Criminological Highlights Vol. 19, No. 3 – April 2021

View this issue as a PDF: PDF iconCrimHighlightsV19N3.pdf

This issue of Criminological Highlights addresses the following questions:

  1. Does the public support the expunging of criminal records?
  2. Is public support for increased police budgets associated with the presence of immigrants in a community?
  3. Do high jail populations contribute to deaths in the community?
  4. When imprisonment rates increase, is it safe to assume that sentencing practices have changed?
  5. Can an intensive 1-day training program change police stop-and-search practices?
  6. Does the diversion of low-risk offenders contribute to the safety and economic well-being of a community?
  7. Should standardized ‘risk’ assessments of probationers allow overrides?
  8. Why has youth crime decreased in the US?

Public support in the US for the expungement of criminal records is growing, especially for those who have been crime free for a reasonable time and who appear to be unlikely to reoffend.

When ‘pardon’ legislation (the Criminal Records Act) was introduced into the Canadian Parliament in 1970 by a Liberal government, a leading Conservative argued that the “main purpose of this bill is not to coddle criminals. … This criminal record has hounded them for years and has created economic difficulties. We are now making an honest attempt to eliminate these difficulties as much as possible” (see the Criminological Highlights collection Evaluating the Benefits of Pardons on our website).  This paper explores public support in the US for procedures that make criminal records inaccessible so that people, at some point after completing their sentences, can live a life outside of the shadow of a criminal record.

It is estimated that perhaps as many as one in three Americans have some form of criminal record (arrest or conviction). Criminal records can affect one’s ability to vote, get a job or housing, or obtain admission to college.  Certain groups (e.g., Black Americans) are much more likely to have a criminal record than others. And in an internet age, criminal records are more easily accessible than they once were.

This paper presents the results of a survey examining American support for criminal record expungement.  The perception of risk appears to be an important factor in understanding whether a person supports the expungement of records (e.g., Criminological Highlights 18(6)#7).  “Risk” can be operationalized in a number of ways, among them being the length of time a person has lived in the community crime free subsequent to criminal justice involvement.  In this survey, only 45% of respondents thought expunging criminal records was, as a general rule, a good way of allowing people to get their lives back on track.   However, 78% thought that people who had been crime free for 10 years should be eligible to have their criminal records expunged.  Probably because they incorrectly believed that “once a sex offender, always a sex offender” (see the Criminological Highlights collection on Sex Offenders and Society’s Responses to Them), many people thought certain types of sex offenders should never be able to get their criminal records expunged. Many fewer wanted to exclude all other types of offender from being eligible for record expungement.  Being employed and having completed a rehabilitation program appeared to be ‘signals’ that people looked for that persuaded them a person was ready to have their criminal record expunged.

Broad support for the expungement of criminal records seemed to be most prevalent among those who believed that people who had done bad things were redeemable and that people could be rehabilitated.  What is perhaps more interesting from a political perspective is that broad support for expungement of criminal records was not related to political orientation, gender, age, race, education or income.

Conclusion:  Rules for expunging criminal records vary enormously across jurisdictions.  In Pennsylvania, for example, those convicted of non-violent crimes more than 10 years earlier automatically get their criminal records suppressed as a result of a recent law that received almost unanimous support in the legislature.  The one-time cost (for software development) to implement this change was about $245,000.  Given the number of records cleared by this method, this means that the cost per person of having a record expunged is about 9 cents.  The survey reported here demonstrates that “Americans are clearly supportive of expunging the records of nonviolent offenders who have remained crime free for a decade.”  In Pennsylvania the cost is one US dime per person.  In Canada the cost to a single applicant for a ‘record suspension’ recently increased to C$657.77.

Reference: Burton, Alexander L, Francis T. Cullen, Justin T. Pickett, Velmer S. Burton, Jr. and Angela J. Thielo (2021).  Beyond the Eternal Criminal Record: Public Support for Expungement.  Criminology & Public Policy, 20, 123-151.

Back to top

The number of new immigrants coming into a country is associated with citizens wanting more funds spent on law enforcement.  The number of foreign-born people already in the community was not an important predictor of support for law enforcement.

Former US president Donald Trump often blamed new immigrants for crime problems.  He appeared to differentiate between the flow of new immigrants (whom he tried to exclude) and the stock of foreign-born residents already in the US (like his wife) whom he appeared not to fear and whose votes he wanted. This paper addresses the role that immigration plays in eliciting support for increased law enforcement funding.

The number of new immigrants has increased in many rich democracies, though there are a few countries (e.g., Poland and Japan) that continue to have small immigrant populations. “In response to rising immigration in Western Europe, animosity toward foreign born residents rose to historically high levels” (p. 1075).  Immigrants are often seen as being less law-abiding, though the existing data would suggest the opposite: areas with large numbers of immigrants tend to have lower crime rates (e.g., Criminological Highlights 18(5)#1).

If, contrary to existing evidence, immigrants are blamed for crime, it would not be surprising that native born people would favour more law enforcement spending if the number of immigrants was increasing. To the extent that “punishment practices [are] at least partly shaped by public opinion” (p. 1076), it is important to understand this relationship better.  According to minority threat theory, “Minority threat is triggered by fear and suspicion that the subordinate group intends to undermine the dominant group’s privilege and advantage” (p. 1077).  It is further suggested that these perceptions may not be driven by actual experiences but by “constructed subjective feelings, fears, and images” (p. 1079).  Prejudice and threat, then, may relate to change – in this case the “flow” of immigrants to a community, rather than the “stock” of immigrants already there.  New immigrants may be seen as a threat; foreign-born neighbours are not.

Residents of 25 rich democracies that participated in the 1996 and 2006 International Social Survey Programme (including Canada, the US, South Korea, Japan, Australia, New Zealand, Israel, and 18 European countries) were asked whether they thought their government should spend more on the police and law enforcement. Using immigration data from 1985, 1995, and 2005 allowed the researchers to estimate the size of the “stock” of immigrants for the year prior to each survey. In addition, it allowed for an estimate to be made of the change in the percent foreign born in the previous decade.  Homicide rate and number of police officers per 100,000 residents were used as control variables at the country level.  At the individual level, various demographic variables were included in the analysis including age, sex, rural/urban residence, political orientation, religion, and employment.

Looking first at the country-level data, there was a very small (r=0.10) correlation between % foreign born and preference for higher law enforcement spending.  However, the correlation between the change in the percent foreign born (in 2006) was much greater (r=0.39).  Including the control factors and looking at change within country, % foreign born (the “stock” of immigrants) did not predict preference for more law enforcement spending.  However, both the change in net migration and the change in the percent foreign born did predict preference for more law enforcement spending. 

Conclusion: In understanding preference for more police and law enforcement, it appears that the increase in the number of immigrants is important, not the number of immigrants already in the country,  Given that the international data suggest immigration is not related to actual crime, it would appear that “fear of crime depends more on unfounded fears and prejudices resulting from changing ethnic heterogeneity than from reasonable concerns about safety” (p. 1095).

Reference: Fink, Joshua J. and David Brady (2020). Immigration and Preferences for Greater Law Enforcement Spending in Rich Democracies.  Social Forces, 98(3), 1074-1111.

Back to top

Locking people up in local jails increases the death rates in those counties in which the jails are located. The findings give further support to the hypothesis that jails are associated with the spread of infectious diseases and other health problems. 

Previous research has demonstrated a link between increases in county rates of incarceration and overall mortality rates in that county.  This should not be terribly surprising. “For example, during the COVID-19 pandemic, jails [in the US] exacerbated the spread of the virus in communities that were already at a disproportionately higher risk of infection” (p.2).  This study builds on previous research by separating out causes for death that would be expected to relate to short- vs. long-term effects of high or increased rates of incarceration in local jails.

The study uses longitudinal jail population data from 1,094 US counties for the period 1987-2017.  These were combined with mortality data for the period 1988-2018 for common causes of death of people who are less than 75 years old.  Various sociodemographic measures like the age and racial makeup of the country, poverty, unemployment and crime rates, and state incarceration rates were used as control factors in analyses linking releases from jail to deaths from these nine causes.

The main variable of interest was average daily jail population in the county divided by the county population. The outcome measures were the 7 most common causes of death, expressed as rates, for those under age 75.

When looking at the short-term relationship between change in the jail population and mortality rates in the county in which the jail was located, “changes in the county jail incarceration rate were most strongly associated with deaths caused by infectious disease, chronic lower respiratory disease, substance use, and suicide” among those age 75 and younger (p. 5).  Deaths associated with these causes tended to increase when jail populations were increasing.

“Associations between county jail incarceration and cause-specific mortality rates weakened for all causes of death as the lag [between the change in jail incarceration and the death rate] increased from 1 year to 10 years” (p. 5).  The decrease in the size of the association was particularly large for causes of death with generally lower latency (e.g., infectious diseases and suicide).   Causes of death like heart disease and cerebrovascular disease – that have longer latencies, generally, between the cause and the effect – showed less pronounced declines.

Conclusion:  The results – showing strong short-term associations between increased jail incarceration and infectious disease, suicide, substance abuse and chronic lower respiratory disease -- is consistent with the finding that high jail incarceration rates are associated with infectious disease spread and acute psychosocial and economic resource deprivation. It should be kept in mind that these death rates reflect deaths in the community, not in the jails.  Hence “these findings suggest that interventions to reduce incarceration might result in broader community mortality benefits….  Given the ongoing COVID-19 pandemic, these findings highlight the immediate need to account for jails as drivers of infectious disease spread in the community” (p. 8).

Reference: Kajeepeta, Sandahya, Pia M. Mauro, Katherine M. Keyes, Abdulrahman M. El-Sayed, Carol G. Rutherford, and Seth J. Prins (2021). Association Between County Jail Incarceration and Cause-Specific County Mortality in the USA, 1987-2017: A Retrospective, Longitudinal Study. Lancet Public Health, 23 February 2021.

Back to top

Judges may not be solely responsible for increases in imprisonment rates.  A study of sentencing, bail, policing policy and crime suggests that changes in sentencing may be almost irrelevant in explaining a 48% increase in the New South Wales (Australia) imprisonment rate. 

In most Australian states and territories, the prison population increased dramatically between 2000 and 2019.  Generally speaking, the rate of increase was larger in the second decade of the 21st century than in the first.   This paper analyzes the reason for this increase in Australia’s most populous state, New South Wales.

The growth in New South Wales was dramatic: from an average of 21,714 prisoners in 2000 to 43,028 prisoners in 2019.   Part of the increase in the prison population was due to a disproportionate increase in the unsentenced (remand) population.  This, in turn, may have been the result of a large increase in the number of people proceeded against in court: an increase of 22% between 2011 and 2019. Bail refusal rates did not appear to account for a measurable amount of the increase in prison populations.  However, the length of time in remand increased for those who were eventually acquitted or given a community sentence.  

One change that would appear, at first glance, to implicate judges as a cause of the increase in imprisonment rates is an increase in the percent of those convicted sentenced to prison from 7.6% in 2011 to 9.8% in 2018.  However, this would appear to relate to characteristics of the cases coming to court.  For example, there were large increases between 2011 and 2018 in the percent of cases that were sexual assault and related offences, abduction/harassment, weapons offences, administration of justice, and drug offences.   Similarly, the percent of those with a prior prison sentence increased from 17.5% of cases to 23.4% of cases.  Multiple indictable offences increased from 22% of all cases to 27%.  All of these case characteristics would be likely to lead to an increased use of imprisonment. 

The likelihood of receiving a prison sentence increased dramatically when the comparison between 2012 and 2018 was done without controls for the nature of the offence or the offender.  However, when controls for case characteristics were made, there was no difference in the odds of receiving a prison sentence, suggesting that these case characteristics explained the increase.  Sentence length and the time that a prisoner was required to serve before being eligible for parole did not change. 

It is possible that some of the increase in imprisonment relates to an increase in drug arrests related to amphetamines.  However, there is no indication of increases in the incidence of other crimes that would account for changes in the nature of cases in court.  Hence, the change in the nature of cases in court seems to relate more closely to changes in the decisions to charge. For example, there was a very large increase in the number of people charged for breaches of community-based orders. 

Conclusion: “The evidence [in this paper] suggests that the increase in the New South Wales prison population since 2011 is largely a consequence of changes in policing policy and (to a lesser extent) crime, rather than changes in penal policy” (p. 576).  Once the changes in the nature of the cases before the courts is taken into account, there was no evidence of an increase in the punitiveness of the courts.  More generally, the data provide a reminder to those concerned with either increases in imprisonment or with over-representation of certain groups, that the placement of a person in prison is the result of multiple decisions and policies, not just the sentencing decision of the judge.

Reference:  Weatherburn, Don (2020).  Is Tougher Sentencing and Bail Policy the Cause of Rising Imprisonment Rates?  A NSW Case Study. Australian & New Zealand Journal of Criminology, 53(4), 563-584.

Back to top

A 1-day training program for police focusing on ethnic/racial bias in stop and search powers was successful on paper, but not in changing the actual behaviour of police on the street. 

In the United Kingdom, as in many other countries, there is concern about ethnic and racial bias in policing.  One of the most frequent specific concerns relates to discretionary police stops and searches.  This paper examines a police training program that “sought to strengthen the foundations of officers’ use of stop and search, and focused on topics including unconscious bias, reasonable grounds for suspicion, and strategies for conducting stop and search encounters” (p. 1260).  Unlike many such training programs, this one was carefully evaluated. 

Police stops have been shown in many countries to focus disproportionately on certain groups – most notably Black people.  In England & Wales, for example, people who identified as Black were about 9 times as likely to be searched as were those who identified themselves as White. As a result of findings such as these, the UK College of Policing (the professional body for policing) developed a program to address concerns about police bias. 

The program – the main part of which involved a 6-hour classroom training session – was designed to include material to make police officers aware of the history and concerns about stop-and-search practices. It also gave training on how to assess reasonable grounds for suspicion, recording requirements, and how a stop-and-search should be carried out. Essentially there was a “competence” component focusing on the necessary skills to determine whether a stop was warranted and an “awareness” component focusing on the issue of bias. 

The implementation was carried out on an experimental basis in six UK police agencies. Individuals within each agency were chosen who had carried out searches in 5 of the previous 12 months and remained in a role in the police service where they might be expected to use stop and search practices.  662 police officers were assigned to the treatment group and 661 to the control group. Questionnaires about stop and search practices were distributed 6 to 54 days after the training.

The results on the questionnaire demonstrated that the treatment increased knowledge of stop and search regulations. In addition, the questionnaire results suggested that there was no significant difference in the likelihood of initiating stops of suspicious people between those who had or had not received the special training.  However, those who had received the training indicated that they would be less likely to proceed and carry out a search, especially if the evidence looked weak. 

In line with national trends in searches, there was a decline in the number of searches carried out in the three months after the training, compared to the three months before. However, the size of this decline was unaffected by the treatment.  In other words, those who received the special training on stop-and-search were just as likely to carry out searches as those who did not receive the treatment.  The training had no reliable impact on the likelihood of searches of minority groups. 

Conclusion: “Overall, the results provide no evidence of any training effects on officers’ recorded street-level practices…” (p. 1277).  In certain ways, this finding is not surprising. The actual training was only 6 hours long.  There was also some variation in the training that was given in an attempt to take into account local conditions. It may be that by focusing exclusively on front line officers without much attention to supervisors and managers and not “embedding [the training] in broader multifaceted reform efforts”, this ‘one shot’ training exercise did not have much of a chance to affect actual practice in the long term. Changing a practice that is embedded in an institutional culture may not be successful unless it is part of a much larger effort.

Reference: Miller, Joel, Paul Quinton, Banos Alexandrou, and Daniel Packham (2020). Can Police Training Reduce Ethnic/Racial Disparities in Stop and Search?   Evidence from a Multisite UK Trial.  Criminology & Public Policy, 19, 1259-1287.

Back to top

Two abrupt changes in the use of diversion in Texas (one decreasing the use if diversion from the criminal courts, the other increasing its use) demonstrate that diverting people from the criminal justice system is good policy.  Dropping the charges against those charged with low level felonies is associated with lower rates of offending in the subsequent 10 years.

In many jurisdictions, many of those who are charged with criminal offences have all charges dropped, often in return for some form of non-criminal justice punishment.  This paper examines the impact on crime of two abrupt changes in the law in Texas related to diversion.  

In 1994, a change in the law made the prosecution of low-level felony drug and property crimes suddenly more likely (and diversion less likely).  In 2007, a failed initiative, that had been expected to pass and that would have increased jail capacity in Harris County, Texas, forced officials to divert more first-time felony defendants so as to avoid jail over-crowding.  Both of these changes were abrupt and, to a large extent, unanticipated.  The study looks at first-time felony defendants charged 2 years before and 2 years after the 1994 change and 2 years before and 7 years after the 2007 event.  The focus is on those most likely to have been affected by the changes that took place. 

The people being brought to court before and after each of these changes were very similar. For example, an overall distribution of the “predicted recidivism risk” score for those before and after the changes that took place were very similar.  What changed was that there was a substantial decrease in the use of diversion in 1994, and a substantial increase in the use of diversion in 2007.  The study then looked at the conviction rates of first-time felony defendants charged before and after each of these two changes.  The characteristics of the cases coming to court did not change. What changed was the likelihood that charges would be dropped against those charged with relatively minor offences. 

The impact of diversion in each of these natural experiments was the same: there was “strong and consistent evidence that diversion leads to statistically significant and economically meaningful declines in recidivism…. Diversion decreases reconviction rates by slightly more than 30 percentage points in the 1994 sample and by 26 percentage points in the 2007 sample…” (p. 17).   Diversion was also associated with increased employment, including employment stable enough to increase the likelihood that a person’s income exceeded the federal poverty level. 

For the group who benefitted from diversion – largely those without prior felony convictions – it appeared that there was a substantial benefit in finding employment.  This is not very surprising given the challenges that any criminal record has on the likelihood of finding employment (see, Criminological Highlights 6(3)#2, 11(4)#4, 17(2)#7). In this study of policy changes in 1994 and 2007, people who were diverted rarely had a prior felony conviction.

Conclusion: This paper demonstrates that diversion of low-level defendants from prosecution increases the likelihood of employment and decreases the likelihood of future offending. It could be argued, therefore, that the paper provides “compelling evidence on a successful intervention [diversion] that both improves defendant outcomes and saves public resources…. [Furthermore] it can be feasibly implemented without significant investments or changes to current infrastructure, making it a practical solution for criminal justice reform” (p. 33).

Reference: Mueller-Smith, Michael and Kevin T. Schnepel. (2021) Diversion in the Criminal Justice System. Review of Economics Studies, 88(2) [Pre-publication pages 1-54] 

Back to top

Overrides of a risk assessment scale used by US probation officers decreased the already mediocre performance of a commonly used risk assessment tool that was designed to determine the appropriate level of intensity of supervision of federal probationers.    

It is tempting to think that it is possible to improve, systematically, the predictions made by a risk assessment scale. If a probation officer, who ostensibly knows a probationer who is being supervised, believes that the risk scale is not giving proper weight to relevant factors, shouldn’t the probation officer override the purely statistical prediction? This paper examines what happens to the accuracy of the prediction when those supervising federal probationers are allowed to override the statistical prediction instrument.

In the US federal system, risk predictions are made with the Post-Conviction Risk Assessment instrument, a 15-item scale using both static and dynamic risk factors.  Probation officers have the explicit power to place supervisees into higher or lower risk categories than that recommended by the scale, though overrides are supposed to be rare.

The study looks at the success in predicting rearrest for a felony or misdemeanor offence within 24 months, but excludes arrests for technical violations.  The study uses data from 259,571 people being supervised on probation in 94 federal judicial districts.  Given that they were on probation for US federal offences (not state offences), it is not surprising that the largest categories of offenders were drug (44%), financial offense (20%), and weapons offense (15%).  The study focuses on the 27,480 overrides (related to about 11% of supervisees), 99% of which moved the supervisee to a higher risk category. Sex offenders constituted 3.1% of the total sample of supervisees; however, 45% of the over-rides were for sex offenders, all of whom received increased risk levels.  The original (scale-determined) risk values placed only 20% of those sex offenders whose risk assessment were subsequently raised into a moderate- or high-risk category.  After the ‘adjustment’, over 98% ended up in a moderate- or high-risk category.   Although this was the group most likely to have their risk scores dramatically altered because of the discretionary decisions of probation supervisors, similar increased perceived risk scores were assigned to other groups of probationers.

Overall, 15% of this sample were rearrested for an offence within 2 years. The policy question addressed by the study is simple: Was the accuracy of the “adjusted” risk measure (adjusted as a result of the supervisor’s override) higher than the accuracy the original ‘scale’ value?  For all groups of offenders – most notably sex offenders – the original scale value predicted better than the ‘adjusted’ risk value. 

One can divide the sample into two groups – low or low/moderate risk (low) vs. moderate or high risk (high).  Using the original scale value, 12% of the low risk group were rearrested and 28% of the high risk group were rearrested.  After the risk scores had been adjusted by the probation officer, about 8% of the low risk group and 17% of the high risk group was arrested for a new offence.   Large numbers of low/moderate risk people (on the original scale) were moved into the high risk category but they did not reoffend. An index of overall accuracy of the predictions showed that prediction accuracy declined as a result of the overrides.

Conclusion:  The results show that subjective modifications of an ordinary “risk” scale make risk prediction worse, rather than better.  In this case, almost all of the modifications increased the “risk” of the offender being supervised.  The result is that the number of false positives – a prediction that offenders would reoffend when, in fact, they do not – increased from about 14% of the sample to about 66% of the sample.   By re-classifying large number of people as ‘high risk’ when they were not, it was almost inevitable that accuracy of prediction would decrease. At the same time, one cannot ignore another important finding: within this group of people whose risk scores were altered by supervisors but who had a relatively low likelihood of rearrest (15%), even the original scale scores did not predict very well.   

Reference: Cohen, Thomas H., Christopher T. Lowenkamp, Kristin Bechtel, and Anthony W. Flores (2020).  Risk Assessment Overrides: Shuffling the Risk Deck Without Any Improvement in Prediction.   Criminal Justice & Behavior, 47(12), 1609-1629.

Back to top

Since the early 1990s, the prevalence of youth crime in the US has decreased dramatically. It appears this is most likely to be the result of reductions in two risk factors for youth crime: the amount of unstructured socializing and alcohol consumption.  

Reductions in youth crime have apparently occurred in the 21st century in a number of countries. In this US study, between 1991 and 2015 there was a reduction in self-reported youth crime of 35%. In Canada, between 1998 and 2015 there was a 37% reduction in crime incidents reported to the police (see Criminological Highlights 18(3)#2).  

The substantial drop in offending might be a result of any of three types of changes in youths’ experiences: (a) increased social control, (b) decreases in exposure to settings where crime is likely to take place, and/or (c) reductions in criminal propensities or motivations to be involved in crime.

This study uses data from a study in which a nationally representative sample of Grade 8 & 10 students from over 100 public and private schools were surveyed and asked questions on involvement in crime. The main dependent variable – prevalence of involvement in crime – was calculated based on the number of different offences (e.g., hurt someone badly, took something of value) the youth admitted to doing in the previous 12 months. 

In addition, a range of possible explanatory measures were assessed.  These included the amount of unstructured socializing, alcohol use, employment of the youth, parental supervision of the youth, attachment to school, as well as the educational expectations, involvement in the community, and preferences for risky or dangerous activities.  Various other control variables (e.g., sex, race, whether the youth was from a single-parent household, parental education) were also included. 

Over time (in models that included the control variables), the prevalence of youth offending decreased by about 35%.  However, when the potential explanatory variables were included in the model, changes in those explanatory variables appeared to account for about a third of the decrease in the prevalence of involvement in crime. When comparing 1991-1995 with 2011-2015, the data suggested that the two largest changes that accounted for the decrease in crime were a decrease in unstructured socializing and a decrease in the use of alcohol.  Considerably less of the decrease in crime was accounted for by changes in sensation seeking. 

Conclusion: The “results show that the decreases in unstructured socializing and alcohol consumption were significantly associated with reductions in youth delinquency between 1991 and 2015” perhaps because “contemporary youth spend much less time away from home with peers in contexts that provide limited supervision and plentiful opportunities for deviance, including alcohol consumption” (p. 129).  Hence “the results suggest an important role for reductions in unstructured socializing and alcohol consumption” (p. 130) in understanding the prevalence of youth crime in the US.

Reference: Baumer, Eric P., Kelsey Cundiff, and Liying Luo (2021).  The Contemporary Transformation of American Youth: An Analysis of Change in the Prevalence of Delinquency, 1991-2015.  Criminology, 59, 109-136.

Back to top

Publication Type

Volume Number


Issue Number