View this issue as a PDF: CrimHighlightsV17N6.pdf
This issue of Criminological Highlights addresses the following questions:
- How does the criminal justice system impose punishments before trial?
- Is the impact of a short prison sentence on future employment any different from the impact of probation?
- Does allowing prisoners to be in the community for short periods of time during their prison sentences threaten public safety?
- Do governments design prisons that will inspire prisoners to lead better lives?
- Does the shortening of prison sentences threaten public safety?
- Is the use of police powers to stop and search members of the public an effective crime reduction technique?
- How fair are risk prediction instruments based on fancy looking algorithms?
- Are there proven techniques to reduce reoffending by those released from prison after serving sentences for sex offences?
Punishment in the criminal justice system starts long before anyone is found guilty and sentenced. Those arrested and detained in prison prior to trial – a larger number in Canada than those receiving prison sentences after conviction – often receive serious punishments as a consequence of their arrest, their experience in courts, and their time in prison, even if they are eventually not convicted or sentenced to prison.
In Ontario, Canada, in 2015, over 91 thousand people were arrested and held for a bail hearing. Although more than a third of them were, ultimately, not found guilty of any offence, it would be wrong to suggest that they – as well as those ultimately found guilty – were not punished prior to the disposition of their cases.
“The hidden nature of the process of arrest, court appearances, and detention obscures the ways in which people may experience important consequences and even punishment on remand, without ever being formally ‘marked’ by a criminal conviction” (p. 2). This paper uses data from interviews (1-3 hours in length) with 120 people (60 men and 60 women) held in pretrial custody in four large correctional institutions in Ontario.
When people were arrested, it was typically unexpected. Although people had a legal right to consult a lawyer following their arrest (typically a phone call to a legal aid lawyer), the vast majority (84%) were not given an opportunity to notify family or kinship networks upon their arrest, making it difficult for people to get personal assistance in responding to their detention. Even when there was no active mistreatment by the police, other punishments (e.g., not having access to prescription medications) followed automatically from the accused person’s removal from daily routines and isolation from normal contacts and supports.
If an accused’s bail decision was not made at the first court appearance, they would be transported to a provincial prison. Court days – often more than one before bail was resolved – typically meant starting the day at 5 a.m. and being transported to court while handcuffed (sometimes also with ankle shackles) and often chained to at least one other prisoner. The conditions of the court’s holding facilities, waiting for the case, and the uncertainty of what would occur, were each seen as stressful. Lunch typically consisted of a single granola bar. The logistics of transportation were such that when the accused were returned to prison, they were too late for dinner. Some prisoners were required to appear in court in whatever clothing they had on when arrested (e.g., a bikini covered by a paper jumpsuit in the case of one woman arrested at a hotel swimming pool). Not surprisingly, 71% of those interviewed preferred, if possible, to appear in court by video link rather than in person.
Prisoners were held in ‘basic’ cells in maximum security institutions. New technologies have made communication with the outside world more difficult. Communication is important because remand prisoners often need to talk to friends and relatives to create a plan for release that will satisfy a court. Prisoners could only call a land line because all calls were “collect” and cellular phones do not accept collect calls. Lockdowns, when visitors were not permitted into the institutions, occurred frequently. These barriers made ordinary visits with those outside difficult, but also made preparation of a bail release plan that could be presented to court an extra challenge.
Conclusion: In recent years, the calculation of credit off one’s sentence for pretrial custody has been a contentious issue in Canada. However, the focus on credit off a sentence “neglects the fact that many people who enter remand imprisonment return to their communities without a conviction…. Participants in this study described the harms of arrest and making court appearances as the most visceral and painful aspects of remand imprisonment…. Each [part of the system – police, courts, and corrections] plays a separate but related role in forming the experience of punishment for remand prisoners” (p. 15). With the common focus on the sentence as the punishment, it is important to remember that “the collective weight of the cross-institutional remand process imposes harms on individuals that can have substantial and negative consequences on their lives in the short and long term” (p. 16).
Reference: Pelvin, Holly (2019). Remand as a Cross-Institutional System: Examining the Process of Punishment before Conviction. Canadian Journal of Criminology and Criminal Justice (online pre-publication).
Imprisoning men for 4 to 10 months, rather than sentencing them to probation, reduces the likelihood that they will be employed in the three years after they return to the community.
Those released from prison are typically disadvantaged economically and have difficulty finding employment. One might expect that imprisonment would interfere with employment for many reasons including: (a) The stigma associated with prison may make employers especially reluctant to employ someone who was imprisoned; (b) Imprisonment may strengthen the person’s criminal self-identity; (c) Prisoners will not accumulate work experience commensurate with that of someone in the community.
This study, carried out in Sweden, examines the relationship between imprisonment and subsequent employment. Data from other Scandinavian studies suggest that those who are imprisoned tend to have weak links with employment even before their imprisonment. Hence there is an obvious challenge in understanding whether the experience of imprisonment itself has negative impacts on subsequent employment.
This paper uses data from all males born in 1975 and 1980 who were residents of Sweden at age 16. By age 30, 28.9% had at least one criminal conviction and 3.7% had been sentenced to prison (typically when age 20 or older). Because of the ability to link administrative data on Swedish citizens, not only were (legal) employment records available but other relevant data (e.g., education records, immigration status, mental health diagnoses, criminal records) were also available. For matching purposes, the analysis was restricted to those sentenced either to prison or probation for the first time at age 20-25. The focus was on those who received an intermediate length prison term of 4-10 months. Shorter prison sentences might have been served in the community with electronic monitoring; longer sentences were more difficult to find equivalent cases that received probation.
Propensity score matching was carried out in two stages. First, the likelihood of going to prison was estimated for all those who received either a 4-10 month prison sentence or probation. Next, a person who was in fact sentenced to prison was matched with 3 people who had the same likelihood of going to prison but instead received a sentence of probation.
Compared to those who were never convicted of a criminal offence, involvement in work between age 20 and age 30 for those who were convicted of a criminal offence and received probation sentences was, not surprisingly, lower. Involvement in work was lower still for those who spent 4-10 months in prison.
When one looks at equivalant groups of men who either received probation or prison sentences, there were no differences in employment records prior to being sentenced. For the first three years after being sentenced, however, those who received a prison sentence were less likely to be involved in work than those who received probation. This difference was no longer significant after 4 years. There was some indication that the effect was driven largely by those who had been working prior to their conviction.
Conclusion: Men sentenced for criminal matters tend to be disadvantaged in many ways, including their ability to get jobs. Being imprisoned for 4-10 months – especially for those who were working prior to being convicted – appears to make it even less likely that they will be employed in the first few years after they have served their sentences.
Reference: Bäckman, Olaf, Filipe Estrada and Anders Nilsson (2018). Locked Up and Locked Out? The Impact of Imprisonment on Labour Market Attachment. British Journal of Criminology, 58, 1044-1065.
Allowing prisoners the opportunity to leave prison for short periods of time appears to make correctional sense: those allowed temporary absences from Canadian penitentiaries are less likely to be unemployed when they achieve full release from prison and are less likely to be returned to prison (either for a new offence or a violation of a condition of release).
Canadian prisoners can be released temporarily from penitentiaries (where they are serving sentences of 2 years or more) by way of either an “escorted temporary absence” (ETA), where they are supervised, or an “unescorted” (UTA), where they are not supervised. Typically, prisoners who have received a UTA have previously had successful ETAs.
This paper examines whether ETAs/UTAs are effective in helping prisoners obtain employment and if they contribute to lower rates of reoffending or revocation of full releases because of violations of conditions of release. In Canada, prisoners are normally eligible for release on (full) parole after serving 1/3 of their sentences and, if refused parole, they are typically released with supervision after serving 2/3 of their sentences. Excluding any temporary absence (TA) for medical or administrative purpose, the study examined TAs granted between 2005 and 2011 to prisoners during their first sentence within this period.
TAs (both ETAs and UTAs) are generally more available to “offenders with less history of serious offending and [who were] less likely to have previous problems in institutions and while on [previous sentences involving] community supervision” (p. 29). Therefore, it was necessary to assess the effects of ETAs and UTAs in comparison to equivalent prisoners who did not receive either form of TA. To create equivalent groups, propensity score matching was carried out whereby background characteristics were used to calculate the likelihood that a person would receive a TA. Twenty-seven variables were used to calculate propensity scores for receiving a UTA (28 for an ETA). These included various forms of risk, current offence and sentence, previous convictions, a statistical prediction of recidivism and reintegration potential, age, whether the offender was Indigenous, and sex. Hence any effects that were obtained are over-and-above the best prediction available to Correctional Service Canada on recidivism and reintegration potential (as well as other variables). It should be pointed out that the failure rate for those on TAs was very low – approximately 1%.
The results are straightforward. The effects of TAs overall as well as ETAs and UTAs were largest (and significant) for employment. Unemployment (defined as no record of employment during the two years after release) was reduced if the prisoner received either an ETA or a UTA. In addition, returns after full release for technical violations or for new offences were reduced if a prisoner had been granted an ETA or a UTA. In general, the effects were larger for UTAs than for ETAs, meaning that UTAs were more effective than ETAs in reducing unemployment, reoffending, and violating conditions of release.
Relatively few prisoners (about 4% of the sample) ever received a UTA. Hence it was not possible to test whether higher numbers of UTAs led to reduced unemployment and lowered re-offending or rates of technical violations. But for ETAs and TAs (overall) it appears that the more temporary absences that a prisoner received, the more favourable the outcome after (full) release from prison.
Conclusion: The results support the view that taking small risks in the form of releasing prisoners on escorted and unescorted temporary absences reduces, in the two years following release, three problems: (a) the likelihood that the prisoner, when released, would be unemployed; (b) the rate of return to prison for a violation of the conditions of release: and (c) recidivism. Generally speaking, more is better: “The more absences an offender received, the less likely they were to have negative outcomes in the community” (p. 37).
Refernece: Helmus, L. Maaike and Marguerite Ternes (2017). Temporary Absences from Prison in Canada Reduce Unemployment and Reoffending: Evidence for Dosage Effects from an Exploratory Study. Psychology, Public Policy, and Law, 21, 23-38.
To minimize the harmful effects of imprisonment and to contribute to the rehabilitation of prisoners, the architectural design of prisons should follow principles that will “inspire prisoners and motivate them to lead better lives” (p. 329).
Concern that prison design may hinder rather than help the rehabilitation of prisoners has been discussed since the late-18th century when John Howard was promoting healthy prisons in various parts of the world. The concern about prison design was raised in an important Canadian report when, in August 1966, the Canadian Committee on Corrections (the “Ouimet Committee”) submitted a report on the design of a new penitentiary that was to be built in Quebec. The committee suggested a number of important changes be made to create a more rehabilitative institution. The Solicitor General of Canada had asked the committee to comment on a draft design because of obvious concerns that it might not foster rehabilitation.
Prisons are often designed to be built rapidly rather than being designed to serve broader rehabilitative needs. Unfortunately, one of the results of the political need to spend as little money as possible and to have the prison operational as quickly as possible is that the long-term interests of prisoners and prison employees are given little weight. Rarely, for example, are prisoners, ex-prisoners, or prison staff consulted on any important aspects of design of prisons. In contrast, when building other institutions (e.g., residential care homes for the elderly) those using and working in the facility are typically part of the development team. Similarly, there is an understanding now, among architects, that the design of hospitals can contribute to the health of those who are in them.
Prisoners, for example, need some opportunities for privacy, for reasonable heating and ventilation, for high-quality family visits. They can benefit from the ability to do work and programming without unnecessary impediments, and they can benefit from having some choices or autonomy over these matters. These same issues differentiate healthy from ‘unhealthy’ hospitals. Often held out as an example of good design, Norway’s most progressive prison was designed “to inspire prisoners and motivate them to lead better lives” rather than to mirror the lives they came from. Hence, although issues of security were not ignored, attempts were made to minimize the punitive aspects of normal prison design (bars, ever-visible security, etc.). Similarly, two high security prisons in Catalonia “attempt to put prisoner rehabilitation at the heart of design” (p. 329) by allowing or even encouraging movement (which is monitored with closed circuit TV) and interaction with others. The actual cells are described as “basic” in part because it is assumed that prisoners will not spend much time in them.
It was also noted that “If prison architects design high-quality facilities for staff – working spaces where they feel safe and able to exercise their power, interpersonal skills and discretion appropriately” (p. 332) – while simultaneously providing pleasant work spaces (and even adequate parking), “a prison is more likely to have a happy and motivated workforce who feel invested in and valued as the considerable assets they are” (p. 333). In contrast, in many English prisons, “staff feel that the provisions made for them are very much a second thought” (p. 333).
Conclusion: “Buildings help to enact ideologies through the social practices they enable and encourage…. The notion of trauma-informed design is beginning to penetrate discussions of planned new women’s prisons [in the UK]. [At the same time] the topic of ‘good’ prison design is… fraught with political minefields” (p. 333). Good prison design need not be seen as ‘softer’ or ‘prettier’; good design need not challenge the institution of the prison. At the same time, however, “designing prison spaces that support rehabilitation and desistance could be a vital component in achieving radical justice reform….” (p. 334).
Reference: Jewkes, Yvonne (2018). Just Design: Healthy Prisons and the Architecture of Hope. Australian & New Zealand Journal of Criminology, 51, 319-338.
US imprisonment rates have been declining for about 8 years. At the current average rate of decline, US imprisonment will be at the 1980 rate in the year 2052. Releasing prisoners before their expected release dates would ensure a quicker return to 1980 levels without increasing crime.
An earlier study (Criminological Highlights 14(5)#6) suggested that reoffending rates for US federal prisoners who were unexpectedly given a reduction in their sentences were no different from the reoffending rates of those who served their full sentences. That paper looked at those released in 2008 as a result of US Sentencing Commission guideline changes that were retroactively applied.
In addition to reducing the disparity in sentencing for those involved in cocaine vs. crack drug offences, the 2010 US Fair Sentencing Act also eliminated or restricted certain mandatory minimum sentences. The US Sentencing Commission allowed these changes to be applied retroactively.
The results were consistent across comparisons. Overall the three-year recidivism rates (arrests for new offences, or a court or supervision violation) were identical – 37.9%. About a third those who “reoffended” committed ‘court or supervision’ violations. This finding – similar recidivism rates for those serving their full sentences and those released early – held for all three major race/ethnicity groups (Black, White, Hispanic), males and females, those with varying amounts of education, those under and over 30 years old, those with minor and substantial criminal history scores, and those whose offences involved or did not involve a weapon. The prisoners examined in this study varied in terms of whether they had been sentenced within, above, or below, the sentencing guideline range. Again, however, there were no differences, within each of these groups, on the recidivism rates for those who served their full sentences and those who received the benefit of the retroactive changes in their sentences. Similarly, the findings held for those with sentences under 10 years in length and 10 years or more.
Conclusion: An unexpected reduction of prison time of (on average) 30 months did not affect reoffending rates for these drug offenders. This finding – no change in reoffending – held for various subgroups that were examined. Obviously, one cannot automatically assume that the results would be the same if other prisoners were unexpectedly released from prison early. The results do suggest, however, that presumptively there is little risk to public safety by extending these early-release policies to others serving time in prison at least for non-violent offences.
Reference: Hunt, Kim Steven, Kevin Maass, and Todd Kostyhak (2018). Recidivism Among Federal Offenders Receiving Retroactive Sentence Reductions: The 2011 Fair Sentencing Act Guideline Amendment. Washington, D.C.: United States Sentencing Commission.
The use of police powers to stop and search members of the public is not an effective way of deterring crime.
The use of stop and search (S&S) powers by the police is controversial for many reasons, including the fact that S&S often is used disproportionately on racial/ethnic individuals. In 2014/15 in England, for example, Blacks were four times as likely to be searched than Whites. A common justification for the use of S&Ss is that they are effective in deterring crime.
This study looks at 10 years of data from London, England, to determine what impact these activities have on crime rates in the location in which the S&S takes place. The impact of S&S on crime “is not the legal justification for most searches” (p. 1213). Rather, S&Ss are typically described as being used for investigatory purposes. Nevertheless, police often talk about S&S powers in terms of controlling crime in specific locations. This makes it difficult to separate out the effects of S&Ss from the possible impact of high concentrations of police in particular locations (e.g., hot-spot policing) whether they are doing S&Ss or not. In other words, if S&Ss were actually associated with lower crime in a particular location, it could be that the effect was due to higher police concentration rather than S&S activities.
For this study, data from 31 of the 32 London boroughs (excluding Westminster – the central area including the Houses of Parliament, Buckingham palace, etc.) were examined. The effects of S&Ss on categories of crime that might be susceptible to S&S as well as ambulance calls for “stab/shot/weapon wounds” were examined for the period April 2004 to November 2014. Daily rates of recorded S&S were obtained for each borough.
The approach was to see whether S&S activities affected crime in the following week or month on the theory that people might adjust their (crime) activity on the basis of what was happening in a particular part of the city. In order to control for general police activities, the number of police in each borough in the time period was controlled for as were long-term changes in crime in the borough.
The effects of searches on crime were examined for the following week and the following month. All searches and specific types of justifications for searches were examined. There were no consistent effects of S&Ss for non-domestic violent crime, burglary, robbery and theft, vehicle crime and criminal damage. In addition, using crime-injury ambulance calls as a second indicator of (serious) violence, there was no evidence that increased S&S reduced violence.
There was, however, a significant impact on drug crime: a 10% increase in total searches resulted in a 1.85% reduction in recorded drug offences in the following month. Though this would appear to support the idea that “drug users and dealers stop offending when perceived risk reaches a certain level… another possibility is that higher rates of S&S prompt people to change their behaviour to make it harder for officers to uncover drugs… or that people carrying drugs are simply displaced to nearby areas that are less ‘hot’ in terms of police activity” (p. 1224-5). Furthermore, of course, “police recorded crime data are unlikely to be the most reliable measure of drug crime… [since they] will depend largely on police activity that discovers people in possession of drugs and not on the underlying prevalence of drug use” (p. 1225).
Conclusion: Though there is some indication of an effect of stop and search on drug crime (which might still be explained by other mechanisms), the effects overall do not suggest that stop and search by police is effective in reducing the kinds of crime that would appear to be most likely to be affected by this approach. Given these results, it may be important to re-focus attention on an issue not addressed in this study: “the need to limit its use to appropriate situations to avoid damage to public trust and police legitimacy” (p. 1227).
Reference: Tiratelli, Matteo, Paul Quinton, and Bed Bradford (2018). Does Stop and Search Deter Crime? Evidence From Ten Years of London-Wide Data. British Journal of Criminology, 58, 1212-1231.
A commercially available algorithmic pretrial risk assessment system, COMPAS, disadvantages accused people who are Hispanic.
“Automated risk assessment is all the rage in the criminal justice system” (p. 1). Previous research (Criminological Highlights 17(2)#1) has suggested that these instruments aren’t better than ordinary people’s intuitive risk assessments and that they disadvantage Black people.
One problem with these risk assessment programs is that the manner in which predictions are made is proprietary knowledge and therefore cannot be directly assessed. Hence if a risk assessment tool does not take into account the unique characteristics of a group (e.g., women, Indigenous people, Hispanics), it may unfairly disadvantage members of that group because factors that lead to increased risk scores for groups on which the instrument was originally validated (often white men) may not be at all relevant for members of other groups.
The prediction algorithm that is the focus of this study – COMPAS – uses about two dozen measures to predict general and violent recidivism. The study analyzes Florida data on 6,172 cases in which COMPAS was used to predict general recidivism and 4,020 cases in which predictions of violent recidivism were made. Recidivism was defined as re-arrest within 2 years. The study examines the recidivism rates of Hispanic and non-Hispanic accused people arrested in Broward County, Florida.
There are two ways in which a scale can disadvantage a particular group, in this case Hispanics. First, the test may not discriminate as well for those in the minority group as it does for those in the majority group on which it was developed. Second, members of a minority group with the same scores as members of the majority group may, in fact, be less likely to re-offend.
COMPAS, it would appear, has both of these problems. For example, for non-Hispanics, the group on which the measure was developed, the higher the score, the more likely it was that a person would reoffend. This was true for both general recidivism and violent recidivism. For Hispanic accused, on the other hand, the likelihood of general recidivism was the same for those with medium and high COMPAS scores. For violent recidivism, the actual recidivism rates were essentially the same for those with low and moderate COMPAS scores.
More dramatic, perhaps, are the comparisons between Hispanic and non-Hispanic accused people. Hispanics and non-Hispanics who were predicted to have high general recidivism differed dramatically in their actual recidivism rates: Hispanic accused were considerably less likely to reoffend than non-Hispanic accused with the same scores. In a similar vein, Hispanic accused who were predicted by the algorithm to have a moderate likelihood of violent recidivism were considerably less likely to reoffend than non-Hispanics with similar COMPAS scores. Other analyses resulted in essentially the same findings: For general and violent recidivism, the COMPAS score was less effective at predicting recidivism for Hispanics as it was for non-Hispanics.
Conclusion: The overall data are very clear. Though the algorithms do not explicitly consider cultural groups, the algorithm is considerably less accurate in predicting recidivism for Hispanics. The data demonstrate that an ethnicity-neutral algorithm over-predicts recidivism for Hispanics. Said differently, the algorithm systematically makes certain Hispanic accused people look more dangerous than, in fact, they are. Clearly “greater care should be taken to ensure that proper validation studies should be undertaken to confirm that any algorithmic risk is fair for its intended population and subpopulations” (p.29).
Reference: Hamilton, Melissa (2019). The Biased Algorithm: Evidence of Disparate Impact on Hispanics. American Criminal Law Review, 56, (in press).
This study replicates, using a randomized control trial, the findings from previous studies of a high intensity community program for sex offenders being released from prison into the community. The program is shown to be effective in reducing reoffending for moderate risk sex offenders followed for an average of about 6 years.
Although many people apparently believe that programs for sex offenders don’t work, this turns out not to be true (Criminological Highlights 9(5)#7). One of the programs that has been demonstrated to be effective in helping sex offenders who were released from Canadian penitentiaries without any supervision or support is “Circles of Support and Accountability” (CoSA) (Criminological Highlights 9(3)#6, 11(2)#6).
This paper extends the previous research by examining recidivism for 100 moderate risk sex offenders all of whom had volunteered to this program. They were randomly assigned to participate in the CoSA program (n=50) or ‘treatment as usual’ (n=50) while still in prison but soon to be on supervision. In this implementation in Minnesota, the sex offender (or ‘central member’ [CM] of the circle) met several times with 4-6 trained community volunteers while still in prison. These same people formed the CoSA for the CM in the community after release.
The volunteers, who formed the other members of the circle, were typically local college students. Because the CoSA approach is labour intensive, getting volunteers who were willing to make a long-term commitment (weekly meetings in a secure public venue with the CM for 6-12 months) was difficult. As a result, people were identified as potential CMs only when sufficient volunteers had been identified for a CoSA. Because there were more sex offenders being released from prison than could be accommodated in the program, random assignment was seen as the fairest way to decide which prisoners would receive the CoSA treatment and which would receive normal community supervision.
The program was carried out in a number of locations in Minnesota. The data for this study (for 50 CoSA CMs and 50 controls) come from CoSAs carried out between January 2008 and June 2016. The cutoff date for the collection of recidivism data was June 2017, meaning that the follow-up period varied from 12 to 113 months (mean = 73 months). Because of random assignment, the follow-up period did not differ across groups.
Prisoners who received the CoSA treatment were significantly less likely to recidivate according to all four measures that were examined. Consistent with other research, sex offender re-arrest was low for both groups (CoSA = 2%, Control=14%). No CoSA CMs were reconvicted of a sex offence, compared to 8% of the controls. Many of the 100 people were rearrested for some offence (CoSA = 60%; Control 80%), and many were reconvicted of some offence (COSA = 50%, Control = 68%).
Conclusion: This study, using random assignment – the “gold standard” of evaluation research – provides strong evidence confirming the results of previous studies of Circles of Support and Accountability. This approach involves a commitment of 4-6 community volunteers to meet weekly for 6-12 months to help those returning to the community after serving a prison sentence for a sex offence. The results are consistent across four measures of recidivism. The program reduced both general and sexual reoffending in a period lasting an average of about 6 years after release from prison.
Reference: Duwe, Grant (2018). Can Circles of Support and Accountability (CoSA) Significantly Reduce Recidivism? Results from a Randomized Controlled Trial in Minnesota. Journal of Experimental Criminology, 14, 463-484.