10

Criminal Justice System

Problems in the U.S. criminal justice system have been well documented. A 2014 report on incarceration rates described a system in a state of crisis that compares unfavorably with those in other industrialized nations (National Research Council, 2014). Estimates of the total number of people currently imprisoned in the United States vary, but according to data collected by the Department of Justice, 1,215,800 people were in state or federal prisons at the end of 2020 (Carson et al., 2021). Other estimates also include people in local jails and those incarcerated prior to trial, as well as those incarcerated in Indian territories. For example, the World Prison Brief (n.d.) estimated that 1,675,400 people were incarcerated in the United States in 2020.¹ The trend of incarceration was upward for decades: the number of people incarcerated in state and federal prisons grew from 200,000 in 1973 to 1.5 million in 2009 (National Research Council, 2014). More recently, the trend has been downward; there are varied explanations for this trend. The United States still has the highest per capita incarceration rate in the world (664 per 100,000), and, by one calculation, every state in the country also has a higher incarceration rate than nearly all democratic nations (Widra & Herring, 2021).

Those in prisons and jails are disproportionately Black, Hispanic, and Native American (Carson, 2021). The prison population reflects persistent sources of disadvantage in U.S. society: large percentages of inmates have

___________________

¹ The Prison Policy Initiative’s estimate for 2020 was 1,797,000, and they offer separate estimates for people imprisoned in U.S. territories and immigration detention (Sawyer & Wagner, 2022).

not completed high school, were first arrested before age 19, have a parent who has been incarcerated, or come from families who have received public assistance (Wang et al., 2022). Approximately one-tenth of this population had been homeless during childhood. A substantial proportion have histories of substance use disorder.

Apart from the high cost of incarcerating this many people—$80 billion a year, by one estimate—researchers have documented long-term harm to those who are imprisoned, even those convicted of minor offenses, and to the communities in which they live (see, e.g., Western & Pettit, 2010). Many analysts and policy makers argue that the current system exacerbates economic inequality, the effects of institutional racism, and problems with violence and family instability that particularly affect high-poverty communities (National Research Council, 2014). Incarceration itself, however, is only the final step in a long series of events and processes in the criminal justice system.² The process begins with the alleged commission of a crime and is followed by arrest and court processes. The court process involves charging, bail, plea bargaining, conviction, and sentencing. Sentencing has the most direct impact on the high incarceration rate. Many of the people in the system repeat the process in what is essentially a revolving door: criminal behavior eventually resulting in incarceration, which is followed by further criminal behavior, which starts the process anew. Every step in this process—including those that support the revolving door, particularly the difficulty that individuals released from prison face in establishing themselves in the labor market and civil society—involves a complex set of actors, social and legal processes and norms, policing practices, and sentencing guidelines.

INFLUENCE OF BEHAVIORAL FACTORS

The behavioral aspects of the steps in these processes have been well documented in a long and extensive literature from sociology, psychology, and economics that explores the causes of criminal behavior. For example, an influential 1968 study used traditional economic analysis to characterize the first step (criminal behavior) as the action of a rational individual accurately perceiving the short- and long-run benefits and costs of committing a crime (Becker, 1968). Becker’s theory built on very old analyses of the benefits and costs of criminal behavior, including the work of Beccaria (1872) and Bentham (1907), as well as the broader literature on deterrence

___________________

² The committee benefited from a workshop presentation by Aurélie Ouss of the University of Pennsylvania that provided background for this section. See https://www.nationalacademies.org/event/07-18-2022/workshop-on-behavioral-economics-exploring-applications-and-research-methods

theory. But Becker was the first to use all the elements of the traditional economic model, portraying an individual who commits a crime as rationally weighing the gains from committing the crime against the costs of doing so, taking into account an accurate assessment of the probability of arrest and the probability of incarceration if arrested. That person may have strong rational reasons for choosing the risk associated with the criminal behavior; in Becker’s view, the “person commits an offense if the expected utility to him exceeds the utility he could get by using his time and resources at other activities” (Becker, 1968, p. 176). Thus, Becker argued that criminal justice policy should be aimed at raising the costs of crime by increasing the penalties (e.g., increased police presence, longer prison sentences).³

In contrast to this theory, empirical studies of people who commit crimes suggest that they do not accurately perceive either the benefits or the costs of their actions, in the short run or the long run; that they often act on impulse; and that they commit acts that they later regret. While people usually have a rough sense of what types of crimes have greater penalties than others, they only imperfectly perceive the specific sanctions for criminal behavior, and their perceptions are heavily influenced by situational factors and context (Apel, 2022). Perceptions of the probability of apprehension following a crime are highly individual and based on particular experiences (including whether they have committed crimes or have been apprehended previously, which increases knowledge), not on full rational calculations. In short, people who may be considering criminal activity estimate sanction risk in ways that are influenced by a variety of personality characteristics and behavioral factors, not just rational calculation (Pickett & Bushway, 2015; Miceli, Segerson, & Earnhart, 2022). One researcher (Pogarsky, 2002) distinguishes between potential offenders who are “incorrigible,” on whom sanctions for criminal behavior have no impact, and those who are “deterrable” and might be deterred by sanctions; he regards the differences between the two as inherently behavioral. This body of work points to the behavioral influences identified in Chapter 3 as among the key building blocks of behavioral economics: limited attention, cognitive barriers, reference dependence, and inaccurate beliefs.

Police officers are also influenced by behavioral forces when carrying out arrests and in other encounters with the public, and these influences have been well documented (Owens et al., 2018). The same holds true for judges and other court officers, who, while necessarily following the

___________________

³ Many elements of Becker’s theory appear in the criminology literature in different forms. That literature has long considered the power of deterrence in models of deterrence theory, and there is also a rational choice model in that literature that is related to Becker’s theory (see Cullen, Agnew, & Wilcox, 2018, for a review of criminological theory). However, these other theories did not build so directly on the traditional economic model and do not always have the same implications for policy.

law, have discretion in court proceedings and sentencing that can reflect behavioral biases. Studies of judicial decisions have highlighted the potential for those decisions to be subjective, to be based on erroneous understandings, and to reflect ad hoc assessments of the circumstances (Bibas, 2004; Danziger, Levav, & Avnaim-Pesso, 2011; Bushway & Owens, 2013; Stevenson, 2018).

APPLICATION OF BEHAVIORAL IDEAS

Despite the widespread recognition of these behavioral issues in the criminal justice system and their importance to policies designed to prevent crime, relatively little research in this field has used the specific lens of behavioral economics (Loughran, 2019, p. 738). We review a few of the available studies to illustrate the types of topics that have been addressed, including many not conducted by economists but which have elements similar to those emphasized in behavioral economics.

Some work that has directly addressed criminal behavior itself is discussed in a review of studies that used behavioral economics to examine decision making by offenders. This review describes valuable findings from the last two decades (Pogarsky, Roche, & Pickett, 2018). For example, studies show that there are biases in how people who commit crimes assess the probabilities of arrest and detection. Biases in assessing probabilities are a central component of prospect theory (see Chapter 1), which argues that individuals’ assessments of probabilities systematically depart from true (objective) probabilities because they are affected by extraneous characteristics.

Other studies show that people are inconsistent in their estimates of those probabilities for different situations and tend to be affected by the salience and framing of the risk rather than the objective probabilities, an example of reference dependence. For example, a person considering stealing a purse from a parked car at night is much more likely to be deterred if they think the purse might have a trackable iPhone in it, even if the probability that it does is very small: this is an example of giving a small probability event excessive weight because it is salient. People also have inconsistent scales of response: they estimate probabilities of arrest very differently depending on the exact numeric scale they are offered. Pickett (2018) has suggested that more effective communication about probabilities of arrest and the use of framing effects in how those probabilities are publicly communicated could lead to specific policies that might reduce crime.

The review (Pogarsky, Roche, & Pickett, 2018) also covers studies of behavioral factors in crime deterrence, and the authors highlight key findings; for example, people who may be considering criminal behavior assess probabilities in a nonlinear fashion, using perceived thresholds, or “tipping points,” in those probabilities to make decisions, and they do not respond

to variations in probabilities below the tipping point. This behavior involves cognitive errors and inaccurate beliefs, as well as reference dependence if those tipping points are taken from some perceived, but possibly mistaken, reference point. The research showed that the tipping point changed depending on the “starting point” for the probability of arrest.

Other work covered in the review (Pogarsky, Roche, & Pickett, 2018) showed that at-risk individuals tend to be averse to ambiguity, a common finding in behavioral science: that is, people act not on the basis of the probabilities of their arrest (whether objective or subjective) but on the basis of how certain they are of those probabilities. “Ambiguity aversion,” the tendency to favor the known over the unknown, was first formalized in economics by Camerer & Weber (1992). They showed that people are more comfortable with decisions when they believe they know more about the probabilities of events even if their subjective beliefs are held fixed. This phenomenon has been shown to affect the impact of police crackdowns. For example, the deterrent effects of such crackdowns are initially small because at-risk individuals initially have not learned what the probabilities of arrest are going to be and hence do not commit crimes because of the ambiguity; however, after learning those probabilities, people are more likely to commit crimes because the ambiguity is reduced (Sherman, 1990).

One study examined whether youth at risk of being involved in criminal activity in Chicago rationally calculated benefits and costs, applying a form of cognitive-behavioral theory (Heller et al., 2017). The therapy was based on the hypothesis that youth from violent neighborhoods learn to automatically respond aggressively to any situation in which they are challenged and that such automatic responses lead to aggressive responses in situations in which they are not needed or appropriate. The study required young men to undergo training sessions where they were taught to slow down, think for a moment, and consider a response rather than automatically responding. This type of behavior modification is typical of approaches developed in the context of cognitive psychology, and it is related to the concept of heuristics that behavioral economists use (see Chapter 2). This seemingly modest behavioral intervention had a major impact and at a relatively low cost: total arrests were reduced by 28–35 percent, and violent crime arrests were reduced by 45–50 percent.⁴

In the area of policing, it is standard practice for officers to make their own judgments in each situation about whether to be aggressive in

___________________

⁴ We note that researchers in criminology have long debated the role of “human agency” in individual decisions to engage in crime, where that term refers to portraying individuals contemplating crime as purposeful and reflecting deliberate choice rather than just reflecting outside social influences (Paternoster & Bushway, 2009; see also Paternoster et al., 2015; Paternoster, 2017). Cullen (2017) proposes a balanced approach with both human agency and social influences considered.

confronting suspects or to emphasize the appearance of fairness in those decisions. In many cases, officers do not accurately perceive the views of suspects, who may interpret police behavior as unfair and violating their expectations for just interactions with law enforcement. In a study with some similarities to the Chicago study of at-risk youth described above, researchers tested an intervention that trained police officers in moderate risk areas to slow down their thought processes during citizen encounters: the training resulted in officers being less likely to resolve encounters by arrests (Owens et al., 2018).

In another intervention with both police and a community’s relationship to police, researchers tested the hypothesis that neighborhood residents’ perception of police officers in their area was affected by whether they had any personal knowledge of the local officers (Shah & LaForest, 2022). The study was carried out in low-income neighborhoods. One intervention involved providing neighborhood residents with mundane information about officers’ favorite food, their hobbies, or why they became an officer; another intervention involved officers handing out “outreach” cards to residents. The intervention reduced crime rates in the first three months after the intervention.

Researchers have also examined court practices and behaviors, including the frequent problem of people who have been charged with a crime failing to appear in court. The most common financial incentive to appear is the setting of bail that will be forfeited if the accused does not appear, but some evidence suggests that bail does not, in fact, have much effect on the probability of appearance (Ouss & Stevenson, 2022). A study in New York City showed that 40 percent of those issued summons to appear failed to do so, leading to warrants for their arrest, which further worsened the accused’s situation (Fishbane, Ouss, & Shah, 2020). The authors hypothesized that many people simply forgot about the need to appear or did not understand the complicated summons they had received, an example of limited attention and cognition. The authors tested an intervention that included use of summons documents redesigned to prominently display the designated court appearance date and also highlight the consequences of not appearing (i.e., arrest warrant) as well as text message reminders. The intervention significantly reduced failures to appear, resulting in 30,000 fewer arrest warrants. The effect was strongest in neighborhoods with low levels of income and high proportions of Black and Hispanic residents. This evidence that many people who fail to appear in court did not intend to do so contradicted the view of many lay people interviewed by the authors, who said that failures to appear are intentional.

A significant body of work has examined the effect of sentencing length on criminal behavior and identified the importance of behavioral concepts. For example, a study of recidivism by those who have been convicted and

incarcerated showed that recidivism was affected by whether the convicted people framed their sentences as a loss or gain relative to their expectations, an example of reference dependence (Bushway & Owens, 2013). Specifically, people who were given sentences shorter than those recommended in sentencing guidelines were more likely to re-offend than those given sentences that were in line with the guidelines. The authors recommended aligning actual sentences with guidelines.

More broadly, researchers have found that making long sentences longer has little impact on crime rates (National Research Council, 2014). There is also evidence that certain types of proactive policing, if carefully designed to target areas and types of offenders with the highest rates of crime, can have a deterrent effect (National Academies of Sciences, Engineering, and Medicine, 2018). These findings are consistent with the behavioral economics concept of present bias (that individuals are more responsive to immediate costs and benefits and less responsive to important long-run factors).⁵ However, much about deterrence is not well understood, and the National Research Council (2012) has called for more research into how perceptions of the risks of capital punishment, for example, are formed.

Several studies have also examined behavioral effects on plea bargaining. The standard model of plea bargaining is that a defendant bases a decision about whether to accept a plea bargain on the probability of conviction if the case goes to trial, the expected length of the sentence if convicted, and the characteristics of the plea bargain itself.⁶ Yet one study showed that bargaining outcomes are also affected by attorney competence, compensation, and workloads; resources; and sentencing and bail rules, all of which are consistent with the traditional economic model. But outcomes were also found to be affected by lack of information about the probability of conviction, opening the door to biased perceptions that are based on salience, framing, and other behavioral factors (Bibas, 2004).

A randomized controlled trial in which students were told to assume that they were innocent or guilty and then were offered plea deals offers additional evidence (Garnier-Dykstra & Wilson, 2021). When the plea deals were framed as a gain (i.e., a greater-than-expected reduction in the seriousness of the charge), the innocent students were more likely to accept the deal (and plead guilty), while the guilty were less likely to do so. This is another example of reference dependence, where the decision is based in

___________________

⁵Becker (1968) also noted that individuals may respond more to the probability of arrest than to the expected sentence but interpreted this as an indication that individuals prefer risk; in contrast, in behavioral economics this phenomenon is interpreted as present bias. The finding that immediate punishment appears to have more impact than long-run, or delayed, punishment is an old finding in the literature (see, e.g., Perry, Erev, & Haruvy, 2002).

⁶ There are a number of models of rational plea bargaining, including by Landes (1971), which frame the issue in terms of a rational economic bargaining model.

part on perceived comparisons with other possible outcomes: in general, innocent defendants are less likely to accept deals than guilty defendants—as traditional theory would predict, since innocence should be positively correlated with the probability of an acquittal at trial. Some innocent defendants were influenced to accept the deal if it was framed as a particularly strong gain. In contrast, however, a randomized trial that used actual probabilities of conviction and sentence length showed that prosecutors and defense attorneys generally acted as would be predicted by the traditional economic model (Bushway, Redlich, & Norris, 2014).

FINDINGS

Behavioral factors that are not generally considered in traditional economic models have an influence at all points in the criminal justice system: they influence the determinants of criminal behavior, policing practices, court proceedings, judicial decision making, and the effects of incarceration. Key findings from the research include:

• People who are considering committing crimes are heavily affected by inaccurate perceptions of the probabilities of arrest. They are also affected by reference dependence and framing when assessing the subjective probabilities of arrest.
• Police officers have inaccurate beliefs about community residents’ perceptions of their actions, which lead them to engage in suboptimal behavior in situations that require decisions about apprehension and arrest.
• Courts do not accurately assess the perceptions of defendants regarding bail penalties, use of sentencing guidelines, and the role of information in the trial process. The reactions of defendants are not what courts assume, and the result can be poorer outcomes for defendants and society.

Despite these findings, the evidence for the effects of behavioral interventions at various steps of the criminal justice system is too scant so far to support definite conclusions on their effectiveness and potential. More research is needed on the role of behavioral factors in the criminal justice system and on the most effective interventions to reduce criminal behavior and to improve policing and judicial processes. Research is needed in several key areas:

• how to directly address the behavioral biases that people have when choosing to commit a crime, including their biased assessments of the consequences of criminal behavior;

• how to more effectively engage police and their communities to reduce the biased perceptions each has of the other; and
• how to improve court processes to provide defendants with accurate information and to reduce their biased perceptions of the outcomes, as well as to provide judges and attorneys with more knowledge and understanding of the behavioral biases of defendants.

REFERENCES

This page intentionally left blank.