk level of reoffending.

With the same data set, Bhati and Piquero (2008) compared prior offending patterns with postprison offending trajectories to determine whether incarceration had an impact on subsequent offending trajectories. Bhati and Piquero also estimated several logistic regression models for possible influential factors on recidivism. They collapsed the states in the survey into four groups in terms of the information level of various release types and reorganized the release types for the state groups in their analysis. They reported that “the prison release mechanism seemed unrelated to the probability of an individual deterred” for most state groups (Bhati & Piquero, 2008, p. 249). Their dummy dependent variable of recidivism and their analytical strategy had the same limitations as Solomon et al. (2005).

In summary, the three previous studies attempted to examine the possible effects of release types on recidivism, yielding mixed findings. In addition, all three studies suffered similar methodological limitations with respect to the measurement of their dependent variable (recidivism); they also failed to consider state variation in sentencing and release programs and practices. Further research is needed to explore the data with different analytical and statistical techniques. The current study is designed to rectify these limitations.

Current Study

As mentioned, using the same data set, the present article attempts to overcome the limitations suffered by the three previous studies in two ways. First, we conduct state-specific analysis given that different states had different release and post-release supervision programs when the survey was conducted. The analysis is limited to states that used mixed-sentencing models, including both determinate sentencing and indeterminate sentencing, and had mandatory release and discretionary parole release programs under the two sentencing models. This within-state comparison method controls for differential correctional policies, programs, and practices between-states. This allows an opportunity to analyze and compare the effects of the two release programs under the two sentencing models on recidivism within similar state contexts. We chose not to compare states with purely determinate-sentencing models (e.g., California, Illinois, and Minnesota) to states with purely indeterminate-sentencing systems (e.g., Michigan and New Jersey) because contextual variation within states may generate different sentencing practices. This may exist even if both states are classified as determinate-sentencing states or indeterminate-sentencing states. As discussed previously, lumping states into homogeneous groups blurs state variations of sentencing practices, which in turn obscures the real cause of the effects on recidivism.

Examination of the data set indicates that six states—Maryland, Virginia, New York, North Carolina, Oregon, and Texas—were identified as having mixed-sentencing models. A majority of prisoners released in these states in 1994 went through the two release programs, that is, parole board release and mandatory release (see Methods section for a more detailed discussion of the data and measures). Although Oregon is considered a mixed-sentencing state, it is different from the other five states in this analysis. Oregon had both discretionary and mandatory release due to a sentencing law change in 1989 when the data were collected, causing old law and new law cases to appear in the data set. All the other five states are defined as mixed-sentencing states because they practiced mandatory and discretionary release at the same time.

Maryland has an indeterminate-sentencing system with voluntary sentencing guidelines that were introduced in 1983. The Maryland sentencing guidelines were “advisory and judges may, at their discretion, impose a sentence outside of the guidelines” (Maryland State Commission, 2009). In addition, Maryland enacted mandatory minimum sentences for certain offenses (Bureau of Justice Assistance, 1998; La Vigne, Kachnowski, Travis, Naser, & Visher, 2003). Prisoners in Maryland can be released through either a discretionary or mandatory release process (La Vigne et al., 2003). Hence, Maryland is also considered a mixed-sentencing state.

North Carolina enacted structured sentencing in 1994, which affected all felony and misdemeanor crimes (except driving while impaired) committed on or after October 1, 1994 (North Carolina Department of Correction, 2008). Virginia abolished discretionary parole release in 1995 (Keegan & Solomon, 2004). Prior to the abolition of parole release, both North Carolina and Virginia had indeterminate sentencing. New York and Texas were considered indeterminate-sentencing states prior to 1994; however, the 1996 National Survey of State Sentencing suggests that all of these states had additional mandatory sentencing policies (Bureau of Justice Assistance, 1998). Thus, North Carolina, Virginia, New York, and Texas had a mixed-sentencing system as well when the 1994 prisoner data were collected.

Although the state-specific method controls between-state differences, another concern is whether after release supervision programs differ between mandatory and discretionary release within the same state. The examination of after-release policies and programs in each of these six states suggests that prisoners released by discretionary parole and mandatory release were supervised in the community for a similar period of time. Indeed, in each of these six states, mandatory releases were subject to the same authority, rules, regulations, and community supervision conditions as discretionary parole releases (Keegan & Solomon, 2004; La Vigne et al., 2003; Watson et al., 2004). In Maryland, for example, all conditional releasees were supervised by the Maryland Division of Parole and Probation. In Texas, both parolees and mandatory releasees were subject to the supervision of the Texas Board of Pardons and Paroles. In the current analysis, there are no obvious spurious effects of after release programs within each state.

Second, the current analysis improves over the three previous studies by taking the length of time that an ex-offender relapsed into account (Kane, 2006). The dependent variable for our study was calculated as the time interval between the 1994 release and the first rearrest for each individual within 3 years. Such a measure more accurately reflects offenders’ recidivist statuses and the degree of the risk of reoffending than a measure with only two categories—recidivated and nonrecidivated.

Our key research question is: Does mandatory parole release under the determinate-sentencing model more effectively reduce recidivism than discretionary parole board release under the indeterminate-sentencing model? We performed survival analyses examining the relationship between the two release types and the time it took offenders to recidivate.

Methods

Data

The data used for the present article originated from the data set Recidivism of Prisoners Released in 1994: United States (U.S. Department of Justice, Bureau of Justice Statistics, 2002). The original data set contained 1994 information on 38,624 sampled prisoners out of the 302,309 released prisoners from 15 states that were tracked for 3 years following their release. Bureau of Justice Statistics employed a stratified sampling procedure, using convicted offenders as the stratification factor to select a representative sample within each state.¹ As discussed above, our analysis concentrates on six states, that is, Maryland, Virginia, New York, North Carolina, Oregon, and Texas.

Our criteria for case inclusion in the analysis were the same as the Bureau of Justice Statistics report (Langan & Levin, 2002), namely prisoners who had a RAP sheet in the state criminal history repository, were alive throughout the 3-year follow-up period, were given a sentence of more than 1 year, and were released either by parole board discretion or mandatory parole release. In addition, only offenders leaving prison for the first time since beginning their sentence were included in the analysis. Offenders who had a life or death sentence or an unknown sentence were excluded from the current analysis. Because Whites and Blacks accounted for more than 98% of the offenders released in 1994, other minorities were left out of the analysis because of very small numbers. Violent crime, property crime, drug offenses, and public order offenses comprised more than 99% of the cases; therefore, the analysis was limited to these four types of crimes. After we eliminated the missing values on the variables for the analysis, the sample sizes were 1,394 for Maryland, 1,853 for Virginia, 1,705 for New York, 1,836 for North Carolina, 1,220 for Oregon, and 1,782 for Texas.

Measures

The dependent variable for the current analysis was recidivism, which was measured by the survival time of prisoners between their 1994 release and the first rearrest within 3 years. The use of time to rearrest provides researchers with a measure that not only indicates the release success or failure but also illustrates both partial and complete success or failure during the observation period.² The original data set did not have an indicator measuring the survival time of each prisoner. The data set, however, provided information on each released prisoner’s entire officially recorded criminal history before and after his or her 1994 release. The current article started by identifying the first rearrest date after the 1994 release for each prisoner based on his or her criminal record. The length of time between the 1994 release and the first rearrest was then calculated to serve as a measure of the dependent variable.

The main independent variable analyzed was release type under determinate- and indeterminate-sentencing schemes. It was a dummy variable coded in the direction of release under a system based on determinate sentencing. Under determinate sentencing, prisoners were released from prison to parole supervision by virtue of state law that determined the length of time prisoners were incarcerated. Under indeterminate sentencing, prisoners were released by a parole board that decided if a prisoner was ready and under what conditions he or she was to be released from prison to a parole supervision program.

We included a number of important static or individual factors and dynamic or institutional factors as control variables that may also affect the likelihood of recidivism. The individual factors were age, gender, and race. Age was measured as the age of each prisoner (in months) at the time of release. Gender was a dummy variable coded in the direction of females. Race was also a dummy variable with 0 = White and 1 = Black, because the study sample only contained these two racial groups.

Following previous studies (Bales, Bedard, Quinn, Ensley, & Holley, 2005; Dejong, 1997; French & Gendreau, 2006; Spohn & Holleran, 2002), the institutional factors included severity of punishment, criminal history, and prison infractions. There were two measures of severity of punishment: time served (in months) before release and percentage of the total maximum sentence the prisoner had served when released.³ Criminal history was measured by the number of arrests prior to the date of release.⁴ Prison infractions were measured by whether a prisoner was formally disciplined for a rule violation in prison. Because Oregon and Virginia did not provide prisoner discipline records and Texas’s records showed that 99.6% of the prisoners had no infraction, the prison infraction variables were excluded from the analyses for these three states. Finally, types of offenses were controlled to eliminate the possible confounding effects of offenses on recidivism. As discussed above, the sample included four types of offenses; that is, violent, property, drug, and public-order offenses. We used violent offense as the reference category to create three dummy variables for the analysis (see Table 1 for descriptive statistics of the variables for each of the six selected states).

Table 1.

Descriptive Statistics of Variables for the Four Subsamples of Offenders.

Variable	Maryland	Virginia	New York	North Carolina	Oregon	Texas
Average length of time from release to first rearrest, months (SD)^a	10.8 (9.1)	13.1 (9.7)	11.8 (9.8)	12.7 (9.8)	8,9 (8.1)	13.6 (9.5)
Rearrest, %	67.3	58.4	58.4	55.4	69.3	48.7
Release type, %
Parole board decision	48.8	42.0	62.2	28.8	57.8	48.6
Mandatory parole release	51.2	58.0	37.8	71.2	42.2	51.4
Race, %
White	26.3	38.3	45.9	34.9	84.8	59.3
African American	73.7	61.7	54.1	65.1	15.2	40.7
Gender, %
Male	94.5	93.2	94.3	90.0	94.1	93.3
Female	5.5	6.8	5.7	10.0	5.9	6.7
Age at release, years, (SD)	32.7 (8.5)	33.1 (8.8)	33.7 (9.0)	31.5 (9.4)	35.5 (9.8)	35.3 (9.5)
No. of prior arrests, (SD)	6.7 (6.2)	6.2 (6.3)	8.3 (8.5)	5.1 (5.1)	12.8 (10.0)	5.9 (5.6)
Infraction record, %
Has record	52.2	—	64.9	44.9	—	.4
No record	47.8	—	19.4	55.1	—	99.6
Unknown	—	100	15.8	—	100	—
Time served in prison, months, (SD)	41.4 (46.9)	27.6 (26.2)	37.5 (36.0)	12.4 (19.5)	20.6 (26.5)	31.2 (33.5)
Percentage of time served for 1994 imprisonment, (SD)	49.2 (21.7)	34.7 (16.4)	51.4 (20.6)	13.3 (11.0)	34.6 (31.3)	23.3 (17.5)
Offense type, %
Violent	39.6	35.1	47.7	37.5	55.4	42.6
Property	23.3	26.0	22.7	26.4	21.8	24.2
Drug	26.2	25.7	18.1	25.5	15.8	20.6
Public order	10.9	13.2	11.4	10.6	7.0	12.5
N	1,394	1,853	1,705	1,836	1,220	1,782

Note. The average length of time from release to first rearrest was calculated only for prisoners who had been rearrested within 3 years.

Statistical Procedures

We applied Cox regression analyses to assess the impact of the two types of release—mandatory parole release under determinate sentencing and parole board discretionary release under indeterminate sentencing—on the survival time after release while controlling other individual and institutional factors for each of the six states. Cox regression is a survival analysis method for modeling time-to-event data in the presence of censored cases. In the current analysis, some of the prisoners committed crimes and were rearrested during the 3 years after their 1994 release. Some of the prisoners, however, had not been rearrested at the end of the 3-year period after release. These cases were called censored cases for which the rearrest was not recorded, even though these prisoners still have a chance to be involved in criminal activities and to be arrested in the future. Cox regression allows the inclusion of multiple covariates in the model while handling the censored cases correctly.

Results

Table 2 presents the results of Cox regressions for each of the six states. The omnibus tests of model fitness (chi-square tests) indicate each of the six models performs well. The results of the Maryland model indicate that release types were a significant predictor of recidivism (b = 0.17) when other important factors were held constant. Offenders under “mandatory release” were likely to have a shorter survival time than those under “parole board release.”

Table 2.

Cox Regressions of Survival Time of Violent Offenders on Release Types Along With Control Variables.

Variable	Maryland	Virginia	New York	North Carolina	Oregon	Texas
Release type^a
Parole board decision
Mandatory parole release	0.17	0.40	−0.31	−0.20	0.014	0.10
	(0.07)*	(0.08)***	(0.08)***	(0.08)*	(0.104)	(0.07)
	[1.19]	[1.49]	[0.73]	[0.82]	[1.01]	[1.11]
Race^a
White
African American	0.32	0.50	0.23	0.36	0.33	0.20
	(0.08)***	(0.07)***	(0.07)***	(0.07)***	(0.09)***	(0.07)**
	[1.37]	[1.65]	[1.26]	[1.43]	[1.39]	[1.22]
Gender^a
Male
Female	−0.13	−0.20	−0.32	−0.26	−0.10	−0.17
	(0.15)	(0.13)	(0.15)*	(0.12)*	(0.14)	(0.14)
	[0.88]	[0.82]	[0.73]	[0.77]	[0.90]	[0.85]
Age at release^a	−0.06	−0.05	−0.05	−0.05	−0.05	−0.05
	(0.005)***	(0.005)***	(0.005)***	(0.005)***	(0.005)***	(0.005)***
	[0.95]	[0.95]	[0.96]	[0.95]	[0.95]	[0.95]
No. of prior arrests^a	0.07	0.05	0.03	0.07	0.05	0.05
	(0.005)***	(0.005)***	(0.002)***	(0.006)***	(0.004)***	(0.006)***
	[1.07]	[1.06]	[1.03]	[1.07]	[1.05]	[1.06]
Infraction record^a
Yes
No	−0.24	—	−0.36	−0.39	—	—
	(0.07)***		(0.10)***	(0.07)***
	[0.79]		[0.70]	[0.67]
Unknown	—	—	−0.06	—	—	—
			(0.09)
Time served in prison (months)^a	0.002	0.001	−0.001	−0.001	−0.006	−0.003
	(0.001)*	(0.002)	(0.001)	(0.003)	(0.002)**	(0.002)
	[1.002]	[1.001]	[0.999]	[0.999]	[0.994]	[0.997]
Percentage of time served for 1994 imprisonment^a	0.000	−0.005	0.007	−0.006	−0.005	0.007
	(0.002)	(0.003)	(0.002)**	(0.005)	(0.002)*	(0.002)**
	[1.000]	[0.995]	[1.007]	[0.994]	[0.995]	[1.007]
Offense type^a
Violent
Property	0.22	0.22	0.40	0.30	−0.17	0.52
	(0.09)*	(0.08)**	(0.10)***	(0.09)***	(0.09)	(0.09)***
	[1.24]	[1.24]	[1.49]	[1.35]	[0.84]	[1.67]
Drug	0.02	−0.01	0.23	0.12	−0.24	0.40
	(0.09)	(0.10)	(0.10)*	(0.10)	(0.10)*	(0.10)***
	[1.02]	[0.988]	[1.26]	[1.13]	[0.78]	[1.49]
Public order	0.04	0.24	0.20	0.14	0.05	0.24
	(0.12)	(0.11)*	(0.12)	(0.12)	(0.13)	(0.12)*
	[1.04]	[1.27]	[1.22]	[1.15]	[1.05]	[1.27]
N	1,394	1,853	1,705	1,836	1,220	1,782
Model χ² (df)	294 (11)***	290 (10)***	422 (12)***	357 (11)***	350 (10)***	252 (11)***

Values are B (SE) [Exp (B)].

p < .05. **p < .01. ***p < .001.

Individual factors (except gender) were all significant in predicting survival time of releases in Maryland. Black offenders were more likely to recidivate earlier than White offenders. Offenders who were younger and who had more prior arrests were also likely to recidivate more quickly than those who were older and had fewer prior arrests. Two institutional factors were also significantly related to offender recidivism: Offenders with previous prison infractions and those who had served a longer time in prison were more likely to recidivate earlier than those who had no records. Finally, the risk of recidivism for property offenders was significantly different from violent offenders. No significant effect was found for other types of offenders. The overall model chi-square was 294 (df = 11), with p < .001.

The results of the Virginia model were very similar to the results of the Maryland model. Offenders under “mandatory release” were also likely to have a shorter survival time than those under “parole board release” (b = 0.40). Race, age, and prior arrest records also had significant effects on recidivism in Virginia that were similar to those observed in Maryland. In Virginia, however, the risk of rearrest for public order offenses was significantly higher than that for violent offenses discovered in Maryland. The overall model chi-square was 290 (df = 10), with p < .001.

For the New York and North Carolina models, the results were contrary to those for the Maryland and Virginia models in terms of the effect of release types. Under the New York and North Carolina models, offenders who had “parole board release” were likely to have a shorter survival time than those with “mandatory release.” The effects of race, age, and prior arrest records in New York and North Carolina, however, were quite similar to those observed in Maryland and Virginia. In contrast, the results show that males in the New York and North Carolina models were more likely to be rearrested than females. Prison infraction records in these two states also had a similar effect on recidivism as observed in Maryland. Percentage of time served for the imprisonment was a significant predictor of rearrest in New York. Finally, as discovered in the Maryland model, released property offenders had a higher risk of rearrest in both the New York and North Carolina models. The New York model showed a modest effect of drug offenses on the probability of rearrest. The overall model chi-square for New York was 422 (df = 12, p < .001); for North Carolina the overall model chi-square was 357 (df = 11, p < .001).

The Oregon and Texas models had different patterns. Release types in both states had no significant impact on recidivism. Race, age, and prior arrest records in the Oregon and Texas models, however, had similar effects as in the other states. In the Oregon model, both time served in prison and percent of time served for imprisonment had significant, negative effects on recidivism, meaning that ex-prisoners who served less time in prison and served a smaller percentage of their prison time were likely to be rearrested in a shorter time period. In contrast, the Texas model showed an effect of the percent of time served for imprisonment similar to that indicated in the New York model. Finally, the Oregon model showed that drug offenders had a significantly lower risk of rearrest than violent offenders, which differed from other states. In Texas, the rearrest probability for violent offenses was significantly lower than for any other type of offense. The overall model chi-square was 350 (df = 10, p < .001) for Oregon model and 252 (df = 11, p < .001) for Texas model.⁵

Discussion and Conclusion

Using available data from the Recidivism of Prisoners Released in 1994: United States (U.S. Department of Justice, Bureau of Justice Statistics, 2002), the present article assesses the effects of indeterminate- and determinate-sentencing models as measured by parole board discretionary release and mandatory parole release on recidivism for six states that had the two sentencing models when the data were collected. Our state-specific survival analyses of the six states (i.e., Maryland, Virginia, New York, North Carolina, Oregon, and Texas) reveal several interesting findings.

First, the six states have different patterns with respect to the effect of the release types. For Maryland and Virginia, the mandatory release program under the determinate-sentencing model was more likely to reduce the risk of recidivism than the parole board release program when other important individual and institutional factors were controlled. In contrast, the New York and North Carolina models show a reversed pattern. In each of these states, parole board release was associated with a lower risk of recidivism than mandatory parole release. Finally, the models in both Oregon and Texas indicate no significant difference between the two release types for the risk of recidivism. These mixed findings may imply that whether a particular sentencing model works is largely contingent on the actual release and after release parole programs in a specific state. The long debate between the effects of the two sentencing models may be too theoretical and political. How a state actually operationalizes and implements its models may be significantly related to the effects of the two models. Scholars, policy makers, and practitioners should not dismiss the indeterminate model and its related rehabilitative efforts. As others have noted, the conclusion that “nothing works” was premature.

Because state correctional programs and policies change over time, it is difficult to review and analyze how the states operationalized and implemented their two sentencing models when the data were collected in 1994. This article, however, makes an effort to conduct such a review. In Maryland, for example, prisoners released on discretionary parole served shorter terms. Serving shorter terms might make them ineligible for prison programming designed to prepare them for release (La Vigne et al., 2003). This might explain why in Maryland, mandatory release under determinate sentencing reduced the risk of recidivism.

As noted above, North Carolina enacted the 1987 Prison Population Stabilization Act, meaning that those under discretionary parole release in 1994 were more likely misdemeanants. This could help explain that in North Carolina the risk of recidivism of parole board releasees was lower than for mandatory parole releasees. Research is needed to analyze the actual operation and implementation of the two models in each state to provide possible explanations of why the two models have different effects in different states.

We believed that collapsing all states for the analysis masked between-state differences. That is, understanding rehabilitation requires attention to the operation of specific state programs. The effects on recidivism may depend more on the workings of post-release supervision policies and rehabilitation programs in specific states than from sentencing models themselves. In this way, our findings differ from the three previous studies that analyzed the same data set (Bhati & Piquero, 2008; Rosenfeld et al., 2005; Solomon et al., 2005). Our state-specific analyses with a measure of survival time of offenders as the dependent variable corrected for the deficiencies of these previous studies. Such analyses are much more sensitive to a state-specific context, and the dependent variable is more time sensitive to the status of reoffending.

The findings also indicate that among some variables, the effects of race, age, and prior arrest records on recidivism are not contingent on state contextualization. They all have significant effects on the risk of recidivism across the six states. Similarly, the prison infraction records variable seems to have a similar effect on the risk of recidivism in three states (e.g., Maryland, New York, and North Carolina) that had records for this analysis. The effects of time served in prison and the percent of time served for the imprisonment were inconsistent and even contrary across the six states. In Oregon, the results indicated that releasees who served less time in prison and served less percent of the imprisonment time had a higher risk of rearrests.

Finally, releasees with property offenses seemed to have a higher risk of being rearrested across five of the six states. The risk for drug and public order offenders varied across the six states when compared with violent offenders. All these findings imply that the historical and cultural context of various states may play an important role in understanding the effects of these institutional and offense variables on the risk of recidivism.

In sum, our analyses suggest that the effects of different sentencing models on reoffending may be largely contingent on the implementation and operation of state programs. Differences observed may be attributable to various supervision approaches within the states, differing expertise of state parole boards, or differing crime categories that are legislatively mandated. Because each state has very different approaches to parole, when parole may occur, and for what crimes offenders are eligible for parole, different classes of criminals have different propensities for recidivism. Future research is needed to disentangle these effects.

Empirical research on these programs and their implementation is needed to better understand how determinate- and indeterminate-sentencing models work in different political, legal, social, and cultural environments. Such research is challenging because it requires the development of detailed measures of the implementation and operation of state programs in each phase of prison release and supervision. Future research will require more precise data collection and collaboration between researchers, practitioners, and policy makers. Although such research is more time consuming and demanding, it is necessary to resolve the long-standing debate regarding which sentencing model is more likely to reduce recidivism.

Footnotes

Notes

Author Biographies

Yan Zhang is an assistant professor in the College of Criminal Justice at Sam Houston State University. She holds a doctorate in criminal justice from Michigan State University. She specializes in quantitative research methods. Her research interests include the social ecological contexts of delinquency, crime and victimization, sentencing policies and outcomes, and spatial temporal interaction of crime.

Lening Zhang is professor in the Department of Behavioral Sciences and director of the Rural Center for Applied Social, Health and Behavioral Research, Saint Francis University. His current research focuses on comparative studies of criminal victimization in China and United States using data collected from a large-scale survey in the city of Tianjing, China.

Michael S. Vaughn is a professor, assistant dean for graduate studies, and director of the doctoral program in the College of Criminal Justice at Sam Houston State University. At SHSU, he serves as the codirector of the Institute for Legal Studies in Criminal Justice. His research interests are primarily in the areas of legal issues in criminal justice and crime and social control.

References

Bales

W. D.

Bedard

L. E.

Quinn

S. T.

Ensley

D. T.

Holley

G. P.

(2005). Recidivism of public and private prison inmates in Florida. Criminology & Public Policy, 4, 57-82.

Bhati

A. S.

Piquero

A. R.

(2008). Estimating the impact of incarceration on subsequent offending trajectories: Deterrent, criminogenic, or null effect? Journal of Criminal Law & Criminology, 98, 207-253.

Brewer

Beckett

Holt

(1981). Determinate sentencing in California: The first year’s experience. Journal of Research in Crime & Delinquency, 18, 200-231.

Brody

(1976). The effectiveness of sentencing. HMSO Study No. 35. London: HMSO.

Bureau of Justice Assistance. (1998). 1996 national survey of state sentencing structures. Washington, DC: Author.

Bureau of Justice Statistics. (2004). Reentry trends in the U.S.: Releases from state prison. Retrieved April 15, 2008, from http://www.ojp.usdoj.gov/bjs/reentry/releases.htm

Cullen

F. T.

Gilbert

K. E.

(1982). Reaffirming rehabilitation. Cincinnati: Anderson.

Cullen

F. T.

(2005). The twelve people who saved rehabilitation: How the science of criminology made a difference. Criminology, 43, 1-42.

Dejong

(1997). Survival analysis and specific deterrence: Integrating theoretical and empirical models of recidivism. Criminology, 35, 561-575.

10.

Empey

L. T.

(1976). The effectiveness of correctional treatment: A survey of treatment evaluation studies. Contemporary Sociology, 5, 582-583.

11.

Fogel

(1975). We are the living proof: The justice model for corrections. Cincinnati: Anderson.

12.

Frankel

(1972). Criminal sentences: Law without order. New York: Hill and Wang.

13.

Frase

R. S.

(1993). Implementing commission-based sentencing guidelines: The lessons of the first ten years in Minnesota. Cornell Journal of Law & Public Policy, 2, 278-337.

14.

French

S. A.

Gendreau

(2006). Reducing prison misconduct: What works! Criminal Justice & Behavior, 33, 185-218.

15.

Garrett

(1985). Effects of residential treatment on adjudicated delinquents: A meta-analysis. Journal of Research in Crime & Delinquency, 22, 287-308.

16.

Gendreau

Ross

R. R.

(1987). Revivification of rehabilitation: Evidence from the 1980s. Justice Quarterly, 4, 349-407.

17.

Greenwood

P. W.

Abrahamse

A. F.

(1982). Selective incapacitation. Santa Monica: Rand.

18.

Hollin

C. R.

(2000). To treat or not to treat? A historical perspective. In Hollin

C. R.

(Ed.), Handbook of offender assessment and treatment (pp. 3-15). Chichester, UK: Wiley.

19.

Ireland

Prause

(2005). Discretionary parole release: Length of imprisonment, percent of sentence served and recidivism. Journal of Crime & Justice, 28, 27-49.

20.

Kane

R. L.

(2006). Understanding health care outcomes research. Boston: Jones & Bartlett.

21.

Keegan

Solomon

A. L.

(2004). Prisoner reentry in Virginia. Washington, DC: Urban Institute.

22.

Langan

Levin

(2002). Recidivism of prisoners released in 1994. Bureau of Justice Statistics special report. Washington, DC: Bureau of Justice Statistics.

23.

La Vigne

N. G.

Kachnowski

Travis

Naser

Visher

(2003). A portrait of prisoner reentry in Maryland. Washington, DC: Urban Institute.

24.

Lipsey

Cullen

F. T.

(2007). The effectiveness of correctional rehabilitation: A review of systematic reviews. Annual Review of Law & Social Science, 3, 297-320.

25.

Lipton

D. S.

Martinson

Wilks

(1975). The effectiveness of correctional treatment: A survey of treatment evaluation studies. New York: Praeger.

26.

Losel

(1995). Increasing consensus in the evaluation of offender rehabilitation: Lessons from research synthesis. Psychology, Crime, & Law, 2, 19-39.

27.

MacKenzie

D. L.

(2001). Evidence-based corrections: Identifying what works. Crime & Delinquency, 4, 457-471.

28.

MacKenzie

D. L.

(2006). What works in corrections? Reducing the criminal activities of offenders and delinquents. New York: Cambridge University Press.

29.

Martinson

(1974). What works?—Questions and answers about prison reform. The Public Interest, 35, 22-54.

30.

Martinson

(1979). New findings, new views: A note of caution regarding sentencing reform. Hofstra Law Review, 7, 243-258.

31.

Marvell

T. B.

Moody

C. E.

(1991). Ultimate impacts of sentencing reforms and speedy trial laws. Williamsburg, VA: Justec Research.

32.

Marvell

T. B.

Moody

C. E.

(1996). Determinate sentencing and abolishing parole: The long-term impacts on prisons and crime. Criminology, 34, 107-120.

33.

Maryland State Commission on Criminal Sentencing Policy. (2009). Sentencing guidelines overview. Retrieved on May 26, 2009, from http://www.msccsp.org/Guidelines/Overview.aspx

34.

McGuire

(2000, July 31–August 1). What works in reducing criminality. Paper presented at the Reducing Criminality: Partnerships and Best Practice conference, Perth, Australia. Canberra, Australia: Australian Institute of Criminology. Retrieved May 26, 2009, from http://www.aic.gov.au/conferences/criminality/mcguire.pdf

35.

Miller

(2008). Getting played: African American girls, urban inequality, and gendered violence. New York: New York University Press.

36.

Morris

(1974). The future of imprisonment. Chicago: University of Chicago Press.

37.

North Carolina Department of Correction. (2008). Post-release supervision and parole commission: Historical perspective. Retrieved September 30, 2008, from http://www.doc.state.nc.us/parole/history.htm

38.

Oregon Board of Parole and Post-Prison Supervision. (2007). About us. Retrieved September 30, 2008, from http://www.oregon.gov/BOPPPS/about_us.shtml

39.

Palmer

(1975). Martinson revisited. Journal of Research in Crime & Delinquency, 12, 133-152.

40.

Palmer

(1976). Abstract of “Martinson revisited.” Crime & Delinquency, 13, 178-179.

41.

Petersilia

Greenwood

P.W.

Lavin

(1977). Criminal careers of habitual offenders. Santa Monica: Rand.

42.

Reitz

K. R.

Reitz

C. R.

(1993). The American Bar Association’s new sentencing standards. Federal Sentencing Reporter, 6, 169-173.

43.

Rosenfeld

Wallman

Fornango

(2005). The contribution of ex-prisoners to crime rates. In Travis

Visher

(Eds.), Prisoner reentry and crime in America (pp. 80-104). New York: Cambridge University Press.

44.

Solomon

A. L.

Kachnowski

Bhati

(2005). Does parole work? Analyzing the impact of post prison supervision on re-arrest outcomes. Washington, DC: Urban Institute. Retrieved April 16, 2009, from http://www.urban.org/url.cfm?ID=311156

45.

Spohn

Holleran

(2002). The effect of imprisonment on recidivism rates of felony offenders: A focus on drug offenders. Criminology, 40, 329-357.

46.

Stemen

D.D.

Rengifo

Wilson

(2005). Of fragmentation and ferment: The impact of state sentencing policies on incarceration rates, 1975-2002. New York: Vera Institute.

47.

Tonry

(1999). The fragmentation of sentencing and corrections in America. Washington, DC: National Institute of Justice.

48.

Tonry

(2000). Reconsidering indeterminate and structured sentencing. Alternatives to Incarceration, 6, 17-21.

49.

Travis

Lawrence

(2002). Beyond the prison gates: The state of parole in America. Washington, DC: Urban Institute.

50.

Travis

L. F.

O’Leary

(1979). Changes in sentencing and parole decision-making: 1976-1978. Hackensack, NJ: NCCD.

51.

U.S. Department of Justice, Bureau of Justice Statistics. (2002). Recidivism of prisoners released in 1994: United States. Ann Arbor: ICPSR.

52.

Watson

Solomon

A. L.

La Vigne

N. G.

Travis

Funches

Parthasarathy

(2004). A portrait of prisoner reentry in Texas. Washington, DC: Urban Institute.