Abstract
Mass incarceration has led to increased interest in understanding the effects of imprisonment. Reviews of criminological theory and research report mixed evidence that incarceration reduces recidivism; indeed, some studies report criminogenic effects. We argue that a better understanding of the heterogeneity of incarceration—including the types and sequences of sanctions and experiences that occur before, during, and after imprisonment—and of incarceration effects among different groups is important for two reasons. First, it can assist with assessing the salience of prior research on the effects of incarceration on recidivism. Second, it serves to identify conceptual and methodological challenges that must be addressed to provide credible assessments of incarceration effects. The paper argues that incarceration likely exerts a variable effect depending on the nature of the prison experience, the counterfactual conditions, including prior sanction history, and the specific populations subject to imprisonment. Implications for theory, research, and policy are discussed.
Introduction
Due to the emergence of what has come to be called “mass incarceration,” scholars and policy makers increasingly have turned their attention to assessing the effects of the dramatic increase in prison populations over the past 30 years. One line of investigation has centered around the question of incapacitation and general deterrence. It has focused on whether increased incarceration has decreased crime rates and, if it has, whether the effect has resulted from these or other mechanisms (Durlauf & Nagin, 2011; Raphael & Stoll, 2009; Spelman, 2006). A parallel line of inquiry investigates the assumption that incarceration reduces recidivism. Although incarceration serves multiple goals, including retribution, reduced recidivism constitutes a central justification for its use and thus has garnered substantial attention (Cochran, Mears, & Bales, 2014; Loughran et al., 2009).
Notably, however, during an era in which policy makers have advocated for evidence-based policy (Welsh & Harris, 2008), credible research evidence that incarceration reduces recidivism is scant. Indeed, Nagin, Cullen, and Jonson’s (2009) review of prison recidivism studies concluded that the bulk of research to date has suffered from methodological limitations that substantially undermine the ability to draw firm conclusions. When their attention turned to experimental or quasi-experimental studies, they found that even with these more rigorous research designs, selection bias typically was not well-addressed. They also found that estimated effects were mixed, with some studies finding no effect of prison on recidivism, some finding that it increases recidivism, and others finding that it decreases it. This state of research is problematic because it leaves unanswered fundamental questions that criminological theory raises about the effects of incarceration (Bhati & Piquero, 2008; Nagin et al., 2009). It is problematic, too, because of the risks that scarce taxpayer dollars are put to inefficient use, opportunities to reduce victimization are missed, and misuse or overuse of incarceration actually may worsen rather than help public safety (Bales & Piquero, 2012; Cochran et al., 2014; Cullen, Jonson, & Nagin, 2011).
The goal of this article is to argue that, in fact, the state of research is substantially worse than what prior reviews suggest and, at the same time, both to contend that incarceration likely has variable effects and to identify specific directions along which research can shed light on incarceration effects on recidivism. A central departure point for these arguments stems, first, from the observation that incarceration constitutes a heterogeneous type of sanction. It constitutes not a single “treatment” but rather a range of possibilities that depend on prior sanctioning events and experiences in prison and after release. For each possibility, the counterfactual—what would have happened to a particular set of individuals had they not been incarcerated—may vary greatly (Cochran et al., 2014). A second point of departure flows from the observation that incarceration effects may vary by population. The effects may differ, for example, among demographic groups, various social groups (e.g., the mentally ill, high-risk inmates), and individuals who return to areas characterized by such factors as higher rates of violence, unemployment, and poverty or other forms of disadvantage.
These arguments are anticipated by several lines of scholarship, including studies of labeling effects (e.g., Palamara, Cullen, & Gersten, 1986) and contingent effects of formal sanctions (Baron, 2013), research on correctional intervention (Andrews & Bonta, 2006), and reentry studies (e.g., Maruna, 2001). More broadly, criminological theory, including shaming theory (Braithwaite, 1989) and defiance theory (Sherman, 1993), suggests warrant for anticipating that incarceration may have differential or criminogenic effects for some groups. Drawing on these and other lines of scholarship, the article’s arguments highlight both the limits of prior research on the recidivism effects of incarceration and the avenues along which future research might productively be directed. It leads, as well, to the central insight that incarceration likely exerts no single, uniform effect but rather different effects depending on the counterfactual and the specific populations subject to incarceration or to other types of sanctions.
Background
Since the 1980s, the United States embarked on a historically unprecedented trajectory of prison growth, one paralleled by the increase in corrections more generally. From 1980 to 2011, the correctional system—including individuals on probation, in jails and prisons, and on parole—increased from 2 million to 7 million (Glaze & Parks, 2012, p. 1). Increases in jail and prison populations contributed substantially to this growth. During the same period of time, the number of inmates in state and federal correctional facilities increased from just over 500,000 to over 2.2 million (Glaze & Parks, 2012, p. 3). The Bureau of Justice Statistics estimates that, as of 2011, the latest year for which national data exist, over 690,000 individuals are released annually from state and federal prisons (Carson & Sabol, 2012, p. 1). Growth in corrections has been paralleled by a dramatic increase in expenditures as well. From 1982 to 2001, total state corrections expenditures rose from US$15 billion to US$54 billion in real dollars, and then in the subsequent decade generally remained in the range of US$48 billion to US$53 billion (Kyckelhahn, 2012, p. 1). Three fourths of corrections budgets go to correctional institutions (p. 2); a clear driver of corrections costs thus stems from investment in jails and prisons.
Substantial scholarship has focused on explaining why criminal justice policy in America became more punitive during and after the 1980s. Increases in violent crime, especially homicide, sparked some of the change, but a wide range of other factors have been implicated as well, including the War on Drugs, racialized depictions of crime, and a transition into a more politically conservative era, one that viewed social disorder as requiring a reduced emphasis on social welfare and a greater increase on formal modes of social control (see, generally, Garland, 2001; Gottschalk, 2006, 2011; Tonry, 2004; Western, 2006). Whatever the precise constellation of causes, the end result—facilitated in part through tougher sentencing laws and the expansion of law enforcement departments—was a dramatic increase in the correctional system, with substantial financial investment directed in particular to expanding prison capacity.
What, then, is the state of evidence concerning the effects of incarceration on recidivism? One widely cited national study of recidivism found that over two thirds of state prisoners are rearrested within 3 years of release (Langan & Levin, 2002). These involved offenses that came to the awareness of law enforcement. In reality, more crimes likely occurred than resulted in arrests, and so it is safe to assume that the true rate of recidivism is higher. As Cullen et al. (2011) have observed, this rate of recidivism is “inconsistent with prisons as a powerful specific deterrent” (p. 54S). Even so, the question is, “What would the recidivism rate be if some sanction, including any of a wide range of intermediate sanctions, other than prison were imposed?” By and large, studies present a varied picture, with some work finding beneficial effects, other work finding criminogenic effects, and still other work finding no effect.
The perhaps more critical finding is that the bulk of prior empirical research has suffered from methodological limitations that substantially reduce the credibility of the estimated effects. Indeed, Nagin et al. (2009) prioritized that finding over the others that emerged from their review. A consistent problem with many of the reviewed studies was their failure to include control variables or to use appropriate matching or quasi-experimental research designs that would reduce the likelihood that estimated effects were spurious. Nagin et al.’s review accorded with that of other systematic reviews (e.g., Gendreau, Goggin, Cullen, & Andrews, 2000; Jonson, 2010; Smith, Goggin, & Gendreau, 2002; Villettaz, Killias, & Zoder, 2006) in identifying that—given the caveat concerning the methodological rigor of recidivism research—most studies of incarceration, especially those with a greater degree of methodological rigor, find that it increases recidivism. That is, incarceration has a criminogenic effect. Recent studies reinforce this assessment (e.g., Bales & Piquero, 2012; Cochran et al., 2014). In addition, studies of imprisonment duration identify little to no effect on recidivism (Loughran et al., 2009; Meade, Steiner, Makarios, & Travis, 2013; Snodgrass, Blokland, Haviland, Nieuwbeerta, & Nagin, 2011).
These findings raise concern about the effectiveness of incarceration in reducing criminal behavior. At the same time, the reviews raise questions about methodological problems in extant research. In addition, there is considerable ambiguity about the appropriate basis of comparison for estimating incarceration effects and, in turn, for generalizing about them. There also is ambiguity about the extent to which incarceration effects can safely be assumed to be uniform, or constant, across different groups, populations, or settings. As we argue below, when viewed in this light, even less is known about the effectiveness of incarceration in reducing recidivism than reviews to date intimate. By extension, there exists an even greater need for systematic assessment of the effect of incarceration on recidivism to advance both theory and policy.
Counterfactuals and the Heterogeneity of Sanctions: Implications for Identifying the Effectiveness of Incarceration on Recidivism
The central starting point for assessing incarceration effects on recidivism involves identifying the counterfactual condition. That is, what would have happened to released prisoners if they had not been incarcerated? An under-recognized problem, however, is that the counterfactual may vary (Cochran et al., 2014). This variability, as we argue below, has considerable implications for assessing incarceration effects and, in turn, for accurately generalizing about them. An estimated effect of a sanction, such as imprisonment, is always specific to the nature of the imprisonment experience and what otherwise would have happened to the individuals in a given jurisdiction or state. However, many studies treat imprisonment as a generic or uniform event, one that does not vary in quantity or quality. They also typically assume that individuals would have received a generic mix of non-custodial sanctions. In reality, the nature of incarceration experiences can vary greatly, and the counterfactual scenario—the sanction that otherwise would have occurred—also can vary greatly. It can, for example, include any of a range of intermediate sanctions (Crouch, 1993; Moore, May, & Wood, 2008; Petersilia & Deschenes, 1994; Spelman, 1995). Put differently, the counterfactual for prisoners is not “no sanction” but rather is any of a range of other possible sanctions (Cullen & Jonson, 2012).
Three dimensions can be used to define or characterize incarceration as “treatment” and, in turn, to aid in identifying the counterfactual condition and thus appropriate comparison group. First, incarcerated individuals previously may have experienced a diverse range of possible sanctions. Second, in-prison experiences may vary and so result in specific types of “treatment,” ones that may exert rehabilitative, deterrent, or criminogenic effects. Third, incarceration may vary with respect to post-release supervision. Some inmates may be released outright, with no supervision; others may be released to a short period of supervision; still others may be released to longer periods of supervision or to more intense supervision.
Below, we describe each of these three sources of heterogeneity and do so to extend arguments in prior research that the issue of sanction heterogeneity is far from an academic or peripheral one (e.g., Cochran et al., 2014; Moore et al., 2008; Nagin et al., 2009; Palamara et al., 1986; Spelman, 1995; Villettaz et al., 2006). Rather, it directly affects research design and generalizability. As we discuss, it directly affects, too, inferences that might be drawn about questions of relevance to criminological theory. For example, an incarceration experience devoid of rehabilitation that nonetheless reduces recidivism by 20% presumptively provides some support for a specific deterrence. Conversely, short incarceration stays at facilities where rehabilitation efforts—such as programs to improve self-control and develop prosocial coping strategies—feature prominently and demonstrate a 20% reduction in recidivism would provide evidence in support of self-control or strain theories.
Heterogeneity in Prior Sanction Experiences
One of the departing points for highlighting the conceptual issues involved in estimating incarceration effects centers around the fact that the courts typically do not ignore the background of criminals. A convicted felon’s prior sanctioning experiences—such as failure on probation or repeated incarceration—thus define the counterfactual conditions. In so doing, these experiences determine what group of individuals should serve as the point of comparison for estimating the effect of prison. For one group, the sole alternative might be jail. In this case, the counterfactual consists of prison effects relative to what would have happened if the individuals were sentenced to jail. For another, probation, intensive probation, or jail might constitute sanctions that the court would consider. Here, then, several counterfactuals exist. The question is the same: “What is the effect of prison relative to what would have happened?” But “what would have happened” consists of three distinct possibilities. Accordingly, a research design would want to identify three different matched groups for comparison.
It is simpler, in some ways, to view the issue from that of a judge’s perspective. The judge would like to know, given the particular offender and his or her prior sanctioning experiences, which type of sanction now would be most effective. Was the individual ever on probation? Did he or she ever serve time in jail or prison? Does he or she have an extensive record of probation placements but no previous stints in jail or prison? Each scenario may incline the judge to support, or not, a term of imprisonment. Prosecutors frequently determine the sanction that is provided, but they may follow a similar logic in their thinking. Of course, it is possible that for some crimes, the prior sanction history is irrelevant. But in many cases, the history of prior sanctions may matter considerably. Consider, for example, juvenile court. National guidelines for juvenile court sentencing call for progressive sanctions, with youth “working” their way up a ladder of progressively more intensive sanctions that emphasize more strongly both punishment and rehabilitation (Howell, 2009). This logic applies to adult court as well, despite the fact that criminal justice tends to be more “offense-focused” rather than “offender-focused” (Feld, 1999).
From a judge’s perspective, the situation in fact is more complicated. The judge may face groups of individuals with varied sanction histories, and these histories may exert varying effects on the judge’s—or the prosecutor’s or court community’s—use of incarceration. For any given group, however, it would be unsatisfactory to read about results from a study of incarceration versus non-incarcerative sanctions. Why? For cases where the judge typically would use probation, the focus for the judge would be on a prison versus probation comparison (i.e., “Do I use prison in this case rather than probation?”). For cases where the judge typically would use intensive probation or some other specific type of intermediate sanction, the focus would be on the use of prison versus that specific sanction. And, not least, for cases where the judge typically would rely on a jail term, the focus would be on the use of prison versus jail. Results from these specific comparisons—for the types of individuals who would receive these specific sanctions and who would be candidates, in the judge’s eyes, for imprisonment—would be useful to the judge for determining whether incarceration constitutes the more effective sanction for reducing recidivism. Such an analysis, of course, may shed little to no light on the expected retribution or the extent to which courts weigh and balance retribution against recidivism (Durlauf & Nagin, 2011). Even so, they would more directly address a central public safety goal, that of reducing the likelihood that sanctioned offenders will recidivate.
To our knowledge, no extant empirical studies systematically investigate these different possibilities. Instead, they typically ignore prior sanction history or, at most, attempt to match on it or focus only on individuals who have never received a given type of sanction before (Nieuwbeerta, Nagin, & Blokland, 2009). The end result is that questions arise about the internal validity and external validity of the results (Mears, Cochran, Greenman, Bhati, & Greenwald, 2011). No simple solution exists save to carefully identify the precise counterfactual conditions involved in a given study. Notably, however, in so doing, opportunities arise for testing criminological theory. For example, consider sanction sequences and the fact that some individuals who have repeatedly been placed on probation eventually may be imprisoned. To the extent that we can identify similar individuals who were not imprisoned but instead continued to receive probation, we can investigate not just the effect of incarceration in this particular counterfactual condition but also the idea that punishment avoidance affects criminal behavior (Stafford & Warr, 1993). In this example, punishment avoidance does not mean “no sanction,” but rather refers to receiving probation when an individual otherwise might be sent to prison. As we discuss further below, the latter focus entails a different approach to estimating sanction effects—instead of viewing prison as the treatment, continued use of probation is viewed as the treatment.
Heterogeneity of In-Prison Experiences
In defining counterfactual conditions, the precise “treatment” needs to be specified; doing so is essential for defining treatment and, in turn, for generalizing results (Farrington, 2003). Unfortunately, most prior research treats incarceration as a binary event—either an individual is incarcerated or not. In reality, incarceration events can vary greatly and so, too, can the “treatment.” Put differently, referring to incarceration as treatment obscures the fact that in reality, there likely are many different incarceration “treatments” (Nagin et al., 2009).
Consider some of the dimensions that may characterize incarceration terms and, in turn, lead to the functional equivalent of different treatments (see, generally, Adams, 1992; Bottoms, 1999; Chen & Shapiro, 2007; Listwan, Sullivan, Agnew, Cullen, & Colvin, 2013; Rhodes, 2004). For example, some prison systems may offer considerably more and higher quality vocational and educational programming than others. They may offer more and higher quality pre-release programming. So, too, they may offer more and higher quality drug or mental health treatment. Some prison systems may be more proactive in reducing inmate victimization. They may have more professional staff. The prison culture may be more humane and more supportive of efforts to facilitate prosocial behavior during and after incarceration. They may be more “reintegrative” (Braithwaite, 1989), promote helpful rather than harmful labeling (Cullen & Jonson, 2014), or be perceived as more procedurally and substantively fair and so encourage prosocial conduct rather than instigate criminal behavior (Sherman, 1993). These and other dimensions define “the” incarceration experience and in turn the nature of the “treatment” captured in studies of incarceration effects.
Put differently, all these sources of variation create the equivalent of different treatments, much as variation in the frequency, intensity, duration, and type of exercise essentially create different exercise programs. Combining all exercise programs into one category would provide little credible basis for assessing the effectiveness of exercise. Similarly, combining all incarceration experiences into one category provides little credible basis for estimating prison effects on recidivism. Studies certainly exist that examine some aspects of the prison experience, such as the length of stay in prison (Loughran et al., 2009) or the experience of prison-induced strain (Listwan et al., 2013). Such studies are, however, the exception and do not substitute for systematic empirical assessment of the diverse types of incarceration “treatments” that exist in prison systems in the United States and worldwide (Mears, 2012).
The salience of addressing variation in the prison experience does not center only around the need to define an intervention carefully for purposes of clarifying the external validity of a study (Farrington, 2003). Rather, it centers as well around the corresponding need to define the counterfactual. Specifically, among inmates who receive a given type of prison “treatment,” what would their experience have been if they had not been incarcerated? Here, we want to know what type of sanction the inmate otherwise would have received and ideally the quantity and quality of that experience. The estimated treatment effect then is specific to the particular type of incarceration “treatment” and to the particular counterfactual experience as well.
Heterogeneity of Post-Release Experiences
Incarceration-as-treatment may vary along a third dimension—the presence, intensity, and quality of post-release supervision (Piehl & LoBuglio, 2005; Tonry & Lynch, 1996). Put differently, supervision further serves to define the incarceration event or “treatment.” It introduces additional heterogeneity. Such heterogeneity in turn requires creation of appropriate counterfactuals and the attendant need to state precisely the setting and population to which observed effects apply. Incarceration with high-quality supervision is, for example, different from incarceration with no supervision. If the latter truly constitutes the counterfactual condition contemplated by a given state or court system, then it should serve as the basis for estimating incarceration effects. If, however, lengthier terms of incarceration would be the likely alternative, then this consideration serves as the foundation for identifying the appropriate counterfactual group. In this latter case, the comparison might focus on inmates who serve lengthier terms as compared with inmates who serve shorter terms (Loughran et al., 2009).
There are two distinct sets of counterfactual conditions involved when we focus on post-release supervision. The first set involves situations where the inmate on supervision is compared with individuals who received different, non-incarcerative sanctions. Here, the challenge lies in part in the fact that ex-prisoners on supervision may differ from those not on supervision. State sentencing policies have changed greatly in recent decades (Visher & Travis, 2011), and the result in part is that some inmates are released to supervision and some are not. But inmate characteristics and how inmates act in prison may influence sentence length and the likelihood or intensity of post-release supervision. Accordingly, identifying matches from individuals placed on probation or some other intermediate sanction is more complicated.
A second set involves situations in which some inmates receive post-release supervision and others do not. This type of study faces the challenge of identifying appropriate matches, but that endeavor likely is easier than that of identifying matches from the non-imprisoned population of felons. Even so, the question arises as to what groups constitute appropriate points of comparison. In addition, the added “treatment” itself varies. Some ex-prisoners may be placed on supervision for extended periods of time and be subjected to more restrictions compared with others (Maruna, 2001, 2011; Petersilia, 2005; Visher & Travis, 2003). The result is a series of incarceration treatments that vary with respect both to time served and other in-prison experiences and to post-release supervision experiences.
Variable Incarceration Effects for Different Social Groups and Settings
As discussed above, sanction effects are defined by specific counterfactual conditions, which in turn are defined by the sequence of prior sanction experiences, the nature of the incarceration experience, and the amount and intensity of post-release supervisions. A further complicating factor when assessing the effectiveness of prison in reducing recidivism is not just the variety of counterfactual conditions—many of which remain largely unexamined empirically—but also the likelihood that incarceration effects vary across social groups.
This possibility is suggested by many lines of scholarship in criminology. For example, more than 30 years ago, Ekland-Olson, Kelly, and Supancic (1983) argued that sanction severity may exert different effects on recidivism depending on the extent of disruption to an individual’s life (see also Listwan et al., 2013). In a related vein, Palamara et al. (1986), in an empirical study, examined how combinations of police and mental health intervention exerted different effects on different types of offenders. Specifically, youth who received both police and mental health interventions were more likely to engage in delinquency than youth who received one or the other. The study underscored the importance of specifying the nature of the sanction and also highlighted that intervention effects may vary depending on the characteristics of the referred youth. Similar to Ekland-Olson et al., the authors argued that intervention effects may be variable. More recently, Baron (2013) empirically identified the potential for the effects of formal sanctions to be contingent on such conditions as peer networks and social context.
Separately, Stafford and Warr (1993) have argued that punishment may exert differential effects on criminal behavior depending on whether an individual has direct or indirect experiences with punishment or, conversely, with punishment avoidance. Here, the idea in part is that a sanction’s effect is contingent not only on prior punishment experiences but also on prior experiences avoiding punishment. It is contingent, too, on whether the experiences were direct (i.e., experienced by the sanctioned individual) or indirect (i.e., experienced by individuals known by the sanctioned individual).
A broad literature on effective correctional interventions also suggests that sanctions may exert differential effects for different groups. Central to this body of work is the identification of general principles that effectively reduce recidivism (Andrews & Bonta, 2006; Lipsey & Cullen, 2007). This work also highlights that interventions—including rehabilitative programs and punishments of various types—may be more effective for some populations, such as high-risk offenders, than for others (Cullen & Jonson, 2014; MacKenzie, 2006).
The general logic underlying this avenue of research is that incarceration may exert differential effects on some social groups depending on such factors as their personality, social context, life experiences, and the costs that incarceration imposes. Notably, however, there remains little empirical research that systematically investigates a range of counterfactual conditions when estimating incarceration effects, as argued above, or, in turn, that systematically investigates whether estimated effects are greater or lesser for some social groups.
Examination of potential conditional effects of incarceration merits attention for at least two reasons. The first is that criminological theory suggests grounds for doing so and thus provides grounds for gaining insight into incarceration effects; conversely, a focus on incarceration effects provides an opportunity to test and develop theory (Agnew, 2005; Nagin et al., 2009). Several examples illustrate the point. The logic of self-control theory suggests that more severe sanctions should do little to reduce recidivism given the putative stability of self-control over the life-course (Gottfredson & Hirschi, 1990); studies identifying deterrent effects of incarceration on low self-control individuals thus would be of relevance to testing and modifying the theory. Strain theory can be viewed as arguing for criminogenic effects of incarceration, depending on the level of strain induced by incarceration (Listwan et al., 2013); to the extent that some individuals experience prison as strainful and that such strain then is amplified by the communities to which they return, prison should generate a differential effect. Identifying such effects would provide grounds for expanding the focus of strain theory, or the specification of strain theory, to encompass events (e.g., sanctions) and contexts (e.g., communities) in identifying individuals most at risk of offending (Agnew, 2005). Reintegrative shaming theory (Braithwaite, 1989) explicitly argues that sanctions may increase offending or decrease it depending on the nature of the sanctioning process, and defiance theory (Sherman, 1993), too, anticipates that the nature of this process, and the perceived legitimacy of sanctioning authorities and administered punishments, may result in variable effects on offending. Studies that specify precisely what types of sanctioning processes and incarceration experiences are “reintegrative” or that induce defiance thus not only are called for by these theories but also provide grounds for determining more precisely when and how a sanction of a given type produces a given change in offending. Similar possibilities can be identified for other criminological theories (see, generally, Baron, 2013; Nagin et al., 2009) and could help both to advance criminological understanding of punishment and to test and modify theories of offending.
A focus on conditional effects merits attention as well because of its importance for policy. If some groups may be more strongly deterred by prison, or if some are more likely to engage in offending because of prison, policy makers and the courts might well want to consider such information in crafting sentencing laws. Indeed, the very existence of a separate juvenile justice system is premised in part on the assumption that less serious punishment, including shorter terms of incarceration, may be more beneficial for youth as compared with adults (Feld, 1999; Howell, 2009). Whether different punishment for individuals that is based on the relative effectiveness of the particular sanctions in reducing recidivism is appropriate is a complicated issue (Ekland-Olson et al., 1983). Even so, the effectiveness of sanctions lies at the heart of any policy discussion. To ignore it, especially if a given sanction increases offending for one group and decreases it for another, would run counter to correctional system public safety goals.
There exist many social groups for which we might anticipate variable effects of incarceration. One line of demarcation involves demographic groups. Race and ethnicity continue to stand as prominent divides in American society and in the criminal justice system, with differences that can be identified culturally, economically, and socially (McPherson, Smith-Lovin, & Cook, 2001; Western, 2006). Similarly, criminological accounts of gender and offending suggest that considerable differences may exist in the factors that contribute to criminal behavior (Agnew, 2005; Gartner, 2011), though little remains known about gender differences in the effectiveness of correctional rehabilitation programs (Lipsey & Cullen, 2007) and sanctions (see, however, Mears, Cochran, & Bales, 2012). Youth inhabit significantly different social worlds than do adults, and such differences, along with developmental changes (Scott & Steinberg, 2008), also may well condition the effectiveness of incarceration or other sanctions.
Another line of demarcation consists of the personal backgrounds of individuals. Indeed, juvenile and criminal court sanctioning historically has emphasized the idea that an individual’s background or characteristics should influence sanctioning decisions. This view permeates the juvenile court, at least in historical accounts (Feld, 1999) but holds largely in contemporary times as well (Scott & Steinberg, 2008). Even if the court does not consider the characteristics of individuals, sanctions nonetheless may exert differential effects depending on these characteristics. Mentally ill inmates, for example, or inmates with a high risk of recidivism, may be differentially affected by incarceration as compared with their counterparts (Cullen & Gilbert, 2013; Morris & Piquero, 2013). Perhaps incarceration, too, exerts different effects for different types of offenders. For example, it may have a greater criminogenic effect on drug offenders as compared with other types of offenders (Spohn & Holleran, 2002).
Not least, there is the social context to which released prisoners return. Ex-prisoners may return to areas that vary greatly in their demographic composition, economic conditions, and levels and quality of informal or formal social control (Visher & Travis, 2003, 2011). Black inmates, for example, may return to areas marked by higher levels of social disadvantage and unemployment (Wang, Mears, & Bales, 2010). All else equal, inmates who return to areas where few employment opportunities exist presumptively could be anticipated to be at greater risk of recidivism (Kubrin & Stewart, 2006). The reentry process, too, may vary for individuals (Maruna, 2001, 2011), such that in some communities, greater support for and assistance with the transition back into society exists, whereas in others, it may be non-existent.
Other dimensions can be identified that could serve to identify groups for which incarceration may have stronger or weaker effects. Such effects may be stronger or weaker in the direction of increasing or decreasing recidivism, or in affecting other outcomes, such as housing, employment, mental health, and family functioning. The central point is that without knowledge about such possibilities, external validity in studies of incarceration effects will remain a considerable problem. Specifically, it will remain difficult to state with confidence what “the” effect of a given incarceration experience, as defined by the duration of incarceration and what occurs during it, likely is across different social groups and settings.
Counterfactuals, Varying Effects, and Evidence on Incarceration and Recidivism
The central argument of this article reduces to the following: When prior research is viewed through the above counterfactual framework, it becomes clear that prior research provides limited insight into the effects of incarceration. The three central limitations center on a failure (a) to examine and describe systematically a range of counterfactual conditions, (b) to examine the potentially differential effects of a given incarceration “treatment” among varied groups of ex-prisoners, and (c) to identify how counterfactual conditions may change over time or vary across place. Each limitation provides the foundation not only for assessing prior research and theory but also for anticipating that incarceration likely exerts variable effects and for guiding future research and theory.
Limitation 1: Vague or Unstated Counterfactual Conditions
The first limitation concerns the limited state of research documenting incarceration effects across a range of counterfactual scenarios. The bulk of prior research, for example, reviewed by Nagin et al. (2009; see also Cullen et al., 2011), does not systematically examine a wide range of counterfactual conditions or even state precisely what counterfactual condition is used to estimate an incarceration effect. Nagin et al. emphasized the limited quality of research on the effect of incarceration on recidivism. Few studies, for example, use rigorous quasi-experimental research designs. Alongside of that limitation, however, is the lack of systematic, rigorous research on the effects of incarceration under different counterfactual specifications.
Ideally, a study would describe the counterfactual in detail with respect to the three dimensions discussed above: (a) the prior sanction history of the population and what the typical sanctioning for it would be absent an incarcerative option, (b) the specific dimensions and characteristics of the in-prison experience that comprises the incarceration “treatment,” and (c) the type of post-release supervision, if any, that also comprises the “treatment.” Failure to specify the counterfactual condition in this way means that it is not only difficult to speak meaningfully about the treatment but also, in turn, difficult to know to what settings or groups the estimated effect applies. That assessment characterizes most evaluations to date of the effects of incarceration on recidivism (see Cochran et al., 2014).
Limitation 2: The Potential for Incarceration to Exert Differential Effects
The second limitation concerns the potential for incarceration—as well as any type of criminal justice program or intervention—to have a differential effect among groups (Andrews & Bonta, 2006; Lipsey & Cullen, 2007; MacKenzie, 2006; Morris & Piquero, 2013). As discussed above, prominent possibilities for variable effects include divides with respect to demographic groups (e.g., race, ethnicity, sex, age), social groups (e.g., abused, mentally ill, high risk), and groups that vary with respect to the types of areas to which they return (e.g., settings characterized by poverty or violence). To our knowledge, almost no research exists that systematically assesses the potentially differential effects of various types of incarceration (e.g., prison terms of varying durations and with varying types of in-prison experiences) on recidivism among these three different sets of groups. Some exceptions exist. For example, Mears et al. (2012) examined whether incarceration, jail, intensive probation, and probation exerted different effects on recidivism among males and females, respectively, and found largely similar effects; however, prison appeared more strongly to increase property recidivism among female inmates, especially when the comparison was with probation. Whether these sanctions, and variants of them, exert different effects for other groups—based on demographic characteristics, social or risk-level factors, or characteristics of communities to which they return or belong—remains largely unknown in sanction effectiveness research. Although not the focus of this article, it merits emphasizing that—notwithstanding progress in assessing the effect of rehabilitation on recidivism (Cullen & Gilbert, 2013)—similar gaps in research exist for understanding the effectiveness of prison programs for different populations.
Limitation 3: Changing or Differing Counterfactual Conditions
Another related limitation of research on incarceration effects bears discussion. To this point, our focus here has been on incarceration as “treatment” and the challenge and importance of accurately identifying the counterfactual condition. The logic revolves around the idea that incarceration stands as a “last resort” punishment. As such, the counterfactual is some lesser form of punishment. Viewed in this way, incarceration constitutes a “treatment,” and the question then is, “What is the effect of this treatment (incarceration, as a form of severe punishment) as compared to the type of treatment (e.g., intermediate sanction, as a form of less severe punishment) that we otherwise would have applied?”
However, the validity of framing the counterfactual this way hinges on assumptions about the default mode of sanctioning. For example, when states, after a period of “get tough” sanctioning, face budget crises that limit the ability to maintain or expand prison systems, the counterfactual may change. In particular, the default mode of sanctioning may be incarceration, not a less severe form of incarceration, for certain types of offenders. In this case, states may seek to use intermediate sanctions, such as electronic monitoring, as a cheaper but hopefully equally effective or more effective alternative to prison. Here, the “treatment” is an intermediate sanction, and the question is, “What is the effect of this treatment (an intermediate sanction) as compared to the type of treatment (incarceration) that we otherwise would have applied?”
The distinction is important and literally defines how a quasi-experimental research design should be undertaken and how incarceration effects should be interpreted (Cochran et al., 2014). In matching analyses, for example, incarceration-as-treatment would involve identifying matched subjects from an intermediate sanction population, such as individuals on probation. The estimated effect here would refer to the set of individuals on probation who would be viewed as appropriate for the incarceration “treatment.” The “treatment” effect thus would be the difference in recidivism produced by incarceration as compared with probation, but only for the subset of individuals on probation who would have been likely candidates for receiving the “treatment” (i.e., incarceration).
The research design would differ if the default, counterfactual condition were incarceration and the focus was on whether probation might constitute a more effective approach for some individuals who typically are sent to prison for their punishment. Here, the analyses would begin with the probation group and matches would be sought from the prison population. In this case, the “treatment” effect would be the difference in recidivism produced by probation as compared with incarceration. However, the effect would refer only to the reduction in recidivism among those individuals in prison who would have been likely candidates for the “treatment” (i.e., probation). Such considerations are critical to undertaking accurate assessments of sanction impacts but have not been systematically addressed in research on the effects of different types of sanctions on recidivism (see, however, Cochran et al., 2014; Nieuwbeerta et al., 2009; see also Bales & Piquero, 2012; Morris & Piquero, 2013; Villettaz et al., 2006).
The central challenge lies, again, in identifying accurately the default mode of sanctioning; treatment then constitutes the alternative to this “business-as-usual” condition. Identifying this condition may be complicated by the variation that can exist across jurisdictions in how they sanction. For example, in some counties, a prosecutor might rely on probation and treatment more so for some offenders because of the quality of supervision and services locally. In other counties with similar cases, a prosecutor might rely more on the prison system to provide treatment because locally there may be little available to address the needs of offenders. Data restrictions may limit the ability of researchers to detect such possibilities. However, the starting point involves recognizing that such patterns may exist and serve as the basis for determining whether incarceration constitutes the default mode of processing or the alternative to a non-incarcerative sanction.
Conclusion
The considerable investment in incarceration over the past three decades, especially during a period when calls for evidence-based practice were ubiquitous, suggests that policy makers believe that prison effectively reduces crime and recidivism. The scholarly record suggests that this assumption is questionable. Compelling arguments, drawing on a diverse range of criminological theories, can be made that incarceration may reduce crime and recidivism or that it may increase it (Bhati & Piquero, 2008; Cullen et al., 2011; Durlauf & Nagin, 2011; Spelman, 2006). However, the empirical record is far less compelling and leads to the conclusion that methodological limitations of research to date preclude any strong claims about the effectiveness of incarceration in reducing recidivism (Nagin et al., 2009; Spelman, 2006).
A central aim of this article was to argue that in fact the limitations of research on recidivism are greater than what prior reviews indicate and to argue that incarceration likely has variable effects that are determined by several considerations. The main limitation in many respects is a conceptual one more so than a methodological one. Accounts that focus on methodology tend to emphasize the failure of many studies to include proper controls that would allow for more confidence that estimated effects of incarceration are “real” rather than spurious. We raise no concerns here about that emphasis. It is an important one. At the same time, we submit that conceptual challenges in assessing incarceration effects have been given less attention and yet must be addressed to arrive at meaningful, generalizable statements of impact.
One departing point for the article involved the observation that courts do not blindly mete out incarceration terms irrespective of an offender’s prior sanctioning histories. These histories may serve to define which types of non-prison sanctions would be considered by the courts. In so doing, these histories and how the courts consider them define the counterfactual. In-prison experiences, such as abuse or rehabilitative programming, and post-release experiences, such as the amount or quality of supervision, may further define the “treatment” that incarceration constitutes. Essentially, these three dimensions—prior sanctioning experiences, in-prison experiences, and post-release experiences—define different types of treatment.
Accordingly, we need estimates of the effects of these different types of treatment. To make such assessments requires identifying the appropriate comparison groups. That can be a challenge. In some places, individuals who have failed on probation may be instantly considered for a jail or prison term. In other places, a term of intensive probation may be more likely. Attention to such considerations is necessary if we are to identify appropriate comparisons and, in turn, arrive at credible estimates of incarceration impacts on recidivism.
Another departing point was the observation that, as with many social programs, incarceration—however defined—may have varying effects for different groups. The corollary is that different types of incarceration experiences may have varying effects for these groups as well. Imprisonment for someone who may lose his or her job arguably creates greater harm to an individual than for someone who has no job, and in turn may result in a different effect on recidivism (Ekland-Olson et al., 1983). It may affect women differently than men (Benda, 2005; Cobbina, Huebner, & Berg, 2012; Davidson, 2011; Mears et al., 2012). Indeed, a range of dimensions exist that may moderate the effect of any type or amount of incarceration, including an individual’s racial or ethnic background, age, whether they have a serious mental illness, their history of and prospects for gainful employment, access to housing and a social support network, residing in disadvantaged or impoverished areas, and so on. Although some studies have tacked into these possibilities, there exists no systematic empirical evaluation of a range of incarceration experiences or events and their relative effectiveness for a range of groups.
These observations lead to several implications. First, as we have argued, they indicate that the scientific evidence for incarceration effects on recidivism is limited, even more so than methodologically focused critiques have suggested. The argument that incarceration is a heterogeneous experience that may have heterogeneous effects for different groups of individuals means that studies that continue to treat incarceration as a generic, or uniform, event are unlikely to provide accurate estimates of the effects of prison. As Sherman (1993) and others (e.g., Cullen & Jonson, 2014; see also Baron, 2013) have argued, what is needed is a theory of sanctioning that can provide guidance about the precise conditions under which punishments in general—and incarceration in particular—may increase or decrease offending. When coupled with credible methodological approaches, including specification of appropriate counterfactuals, the possibility of a science of punishment becomes more likely. With the rise of mass incarceration, the time for such a science would seem to be none too soon. Any efforts in this direction will need to draw on the insights of policy makers and practitioners because it is their decisions and practices that define the counterfactual conditions that determine the sanctions that constitute “usual practice” and those that constitute alternatives and thus “treatments.”
Second, and on a more positive note, the argument here provides guidance for how future studies might provide greater insight into the effects of incarceration. In particular, they point to the need to identify precisely the counterfactual conditions that are most likely and then, for purposes of external validity, to describe carefully these conditions and the population under study (Farrington, 2003). They point, too, to the importance of investigating whether identified effects may vary among different groups and, to the extent that they do, why.
Third, systematic investigation of incarceration effects in this manner affords opportunities not only to advance policy-relevant research but also to provide insights into questions of central importance to criminological theory. Deterrence doctrine, for example, is premised on assumptions about the effects of the severity of crime, but the conceptualization and measurement of severity remain significant challenges for criminologists (Cochran et al., 2014; Ekland-Olson et al., 1983; Loughran et al., 2009; Morris & Piquero, 2013). Systematic investigation of prison stays—including examination of lengths-of-stay and how punitive they feel to inmates—provides a foundation for assessing severity and its effect on recidivism.
Systematic investigation, too, of sequences of sanctions—for example, probation followed by probation versus probation followed by prison—provides a foundation for further investigating critical theoretical questions about deterrence (Nieuwbeerta et al., 2009). Perhaps initial severe sanctions, such as prison, followed by less severe sanctions, such as probation, are a more effective approach to reducing recidivism than progressing from less severe to more severe sanctions. Perhaps, too, shorter but more frequent prison terms may provide more of a deterrence to individuals (Durlauf & Nagin, 2011). It may be that, as Kleiman (2009) has argued, a greater reliance on swift and certain punishment can more effectively reduce recidivism as compared with an exclusive investment in incarceration.
It is difficult to know a priori which result would occur, in part because we know relatively little about the relative effectiveness of various combinations of swift, certain, and severe punishment. It is difficult, too, because we also know little about the extent to which those who are punished perceive different types of sanctions as severe. As discussed above, some studies suggest that probation and other intermediate sanctions may be perceived in some cases to be more punitive than prison. Regardless, these and many other possibilities bear investigating, and stand out clearly as soon as we recognize that any given sanction occurs within a context of prior sanctioning events, and that these may moderate in some way that sanction’s effects (Stafford & Warr, 1993). Given that individuals in prison frequently compile a long record of prior sanctions, including stays on probation and in jail (Petersilia, 2005), the study of sequences not only is possible but also offers a key avenue for informing policy.
Theories other than deterrence could be tested and elaborated on through investigation of incarceration effects on recidivism. Strain theory, for example, may help account for why some types of incarceration experiences may decrease recidivism (Listwan et al., 2013). At the same time, incarceration may constitute a significant strain that increases recidivism (Agnew, 2005). Understanding how and why could shed light on ways that strain theory could be modified to make more precise predictions about the types of conditions in prisons, or the types of individuals under these conditions, that make recidivism more likely or less likely. Similar possibilities hold for advancing scholarship on self-control, labeling, social bonds and social support, and other prominent theories of crime (Cullen & Jonson, 2014; Nagin et al., 2009).
Fourth, the assessment and recommendations for research here simultaneously raise questions about the wisdom of investing heavily in incarceration as a means by which to reduce recidivism. We recognize that no clear answer exists. Policy makers wish to reduce crime. Victims and society at large typically want a measure of retribution. Incarceration undoubtedly helps reduce recidivism for some individuals, to reduce crime rates to some, largely unknown, extent through incapacitation or general deterrence, and to achieve a sense of moral justice. At the same time, it likely increases recidivism for other individuals, potentially increases crime rates, especially if it diverts funds away from approaches that might effectively reduce crime, and can leave victims and the public fundamentally dissatisfied with the justice system’s response to offending. It also can have a disruptive influence on communities (Clear, 2007). Not least, it is an expensive means of promoting public safety. Given that over two thirds of prisoners are rearrested within 3 years of release, the return on investment appears to be uncertain at best, especially in light of the largely unknown effects on reducing crime rates, whether through incapacitation, deterrence, or other mechanisms (Cullen et al., 2011).
Should states revise their sentencing policies? What we can say is that to be consistent with calls for government accountability and reliance on evidence-based practice, the federal government and states will need to increase their investment in the types of research infrastructure that would allow for conclusions about when incarceration is most likely to be effective and when other approaches may make more sense. Without such infrastructure, the near-certainty is that society will spend too much on punishment for too little return. And it may well create more social harm than good. That such a possibility exists—that there would be an era of mass incarceration in an era of calls for government accountability and that greater harm rather than good may well result—remains striking. An opportunity existed to create a wide range of sanctions and systematically investigate their effectiveness for different groups. That path was not followed. It is, however, a path that remains open for investigation.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.