Criminological Highlights Vol. 22, No. 2 - April 2025

Item 1

Compared to similar people who are released back into the community almost immediately after being arrested, accused people who are held in pretrial detention for more than a week before they are released have a higher likelihood of missed court appearances, new arrests, and new arrests for violent crimes.

Detaining people – even for relatively short periods of time – is an important part of criminal justice punishment. In Canada, for example, 46.3% of federal and provincial/territorial prisoners were awaiting trial on an average night in 2022.

There is a fair amount of evidence that pretrial detention itself is harmful and does not appear to reduce overall offending (e.g., Criminological Highlights 17(2)#7, 17(3)#1, 17(5)#3, 18(4)#1, 21(4)#4), 21(5)#7, #8).

In some jurisdictions, such as the three US counties contributing data to this study, people are often arrested and imprisoned, but released very quickly (i.e., within one day or less). But many accused people are held for at least a week before being released. This study examines the impact of a stay of more than 7 days in remand custody before being released into the community. One group of people charged with an offence were released almost immediately or the day after being arrested. They were compared to a group of people who were charged with an offence and remained in custody for at least 7 days before eventually being released into the community. Clearly, these two groups (released in less than 1 day vs. released after more than 7 days) are different. Therefore, in all analyses, various factors (e.g., previous criminal justice outcomes, previous incarcerations, number of charges, type of offences, race, age) were controlled statistically. The study looked at how these accused people behaved during the roughly 6-month window when they were at risk of misbehaving after release into the community.

Expressed as estimates of what would happen with these two outcomes (controlling for other factors), the results suggest that about 25% of those who were detained for more than 7 days would fail to appear in court at least once, compared to 18% of those detained one day or less. Those detained for more than 7 days were also more likely than those released almost immediately to be arrested for a new criminal offence (28% vs. 21%) and a new violent criminal offence (9% vs 6%). Finally, consistent with a substantial amount of research, being detained for longer is associated with a higher conviction rate (50% vs. 41%).

The findings are a reminder that pretrial detention is, indeed, a punishment (in terms of what it does to the accused person) and is also a cause of crime. 

Conclusion: “Pretrial detention is unlikely to achieve the crime prevention goals set out for jails. Rather, being detained pretrial for more than 7 days (compared to 1 day or less) appears to increase missed court appearances, arrests, and convictions” (p. 115). When people suggest that their community has a soft bail system and that pretrial release, therefore, is an important cause of crime, they are probably focusing on two things. First, they are only thinking about the short-term impact of being incapacitated. Second, they may be drawing attention to the inherent imperfections of the criminal justice system: some people who are released on bail do, of course, commit new offences. The question is not whether people released on bail sometimes commit new offences. The question is whether, in the long run, those released from pretrial detention commit more crime than those detained for a long period while awaiting trial.   This paper suggests that pretrial detention for over 7 days is likely to lead to an increase in crime when compared to the alternative (release).

Reference: DeMichele, Matthew, Ian A. Silver, and Ryan M. Labrecque (2025). Locked up and Awaiting Trial: Testing the Criminogenic and Punitive Effects of Spending a Week or More in Pretrial Detention. Criminology & Public Policy, 24, 99-121.

Item 2

A scale that appears to have the same error rate for two groups (Indigenous and non-Indigenous prisoners) can still discriminate against Indigenous prisoners because the errors for the two groups are substantially different.

Scales developed by psychologists are often used to predict misbehaviour in prison or reoffending when people are released from prison. Often, their use is justified for a variety of groups because some statistics indicate that their “error rate” is the same across groups. This paper examines whether the types of errors that these scales make are the same for different groups.

In assessing the accuracy of a scale, there are two types of errors that can be made. A scale can predict that a person will misbehave when in fact they do not (“false positive”), or it can predict that a person will not misbehave and they do (“false negative”). It’s easy to have no false positives: simply predict that nobody will misbehave, just as it is easy to have no false negatives (predict everyone will misbehave). The “meaning” of any accuracy index, then, is quite complex. Many scales used to predict whether someone will misbehave try to reduce “accuracy” to a single number. The measure often used by psychologists – “Area Under the Curve” (AUC) – is one of them.  An AUC of .5 is equivalent to flipping a coin, and 1.0 is a perfect prediction.

A Canadian study of the “Psychopathy Check List,” for example, found that the AUC for Indigenous and non-Indigenous penitentiary prisoners was about the same. However, actual accuracy was fairly low for both groups (AUC ≈. .65) despite it being “statistically significant” (See also Criminological Highlights, 18(5)#6). This is very similar to results described in a Correctional Service Canada “Research in Brief” (2023: RIB-23-35) report of its “Criminal Risk Index” prediction of reoffence rates (AUCs of .64 for Indigenous and .69 for non-Indigenous). 

The current study was carried out in New Zealand and used Psychopathy Check List data from Indigenous and non-Indigenous prisoners assessed between 1998 and 2020. Recidivism was operationalized as a new offence resulting in conviction within 5 years of the prisoner’s release. Previous research, using these same data, had demonstrated “statistically equivalent recidivism discrimination performance within Maori [Indigenous] and [non-Indigenous] and [the overall sample]” (p. 456). This conclusion, like that of the Correctional Service Canada study, was based on the finding that the AUCs were similar.

However, a closer analysis of the New Zealand findings demonstrates that Indigenous prisoners were more likely to score high on this measure and were more likely to be reconvicted than non-Indigenous. The analysis presented here also demonstrated that the types of errors were different for the two groups. Of the Indigenous prisoners who were not reconvicted after release, 62% had been predicted to reoffend. For non-Indigenous prisoners, “only” 28% of those who did not reoffend had been predicted to offend. Looking at those who were reconvicted, 25% of the Indigenous had been predicted not to reoffend, compared to 38% of the non-Indigenous prisoners. From a prisoner’s perspective, a similar error rate isn’t very helpful. In situations like this (where the groups differ in various important ways), a similar overall error rate does not describe what kinds of errors are being made. One group may be disadvantaged more than another even though their error rates look similar.  

Conclusion: These findings should be seen as part of a long series of empirical findings suggesting that predictive instruments developed using one group do not necessarily work the same – or as well – with other groups. Questions that predict the outcome variable for one group may not predict for other groups. And, as this paper shows, the types of errors that are made may be different for different groups. An index of accuracy may treat as equal different types of errors (false positives and false negatives) for different groups. From the perspective of a prisoner, being falsely predicted to reoffend (more likely in this study to be made for Indigenous prisoners) is not the same as being falsely predicted to be safe (more likely to be predicted for non-Indigenous prisoners).

Reference: Botha, Ryan, Tara Nichols, and Devon L. L. Polaschek (2025). Cross-Cultural Fairness of the PCL:SV – A New Zealand Analysis. Criminal Justice & Behavior, 52(3), 447-463.

Item 3

The over-representation of Black youths in the justice system starts early in the process. Youths without a record of previous contact with the police who were apprehended for a minor offence in a southern Ontario (Canada) region were unlikely to be charged. However, Black youths were more likely to be formally charged than white youths. White youths were more likely to receive a caution.

Canada’s 2003 Youth Criminal Justice Act encourages police not to send youths apprehended by the police to court. Section 6(1) requires police officers to consider whether, before starting judicial proceedings, taking no action or some other action (other than taking the youth to court) would be sufficient. This paper examines whether Black and White youths are equally likely to benefit from these police decisions.

The region in which this study was carried out had created a formal pre-charge youth diversion program for relatively minor offences. Hence, police officers dealing with a youth who had been apprehended for an offence could choose one of three approaches: charge the youth and take them to court, divert the youth to the diversion program, or simply caution the youth. Cautioning the youth was used more than half of the time for both female and male youths. Males were more likely to be charged than females (17% vs. 11% of all cases).

Looking only at those offences that were eligible for the diversion program, it was found that Black youths were more likely than White youths to be formally charged and sent to youth court and less likely to be released with a simple caution.  This same finding – harsher treatment for Black youths than White youths – held when the author looked at specific (high volume) offences – minor theft and possession of marijuana. (The data covered years prior to the legalization of simple possession of marijuana.)

Females apprehended by the police were more likely to be cautioned by the police if they were White rather than Black. Black females were more likely to be sent to the formal diversion program and less likely simply to be cautioned than White females.  Black male youths apprehended by the police for a minor offence were also more likely to be charged and sent to court and were less likely to be let off with a simple caution than were White male youths. Since being cautioned is almost certainly less punitive than being sent to a formal diversion program (and being sent to court is the most punitive outcome), it would seem that Black youths – both male and female – received more punitive outcomes than their White counterparts.

“The police [in Ontario] have the unique power to stop, question, search, investigate, arrest, detain, and monitor anyone they deem a potential criminal… [Their] discretion is often vulnerable to bias” (p. 405). Given numerous other studies showing differential treatment of suspects by the police, these findings are not surprising. However, given that they involve youths in a jurisdiction where the police are urged not to charge any youth who is apprehended for a minor offence (like the offences covered by this study), it is important to discover that Black youths are still treated more harshly. Even if “harsher” in this setting means sending them to a diversion program rather than simply being cautioned and released, the difference matters as it still imposes structured requirements, mandatory attendance, and various behavioural conditions.

Conclusion: Black youths who had no record of being apprehended by the police were treated more harshly than comparable White youths. One consequence of this differential treatment is that Black youths, when first apprehended by the police, are more likely to acquire a record of offending than are White youths. Hence, if they are apprehended again as youths, Black youths will be more likely to look like “repeat offenders” than will White youths, making them especially likely to be treated formally and sent to court.

Reference: Samuels-Wortley, Kanika (2022) Youthful Discretion: Police Selection Bias in Access to Pre-Charge Diversion Programs in Canada. Race & Justice, 12, 387-410.

Item 4

Being convicted of a minor (misdemeanor) offence obviously has fewer consequences than being convicted of a more serious (felony) offence. However, for immigrants, misdemeanor convictions can have very serious consequences, including the risk of deportation. Immigrants are also more likely than non-immigrants to be imprisoned for such offences.

In the US, “there is no group for whom the consequences of misdemeanor arrests are more severe than for non-US citizens” (p. 677). In many ways, being an immigrant in the US is equivalent to having a racialized legal status. 

Prosecutors, for example, are said to be “less apt to offer diversion programs [to undocumented immigrants for minor offences] because they believe that [immigrants] cannot complete the program prior to removal [from the US]” (p. 659). Similarly, it seems that it is acceptable for a judge to suggest that a crime by an immigrant is more serious than the equivalent offence committed by a non-immigrant (p. 660). Not surprisingly, then, “immigration consequences [deportation] are an explicit goal of a conviction in some jurisdictions and immigration officials often encourage prosecutors to seek plea agreements that guarantee removal [from the country]” (p. 661). Finally, it is important to remember that, unlike cases with serious felonies, “police officers are often the sole source of evidence for many minor crimes” and charges are typically “dismissed only after imposing significant burdens on defendants” (p. 658).

This study looks at the criminal justice treatment of immigrants in two states. In Texas, local police participate in immigration enforcement, often with the goal of removal from the country. California is very different. Their local police are banned from cooperating with immigration officials in cases involving minor crimes by apparent immigrants. California also lowered the maximum sentence for many misdemeanors to 364 days’ imprisonment, just under the federal 1-year threshold that can automatically trigger immigration repercussions. The recorded data on those charged with offences in both states include citizenship and whether the accused was foreign born. But only Texas records whether the accused is a legal or illegal immigrant.

In both states, non-US citizens are more likely to be charged, convicted, and incarcerated than are US citizens. When various controls (demographics, arrest information, criminal history) are included in the estimates of the effects of citizenship on criminal justice outcomes, the effects of citizenship on conviction and incarceration still hold, as does the impact of citizenship on arrest in California (the small difference in the same direction is not significant for Texas). In Texas, where non-citizens could be categorized as either lawfully present or undocumented, the data reveal a clear distinction between the two: the undocumented accused people are most likely to be convicted and incarcerated.  “The incarceration difference between citizens and undocumented immigrants is more than double the estimated incarceration gap between White and Black misdemeanants” (p. 675).

Conclusion: The results from these two large, but politically different, US states “reveal that noncitizens, and especially undocumented immigrants, are exposed to disproportionate punishment” (p. 677) when apprehended by the police for possible misdemeanor offending. “Although the punishment of noncitizens disproportionately impacts racial minorities, [the] results suggest that citizenship is a unique and consequential axis of legal inequality, even in local criminal cases that involve only minor offences” (p. 679).

Reference: Light, Michael T., Jason P. Robey, and Juingmyung Kim (2024). Citizenship, Legal Status, and Misdemeanor Justice. Criminology, 62, 655-703.

Item 5

People from poor neighbourhoods who commit offences in nearby wealthy neighbourhoods are less likely to benefit from prosecutorial charge reductions than are people from the same poor neighbourhoods who commit offences in their own neighbourhoods.

Prosecutorial decision-making – which generally is not subject to any appeal or review process – can be an important contributor to inequalities of outcome in criminal courts. This paper looks at the interaction between who the accused is (defined by what kind of neighbourhood they are from) and what kind of person they victimized (defined by the wealth of the neighbourhood where the crime took place).

Plea negotiations provide the site where prosecutorial charge reductions can most clearly be seen and measured. Data from New York City were analyzed for 68,113 cases in which the accused pleaded guilty and information on the defendant’s residence and characteristics was available. These were, in effect, cases in which the prosecutor decided the case outcome.  The focus was on charge reductions: whether the statutory severity of the most serious charge was reduced during the prosecutorial process. The socioeconomic status of neighbourhoods was operationalized by using the median income of the neighbourhood. The analyses focused on low- (median income of $40,900 or less) and high-income (median income of $98,160 or more) neighbourhoods. The results were controlled for various factors (e.g., criminal record of the accused, severity of the offence, age, race, gender).

The study found that 53% of the defendants received a charge reduction. While 64% of the defendants came from low-income neighbourhoods, only 5% were from high-income neighbourhoods. As one might expect from other research, young, male, and Black defendants had lower odds of getting a charge reduction than older, female, and White defendants. Interestingly, however, defendants from low-income neighbourhoods were slightly more likely to receive charge reductions while those arrested in high-income neighbourhoods were less likely to receive charge reductions. 

Charge reduction outcomes also reflect a complex interplay between a defendant’s socioeconomic background and the neighbourhood where the arrest occurred. Defendants from low-income areas who were arrested in high-income areas were less likely to receive charge reductions than those who were arrested in their own or similar (low-income) areas. Those defendants from high-income areas who were arrested in low-income neighbourhoods had the same likelihood of receiving a charge reduction as those arrested anywhere else. In other words, the defendants least likely to receive a charge reduction were those from poor areas charged with offences that took place in wealthy areas.

Prosecutors seem to view those from a low-income area who commit offences in wealthy areas as particularly deserving of harsh penalties. This could be the result of prosecutors’ views that such offences are more serious or (assuming that prosecutors see harsh charges as a way of deterring offending) that this is an effective approach to providing additional protection to residents of high-income neighbourhoods. “Offenders from low-income areas may be perceived as especially dangerous and threatening when they are viewed as ‘out of place’” (p. 883).

Conclusion: The decisions by prosecutors to reduce the severity of the initial charge against a defendant “are not only impacted by ecological stereotypes tied to defendant home neighbourhoods, but also the combination of home and arrest areas, with unique disadvantages for low-income residents who target high-income areas… [These findings] detail the subtle and indirect ways that environmental influences can shape punishment disparities and unintentionally perpetuate social inequality in the criminal legal system” (p. 884).

Reference: Hernandez, Raquel A. and Brian D. Johnson (2024). Traveling to Criminal Opportunity: Defendant Mobility, Socioeconomic Context, and Prosecutorial Charge Reductions. Journal of Research in Crime & Delinquency, 61, 860-899.

Item 6

When the conditions of release from pretrial detention take into account the accused person’s ability to meet these conditions, not only are more people released from pretrial custody, but the accused people are also less likely to be eventually found guilty. They were not more likely to be re-arrested or convicted of a new offence.

Prior to a court decision in 2018, San Francisco’s bail system did not consider an accused person’s ability to pay when determining the amount of money they had to post to be released from custody after arrest. A 2018 court decision required that the accused’s ability to pay be considered when setting the bail amount.

Being detained in custody while awaiting trial obviously reduces the ability of the accused to contribute to their own defence, to show good conduct to the court that is supervising their release, and to participate in treatment programs. Accused people held in custody may also feel pressure to accept unfavourable plea offers. This study looks at the impact of that 2018 court decision on the outcome of cases and the accused’s subsequent behaviour.

To assess the impact on those most likely to be affected by the new (more liberal) policy, accused people who were released after the change in policy were matched on a wide range of variables with those who were charged before the new policy was imposed by the court.

The 2018 court decision reduced the number of people who were detained or released as a result of posting cash bail, and it increased the number of people released on some sort of supervision. In addition, those released after the court decision were less likely to be convicted than those whose cases were decided before the court decision. It would appear that those who attained release as a result of the court-imposed change in policy were “less likely to accept a plea bargain than very similar individuals who were detained prior to [the court decision], controlling for other changes [that might have affected plea bargains]” (p. 137).

A separate analysis examined the impact on subsequent contact with the justice system.  Controlling for various characteristics of the case, there were no differences in the likelihood of re-arrest or a new conviction within 6 months of being released. With people expressing concern about the release of those awaiting trial, this finding is important because increasing the number released did not lead to any increase in crime.  The specific programs available in the jurisdiction (San Francisco) for the supervision and treatment of those awaiting trial may have contributed to this finding.

Conclusion: It would appear that changes in the practice of pretrial detention that resulted in more people being released prior to trial did not affect the likelihood of subsequent contact with the criminal justice system.  Those who were processed under the new provisions (favouring pretrial release) were, however, less likely to be convicted, in large part because they did not accept unattractive plea offers.

Reference: Lacoe, Johanna, Alissa Skog, and Mia Bird (2025). Bail Reform and Pretrial Release: Examining the Implementation of In re Humphrey. Criminology & Public Policy, 24, 123-148.

Item 7

Risk assessments are often carried out on people who are charged with a crime to determine whether they would appear for their required court appearances if they were to be released. A standardized scale that had often been shown to be “valid” for “Americans” was, in this study, shown to be worthless for this purpose when applied to Native American accused people.

When a person is arrested, they are presumed to be innocent and should, therefore, be released until their trial unless it can be shown that they are likely not to appear in court or they are likely to commit a new offence. This paper examines whether a standardized scale (the “Public Safety Assessment”) is as effective for Native Americans as it is for Whites.

There is “no universal threshold of acceptability for the predictive validity of pretrial risk assessments, in terms of strength or practical significance” (p. 399). Hence, the question of whether a risk assessment “works” is often reduced to a relative one: Does it work better than an intuitive guess? Alternatively, does it work as well for one group (e.g., Indigenous people) as for another (White residents of the community).

The Public Safety Assessment uses easily available information (e.g., age, current offence, criminal record, previous failures to appear) to predict whether a person will appear for trial and/or be arrested for a new criminal offence if they are released. In previous research, this scale has been shown to predict both new criminal arrests and failures to appear. However, most of the previous research has been carried out in urban settings, and little of it has compared the value of the scale for predicting the behavior of Indigenous and non-Indigenous samples.

This study was conducted in a largely rural South Dakota (US) county, where Native Americans make up about 10% of the general population and about 50% of the jail population. It examined cases involving 4,570 people who were booked on new charges between 2018 and 2021. Notably, 63% of these individuals were Native American. The study examined two outcome variables: whether the accused person failed to appear in court during the pretrial period and whether the accused person was charged with a new criminal offence.

The measure of accuracy – the AUC (“area under the curve”) – can be thought of as the likelihood that a randomly chosen positive instance (e.g., in this study, a person who actually did fail to appear) would have a higher score than a randomly chosen negative instance (a person who appeared as required). If the scale predicted perfectly, the AUC would be 1.0, whereas if the scale had no predictive value, the AUC would be 0.5 (the likelihood of a positive instance having a higher score being 50%). As shown elsewhere (this issue, Item 2), AUC scores have an important weakness: They can obscure the type of error that is made. Hence, the AUC, like many single measures, is not a fully adequate indicator on its own.

In this study, when looking at “failure to appear,” the overall AUC – pooling all groups – was better than chance (AUC = .55) but not by much. More important was the fact that the AUC for Whites was .64 and for Native Americans was slightly worse than it would be if one flipped a coin (AUC = .48). Given that 8 of the 11 items on the scale related to prior justice system involvement, it is not surprising that the AUCs were slightly higher when “new criminal arrest” was the outcome measure examined. For this measure, the AUC for Native Americans was higher, but still below that for Whites.

Conclusion: The predictions were not very accurate for any group and were clearly worse for Native Americans. Said differently, if this scale were used with Native Americans, they would be more likely to be misclassified than would Whites. These results are not surprising: What they show is that the determinants of a “fail to appear” are different for Indigenous people than are the determinants for White Americans. The lesson is clear: Predictive instruments like this one need to be tested regularly on any group they are used for.

Reference: Zottola, Samantha A., K. Stewart, V. Cloud, L. Hassett & S.L. Desmarais (2024). Predictive Bias in Pretrial Risk Assessment: Application of the Public Safety Assessment in a Native American Population. Law & Human Behavior, 48, 398-414.

Item 8

In many jurisdictions, people have the right to remain silent when asked questions by the police.  Although this may be their legal right, this study demonstrates that when accused people invoke their right to silence, they are perceived, by ordinary citizens, as being more likely to be guilty than if they simply waived their right to silence and denied committing the crime.

The theory behind the right to remain silent when questioned by the police is that accused people do not have to put themselves in a situation in which they might incriminate themselves. Many accused people waive their right to silence because they “believe that their innocence will shine through to investigators and ultimately protect them” (p. 369).

The problem with the “right to remain silent” is that people – those in the criminal justice system and those who might ultimately be deciding the cases (e.g., juries) – may assume that someone who is truly innocent would not invoke this right since they have nothing to hide. Indeed, there is some research that suggests that guilty suspects invoke the right to silence more often than do innocent suspects. It would not be surprising, therefore, if both ordinary citizens and police investigators inferred guilt from the decision of an accused person to remain silent.

This paper reports the results of two experiments: one carried out with undergraduate psychology students and the other with undergraduate and graduate students enrolled in courses related to law enforcement and the criminal justice system. Study participants were given a set of vignettes describing criminal investigations for various serious crimes. The vignettes described the crimes as reported by victims and/or witnesses and were designed to be somewhat ambiguous (so that people could not easily decide that the accused was either guilty or innocent). Each of the three vignettes (aggravated assault, drug-related death, and drive-by shooting) contained information about police questioning of the suspect. The suspect was described as remaining silent throughout the interview with the police, telling the police that they were going to invoke their right to silence, or simply denying they committed the crime. The experimental conditions (silence, invoking the right to silence, denial of guilt) varied such that each vignette was, across the experimental participants, equally likely to be used in each of the three conditions.

The main measures obtained from each participant for each vignette were the perception of whether the suspect was guilty as well as various ratings (e.g., honesty, trustworthiness) of the suspect. In both studies (which differed mainly in the group from which people were recruited), there were no important differences in the responses to accused people who formally invoked their right to silence and those who were completely silent during police questioning.

The important finding was that those who waived their right to silence and simply asserted their innocence were less likely to be seen as guilty and received fewer negative ratings than those who were described as having sat in silence or who explicitly invoked their right to remain silent.

Conclusion: The right to remain silent when questioned by the police clearly has serious costs associated with it: The suspect who implicitly or explicitly invokes the right to remain silent will be seen, by ordinary citizens, as more likely to be guilty than they would be if they waived this right and simply asserted their innocence. The notion, therefore, that it is impermissible to penalize someone for invoking their right to silence sounds better than it actually is in practice.

Reference: Lawrence, Megan L., E.R. Saiter, R.E. Eerdmans, and L. Smalarz (2024). The Miranda Penalty: Inferring Guilt from Suspects’ Silence. Law & Human Behavior, 48, 368-384.

This issue of Criminological Highlights was prepared by Anthony Doob, Rosemary Gartner, Maria Jung, Tyler King, Jihyun Kwon, Jane Sprott, and Danielle Van Wagner