Parole, Desistance from Crime, and Community Integration (2008)

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.

2 Dimensions of Desistance CONCEPT AND DEFINITIONS Zero Offending or Less Offending? “Desistance” from crime is a common term in criminal justice re- search. Most offenders, after all, eventually stop offending. Yet there is relatively little theoretical conceptualization about crime cessation, the various reasons for desistance, or the mechanisms underlying it. Conceptually, there are several fundamental questions about desistance. For example, can desistance occur after one act of crime? If so, are the pro- cesses of desistance from a single act of crime different from those following several acts of crime? Is there such a thing as “spontaneous remission”? If so, can the term be precisely defined? One definition of spontaneous remission is desistance that occurs absent any external intervention (Stall and Biernacki, 1986). How can permanent desistance (the absence of acts of crime forever) be distinguished from temporary desistance (the absence of crimes for some amount of time)? Should a change from serious criminal acts to less serious ones be char- acterized as a type of desistance? In a similar vein, if criminal acts cease, but other problem behaviors (such as alcoholism or other drug use) persist or increase, what does it mean about the nature of desistance? All of these   This section is drawn from Laub and Sampson (2001) 19 

20 PAROLE, DESISTANCE FROM CRIME, AND COMMUNITY INTEGRATION issues raise critical questions about the meaning of desistance, and, conse- quently, about how to measure it or determine whether it is occurring. Whether using official records or self-reports, it is clear that how one defines desistance affects the measures and determinations about who desists and the relative size of the desisting population (for discussion of these issues, respectively, see Bushway et al., 2003; Brame et al., 2003). Kazemian (2007) reviews various operational definitions in past studies of desistance, which include convictions at age 21 but not between ages 21 and 32 (Farrington and Hawkins, 1991), the absence of a new officially recorded offense or probation violation in a 2-year period (Kruttschnitt et al., 2000) and the absence of self-reported illegal earnings during a 1-year period (Pezzin, 1995; see also Kazemian, 2007: Table 1, for more details from additional studies). Kazemian (2007, p. 8) concludes that the “sub- stantial degree of variability in the conceptualization of desistance . . . has led to disparate results regarding the causes and correlates of desistance from crime.” The issue of reintegration in society is considered as part of desistance in this report because the committee believes it may be a necessary condi- tion for the maintenance of desistance. A recent National Research Council (2005) report notes that there is little standardization of how outcomes such as desistance or recidivism should be measured in the evaluation of programs. Moreover, the findings on desistance from crime as a result of informal social controls come from longitudinal research, not program evaluation research. Empirical work is needed to examine how different definitions of desistance, as well as different research approaches, affect research outcomes. There is currently no agreed-on definition of desistance, but there is a growing consensus among researchers that it is best defined as a process, not an event, in which the frequency of crimes decelerates and exhibits less variety (see Bushway et al., 2001; Laub and Sampson, 2003; Maruna, 2001; Uggen and Massoglia, 2003; Weitekamp and Kerner, 1994; Loeber and LeBlanc, 1990; LeBlanc and Fréchette, 1989). For example, Maruna (2001, p. 17) defines desistance not as an event that happens, but, rather, as “the sustained absence of a certain type of event (in this case, crime).” A National Research Council report (1986) noted that care must be taken not to erroneously attribute the absence of further crime events near the end of an observation period or at the end of a specific age to (career) desistance rather than to the random time between events. Improved mea- sures of the permanent absence of offending, which remains the clearest definition of desistance from crime, are needed. Laub and Sampson (2001, 2003) have argued that it is important to analytically distinguish termination of offending from the concept of desis- tance: “termination” refers to the time at which one stops criminal activity, 

DIMENSIONS OF DESISTANCE 21 while “desistance” is the causal process that supports the termination of offending. While it is difficult to ascertain when the process of desistance actually begins, it is apparent that it continues after the termination of of- fending. In their view, the process of desistance maintains the continued state of nonoffending. More Prosocial Outcomes In addition to the cessation or reduction of criminal activities, the concept of desistance as a process generally also encompasses positive outcomes in terms of individuals’ behavior and integration into society. “[The] successful establishment of bonds with conventional others and participation in conventional activities are major contingencies on the path that leads to termination of a criminal career” (Shover, 1996, p. 126). More recently, Uggen and Massoglia (2003, p. 317) have argued that “desistance is a process characterized by particular behavioral states or markers” that is marked by the assumption of “adult occupational and family roles” (2003, p. 317). Along similar lines, Maruna (2001, p. 7) has contended that de- sistance is only possible when ex-offenders “develop a coherent, prosocial identity for themselves.” Thus, desistance is also generally viewed in terms of social integration or reintegration. Marriage Family and work seem to be especially important in the desistance process. The association of marriage with lower crime among men has been widely reported in both quantitative and qualitative studies (Farrington and West, 1995; Horney et al., 1995; Irwin, 1970; Laub and Sampson, 2003, Maume et al., 2005; Sampson and Laub, 1993; Shover, 1996; Warr, 1998; for a general overview, see Laub and Sampson, 2001). Marriage, especially strong marital attachment, has thus been identified as a significant factor in desistance for men. Recent research has extended this finding to women (King et al., 2007), but the researchers find the effects for marriage are less robust for women than they are for men. The King et al. study is an important one because it uses propensity score matching to estimate the causal effects of marriage. They use a contemporary dataset (The National Youth Survey), and they extend the analysis of marriage and crime to men and women. They find marriage reduces offending for males, especially for those men with a low propensity to marry. They find that marriage reduces offending for females, but only for those with a moderate propensity to marry. A change in criminal behavior may not necessarily result from marriage alone; rather, a change may occur in response to an enduring attachment 

22 PAROLE, DESISTANCE FROM CRIME, AND COMMUNITY INTEGRATION that emerges from entering into a good marriage. From this perspective, the growth of social bonds is like an investment process (Laub and Sampson, 1993, pp. 310-311; Nagin and Paternoster, 1994, pp. 586-588). As the in- vestment in social bonds grows, the incentive for avoiding crime increases because more is at stake. Empirical support for the idea of marriage as an investment process comes from research findings that early marriages char- acterized by social cohesiveness led to a growing preventive effect (Laub et al., 1998). Marriage also influences desistance because it often leads to significant changes in everyday routines. It is well known that life-styles and routine activities are a major source of variation in exposure to crime and victim- ization (Hindelang et al., 1978; Cohen and Felson, 1979). For example, participation in unstructured socializing activities with peers increases the frequency of deviant behaviors among those ages 18 to 26 (Osgood et al., 1996). Marriage has the potential to radically alter routine activities, es- pecially with regard to one’s peer group. As Osgood and Lee (1993) have argued, marriage entails obligations that tend to reduce leisure activities outside of the family. It is reasonable to assume that married people will spend more time together than with their same-sex peers. There is sup- porting empirical evidence that the transition to marriage is followed by a decline in time spent with friends and with exposure to delinquent peers (Warr, 1998, p. 183). Marriage, therefore, has the potential to cut off an ex-offender from peers at risk of re-offending. A recent study has addressed the issue of the causal direction of the relationship between marriage and desistance from crime. That study (Sampson et al., 2006) adopted a counterfactual life-course approach using yearly data from a sample of a group of 500 high-risk boys followed pro- spectively from adolescence to age 32 and a subsample of 52 men followed to age 70. The researchers found that being married is associated with an average reduction of approximately 35 percent in the odds of a criminal act for the married men in comparison with the nonmarried men (see also King et al., 2007). The researchers argue that marriage has a “knifing off” effect in the lives of disadvantaged men. They theorize that marriage may offer the op- portunity for new relationships and social supports; may change the routine activities engaged in by these men, especially unstructured time with peers; may create a more structured set of activities that are in many ways “su- pervised” by the spouse; and may transform the identity of the offender in ways that allow a life of greater responsibility as a spouse and parent (Laub and Sampson, 2003). While these results are robust, supporting the inference that states of marriage causally inhibit crime over a person’s life, the data used in the study are not recent. Using more recent data from the 1970s, Piquero and 

DIMENSIONS OF DESISTANCE 23 his colleagues (2002) found that conventional marriages inhibited continu- ity in criminal activity generally for both whites and nonwhites. However, the researchers found that conventional marriage did not appear to sig- nificantly inhibit violent arrests among nonwhites. They also found that common-law marriages increased crime (both violent and nonviolent) for nonwhites, but was insignificant for whites. Using data based on interviews with 658 newly convicted male offend- ers sentenced to the Nebraska Department of Correctional Services during a 9-month period in 1989-1990, Horney et al. (1995) also found that while marriage reduces offending, cohabitation increases it (Horney et al., 1995, p. 659). This finding is important in light of the fact that by 2002, only 48 percent of black families were headed by married couples, down from 70 percent during the 1960s (Kinnon, 2003; Wilson, 2002), and that one factor influencing this decline is the rise of incarceration rates, among black men since 1980 (Cready, Fossett, and Kiecolt, 1997). It would be important, therefore, to study the effects of marriage on reentry outcomes among a contemporary population of released prisoners. Data on the effects of marriage on this reentering population are needed. Building on research about the positive effects of healthy marriages, the Administration for Children, Youth and Families (ACF) in the U.S. Depart- ment of Health and Human Services (DHHS) has designed and funded the Community Healthy Marriage Initiative (CHMI), a nationwide program that has been implemented in over 20 sites (U.S. Department of Health and Human Services, 2007). A national evaluation of the CHMI has been funded through 2010 to compare outcomes in the CHMI communities with outcomes in comparison communities that are well matched to the project sites. The study focuses on evaluation of community-level approaches to encourage changes in norms that increase support for healthy marriages. In 2006, with funds provided under the Deficit Reduction Act of 2005, an estimated $4 million, was made available for responsible fatherhood, marriage, and family strengthening grants for incarcerated fathers and their partners. The program supports services to promote or sustain healthy mar- riage primarily to unmarried couples or married couples where one parent is incarcerated or has been recently released from prison or jail or is on probation or parole. The Office of the Assistant Secretary for Planning and Evaluation (ASPE) in DHSS will support an evaluation of the performance of these grants. Technical assistance will be provided to grantees that need help in complying with evaluation requirements. Work Like marriage, strong ties to work can lead to desistance from crime. One study using longitudinal data found that job stability was strongly re- 

24 PAROLE, DESISTANCE FROM CRIME, AND COMMUNITY INTEGRATION lated to desistance from crime (Sampson and Laub, 1993). In a similar vein, a study using qualitative data showed that acquiring a satisfying job was an important contingency in the lives of men who desisted from crime (Shover, 1996). However, as in the case of marriage, the nature of the relationship between work and desistance is not known: it is possible that “selection effects” are affecting that link: some factor or factors that predispose per- sons to find and remain in stable employment may also predispose them to desist from crime. One of the most convincing attempts to counteract selection bias comes from an analysis of data from a national work experiment that drew par- ticipants from poor urban neighborhoods in nine U.S. cities. Overall, the people who were given jobs through the experiment showed no reduction in crime relative to those in a control group (Uggen, 2000). However, age significantly interacted with employment to affect the timing of illegal earn- ings and arrest. Participants aged 27 or older were more likely to desist when provided with even marginal employment. For younger participants, the experimental job treatment had no effect on desistance. The findings from Uggen (2000) are important because the experimental nature of the data addresses the selectivity issue that has plagued much of the research on desistance. By specifying event history models accounting for assign- ment to, eligibility for, and participation in the National Supported Work Demonstration Project, this study provides more refined estimates of the effects of work as a turning point in the lives of offenders. Unfortunately, data for this study were collected from 1975-1979. Given changes in labor markets since the 1970s and current employment prospects for uneducated whites and minorities, the finding that work is a turning point may also be outdated. More recently, Rossman and Roman (2003), in their evaluation of the Opportunities to Succeed (OPTS) Program, found that full-time employ- ment among releasees increased with greater interaction with case managers and with higher levels of participation in drug treatment programs and that an increase in levels of employment was a predictor of reductions in drug dealing, violent crime, and property crime. Bloom (2006) notes that there have been few rigorous studies of recent, employee-focused reentry models. A similar conclusion is reached by Visher, Winterfield, and Coggeshall (2005). New models of work programs for releasees, such as the Safer Foundation’s Ready4Work model that is also being tested by the U.S. Department of Labor, focus on providing coordi- nated services both before and after an inmate is released. Evaluations of such models show results that are only somewhat positive. 

DIMENSIONS OF DESISTANCE 25 DESISTANCE AND PAROLE For individual offenders, the major functions of imprisonment are ret- ribution, removing dangerous people from the community, rehabilitation, and specific deterrence of future offending after release. The goals of parole supervision are similar and can be broadly conceived as “service” and “sur- veillance”: the rehabilitation of releasees and facilitation of their reentry in a community and the deterrence of crime and crime-related behaviors. In this framework, parole then serves as a sorting function by identifying releasees who fail quickly and returning them to prison in order to protect communities and by making services available to those who might most benefit from them.   Successful reintegration of offenders protects the com- munities in which they reside. Thus, the purposes of parole supervision are more utilitarian, integrative, and rehabilitative than the purposes of incarceration; retribution recedes into the background. Understanding the possible links between desistance from crime and community reintegration has important implications for designing cor- rectional policies, especially parole. One strategy is the greater use of com- munity-based corrections in order to assist offenders in their reintegration in a community. According to a life-course theory of age-graded informal social control (Sampson and Laub, 1993), effective sanctions should include methods of building social bonds to family, employment, and the commu- nity when offenders are released from prison. That is, they would combine treatment and services that promote connections with the community with appropriate emphasis on surveillance. Emphasizing the importance of treatment and services for the reduction of criminal behavior does not mean that community safety need be compro- mised. Community corrections strategies can provide effective surveillance and control of offenders to maintain the safety of a community while avoid- ing the placement of offenders in the potentially criminogenic environment of prison, an example of which is found in a recent study by Bhati (2007). For a sample of prisoners released from state prisons in 1994, Bhati used preincarceration arrest patterns to construct anticipated postrelease of- fending trajectories. A comparison of these counterfactuals with the actual postrelease offending patterns suggests that about 4 percent more returned to offending than was projected (i.e., they had a criminogenic experience in prison). Overall, 40 percent were deterred from future offending (i.e., were not rearrested after release), while the majority (56%), were merely incapacitated, that is, neither deterred nor rehabilitated by incarceration since they were found to revert back to anticipated offending patterns. Program components aimed at improving informal social controls and providing social support may reduce such criminal behavior, thus reduc- ing the need for future incarceration or surveillance of releasees (see Laub 

26 PAROLE, DESISTANCE FROM CRIME, AND COMMUNITY INTEGRATION and Allen, 1999). Moreover, community sanctions can promote offender accountability to their victims and to the community at large. In a paper presented at the workshop, Taxman (2006) notes that recent literature in the field has highlighted the way that natural support systems and informal social controls can augment and enhance the formal social controls imposed by correctional agencies. Informal social controls may encourage people to commit to a prosocial life-style and to develop new networks that include law-abiding citizens, not only through employment and families, but also through other types of community engagement, such as participating in the work of neighborhood associations or volunteer work (see also Uggen and Manza, 2004). In the long run, if involvement in serious crime and delinquency au- tomatically curtails future opportunities, releasees will have fewer incen- tives and options to desist and lead conventional lives. Many new laws in effect have permanently disenfranchised ex-offenders from employment opportunities, housing assistance, welfare benefits, and political rights (see Petersilia, 2003; Travis, 2005; Manza and Uggen, 2006). Such policies serve as barriers for ex-offenders to reconnect to the institutions that have been demonstrated as important in the desistance process, especially work and family. Thus, current policies regarding ex-offenders may produce unin- tended criminogenic effects by further damaging offenders’ already weak social bonds and cutting them off from promising avenues for desistance and reintegration into communities. WHO OFFENDS? WHO DESISTS? There are multiple pathways and factors involved in desistance from crime, including marriage and work, as noted above. Transformation of personal identify—that is, cognitive change as a precursor to behavioral change—has also been documented (Maruna, 2001; Giordano et al., 2002). Reduced exposure to delinquent peers fosters desistance from crime for youthful offenders (Warr, 1998). Perhaps the most obvious and simplest pathway to desistance from crime is aging: offending declines with age for all offenses (Glueck and Glueck, 1974; Gottfredson and Hirschi, 1990; Laub and Sampson, 2003). However, in spite of the evidence to date, inter- ventions designed to help men and women desist from criminal behavior have been slow to target these factors, with the exception of those related to employment—job readiness, training, and placement programs. Such programs are easier to implement than programs concerned with marriage and family and peer associations. One factor that appears to increase desistance from crime is reduced consumption of illegal drugs. The Federal Bureau of Prisons Office of Research and Evaluation (2000) evaluated its residential drug treatment 

DIMENSIONS OF DESISTANCE 27 program, which includes a transitional component that keeps former pris- oners engaged in treatment as they return to their home communities.  A rigorous research design included three methodologies to account for selec- tion bias. More than 2,000 individuals were included in the research.  After 3 years, treatment subjects were significantly less likely to be rearrested or have their parole revoked, than the control subjects (52% and 44%) and less likely to use drugs (58% and 50%). These results mean reduction in recidivism of about 16 percent. Arrests for all offenders also showed dif- ferences. Employment rates showed no differences. Moreover, in a 12-year follow-up of a sample of cocaine-dependent releasees, Hser and colleagues (2006) found that men who were continually abstinent for at least 5 years reported less past year involvement in crime, unemployment, and abuse of other substances than those who continued to use cocaine. Other factors have also been proposed as important in the desistance process. With regard to education, Locher and Moretti (2004) found that education decreases arrest and incarceration, based on prisoner, arrest, and self-report data. Other factors for which there is little or mixed evidence include residential change, religion, criminal justice sanctions, criminal jus- tice supervision (probation and parole), and a wide range of correctional and community interventions. THE PROCESS OF INDIVIDUAL CHANGE There is remarkable heterogeneity in criminal offending. Some offend- ers have short careers in one or a variety of crimes; other offenders have very long careers. A few specialize in one type of criminal activity, while most appear to commit a variety of crimes. From a theoretical perspec- tive, rather than thinking in the simplistic 0-1 categories of offenders and nonoffenders, it is useful to view criminality as following a path consisting of one or more crime and noncrime cycles (Glaser, 1969). Research on ex- offenders has shown that men typically do not commit crimes continually for long periods of time: instead, they “follow a zig-zag path . . . going from non crime to crime and to non crime again. Sometimes this sequence is repeated many times, but sometimes they clearly go to crime only once; sometimes these shifts are for long duration or even permanent, and some- times they are short lived” (Glaser, 1969, p. 58; see also Piquero, 2004). Processes of desistance have emerged that are common across a variety of problem behavior areas, including crime (Fagan, 1989). First, the deci- sion to stop appears to be preceded by a variety of negative consequences, both formal and informal, such as a prison sentence or the loss of a job. Second, multiple processes appear to be involved in sustaining and reinforc- ing the decision to change. Research on alcoholism, smoking, and obesity show commonalities in a process of three basic stages of behavior change: 

28 PAROLE, DESISTANCE FROM CRIME, AND COMMUNITY INTEGRATION motivation and commitment, initial behavior change, and maintenance of change (Brownell et al., 1986). Given this pattern, a realistic goal for ex-offenders, especially for high-rate offenders released from prison, is not zero offending, but reduced offending (reduced in terms of frequency and seriousness) and increased lengths of nonoffending periods. Empirical research on desistance has consistently demonstrated that this goal can be achieved. There appears to be an important distinction between lapses (slips) and relapse, and much could be learned about the processes of change if more were known about which slips lead to relapses and which do not (Brownell et al., 1986). There is some evidence to suggest that the determinants of lapses are different from the determinants of relapses. For instance, lapses are more commonly associated with situational factors, and relapses are related to individual factors such as negative emotional states or stress events (Brownell et al., 1986). It would also be valuable to know more about the timing of lapses in the change process, how this process applies to desistance from crime, and how it operates in a context of severe offi- cial sanctions, such as reincarceration, which typically are not part of the dynamics of behavioral change for addictive disorders. What is most important from this perspective is that the goal of desis- tance programs is not necessarily zero offending, but less offending and less serious offending. Less crime does not mean no crime: it is important for policy makers and program administrators to have realistic goals and to have forms of punishments and rewards available that will support those realistic goals. DESISTANCE PROGRAMS Kinds of Analysis Programs to encourage desistance from crime have both advantages and disadvantages, including budgetary costs. A cost-benefit analysis is one standard way to assess programs. In doing so, however, one must keep in mind that the benefits of a desistance program are often not confined to reductions in criminality. Similarly, post-incarceration programs and poli- cies not specifically aimed at reducing recidivism—such as those for drug treatment, mental health care, adult education, job training, and measures to increase the effective market wages for low-skilled labor—may result in desistance as a benefit. For example, with regard to substance abuse, re-   For an extensive review, see Laub and Sampson (2001) and citations therein. In addition, see Laub and Sampson (2003), Sampson and Laub (2003), Sampson, Laub, and Wimer (2006), and Ezell and Cohen (2005). 

DIMENSIONS OF DESISTANCE 29 search has found that nonusers are less likely to return to prison (Wexler et al., 1999). However a program is labeled and by whatever agency it is run, good policy analysis considers all of its benefits (advantages) and compares them with all of its costs (disadvantages). Economists’ standard measure for assessing the nonfinancial benefits and costs of a program is “willingness to pay”: how much the beneficiaries of a program would be willing to pay from their own resources for the advantage it generates for them, and how much the people disadvantaged by that program would have to be paid to compensate them for the dis- advantages (see Cook and Ludwig, 2000 and the references cited therein; Cohen et al., 2004; Nagin et al., 2006). For example, the merely financial losses to crime victims (e.g., the value of stolen property, medical expenses, lost wages) do not exhaust the social costs of crime and thus the benefits of crime control. Victims also experience pain and suffering, and even persons not yet victimized incur costs to avoid crime (e.g., by staying indoors or by moving) and suffer anxiety about the risk of crime. Moreover, the costs of crime-avoidance behaviors are not restricted to those who engage in them: when a drugstore restricts its hours or moves out of a neighborhood because of the owner’s fear of crime, residents of that neighborhood lose both their convenience as shoppers and economic opportunity as potential employees. The relevant question, from a policymaker’s perspective, is what a reduction in the risk of victimization is worth to those who enjoy it: for example, how much more rent would a person be prepared to pay to live in a lower crime neighborhood, all else being equal? Different programs compete for scarce resources: dollars, staff, mana- gerial and political attention, and even office space. In allocating funds to increase desistance (or for any other purpose), one can also ask which of the two or more programs makes the greatest contribution to the specific goal for each dollar or other resource unit. This analytic approach is known as cost effectiveness. These two styles of analysis—cost benefit and cost effectiveness—do not exhaust the full range of considerations that might go into the choice of public programs. For example, they exclude any consideration of justice or fairness. Nonetheless, they both provide useful guidance when the ap- propriate data are available and can serve as conceptual frameworks for thinking about programs, even when, as is often the case, those data are not available. For example, current correctional budgets are highly skewed toward institutional corrections (prisons and jails) in comparison with community corrections (probation and parole). A cost-effectiveness analysis would ask how many crimes (weighted for severity) could be prevented by add- ing $1 million to an institutional corrections budget (enough to keep ap- proximately 40 prisoners for an additional year) in comparison with adding 

30 PAROLE, DESISTANCE FROM CRIME, AND COMMUNITY INTEGRATION that same $1 million for more intensive community supervision (enough to monitor 125 high-risk parolees in the community). To consider the example, suppose that the average prisoner would have, if released, five times the criminal activity rate (lambda) of the aver- age high-risk probationer and that doubling probation supervision would cut crime among high-risk probationers by 10 percent. If the lambda of the high-risk probationer is x, incarceration would prevent 40 (5x) crimes per year, or 200x crimes, while intensive probation supervision would prevent 0.1 (1,000x) crimes per year, or 100x crimes. Thus, with the stated assump- tions, prison is more cost-effective as crime control than probation. But, if intensifying probation would cut crime among high-risk probationers by 30 percent, intensive probation would be a more cost-effective use of that $1 million. It should be noted that this analysis can be done without monetiz- ing the costs of crime (the benefits of crime control). In contrast, if the decision facing a lawmaker is whether to add $1 mil- lion to the corrections budget, there is a different set of questions than those above. One would need to know how much crime that sum would prevent, the value of that reduction to potential victims, and the nonbudget costs and non-crime-reduction benefits associated with that expenditure. Assessing Programs In thinking about how to assess desistance programs, it is critically important to keep in mind the heterogeneity of the population of released offenders: from those who have committed one nonviolent offense to those who may have served more than one sentence in prison for multiple vio- lent offenses. Programs that may work for one kind of parolee may not work or even be appropriate to try for other groups. As but one example, Piquero and Pogarsky (2002) has argued that the threat of sanctions would be most ineffective on the two extremes of the offending spectrum, that is, those who either have extremely high self-control and those with very low self-control. Moreover, even for appropriate programs, poor program implementation is often a barrier to both program effectiveness and to program evaluation. It is useful to think of the two obstacles (heterogeneity and implemen- tation) in terms of the “demand” and “supply” for rehabilitation. Limited demand, that is resistance to change among parolees, limits what even the   By contrast with cost-benefit analysis, cost-effectiveness analysis necessarily ignores im- portant ancillary questions, such as the value to victims of having the perpetrators punished with incarceration rather than probation or the damage to prisoners and their families from incarceration. 

DIMENSIONS OF DESISTANCE 31 most carefully implemented program can achieve, while limits on supply— the state-of-the-art in providing rehabilitative services—puts a cap on what can be achieved even by programs aimed at parolees amenable to change. On the demand side, an important and often ignored dimension of heterogeneity among released prisoners is first release. As noted in Chapter 1, inmates released from their first prison sentence have lower recidivism rates than those released from prison for the second, third, or fourth time. A recent study found that prisoners released for the first time accumulated 18-25 percent fewer arrests during the first 3 years out of prison than those who had been to prison and released at least once before—control- ling for sex, age, race, imprisonment offense, prior arrests, and time served (Rosenfeld et al., 2005). Over time, the fraction of released prisoners who are first-timers has declined, while the fraction who have are being released from their second or other incarceration has increased. This distinction is important (Tonry, 2004, p. 189): . . . most people sent to prison, and hence released, will be persistent of- fenders, and accordingly . . . on average people released from prison will present relatively high risks for recidivism. . . . Is it not odd . . . that in our time, we have forgotten the lower risk first-timers, and do not even think in our research to ask separately about them [emphasis added] By not distinguishing first-timers from the repeat offenders —who have shown that they are more difficult candidates for reentry—supervision and reentry programs may be doomed from the start. Approaches that work with the first-timers may not work with the repeaters and so may seriously affect assessment outcomes of desistance programs.