A comprehensive study of recidivism rates among canadian federal offenders: August 2019

Lynn A Stewart
Geoff Wilton
Sebastian Baglole
&
Ryan Miller
Correctional Service of Canada
August 2019

A large-scale project like this cannot be completed without benefiting from the assistance of many individuals. We wish to acknowledge the timely and conscientious work of Chelsea Sheahan, Kaitlyn Wardrop, Angie-Lee Costeira, and Kate Pardoel in coding and additional data analysis. Chelsea de Moor assisted by developing an annotated bibliography. Tanya Rugge, Guy Bourgon, and Larry Motiuk provided valued advice on the initial methodology. Our thanks to Larry Motiuk, Bruno Jean, Mark Olver, Leslie Anne Keown, Shanna Farrell MacDonald, Kelly Babchishin, and staff at Correctional Operations and Programs and Aboriginal Initiatives who provided feedback on the draft.

Key words: recidivism rates, federal offenders, women offender recidivism, Indigenous offenders recidivism by age bands, recidivism by reoffence type

This study provides a comprehensive estimate of recidivism rates of federally sentenced offenders based on reconvictions that resulted in returns to federal custody or in provincial or territorial sanctions. Using official records from the Canadian Police Information Centre (CPIC) database combined with data from the Offender Management System (OMS), we examined the rates of reoffending of any kind and violent reoffending over various follow-up periods for the following groups: all men, all women, Indigenous men and women offenders, Black offenders, and South East Asian offenders. A five year annual release cohort from the 2007/2008 through 2011/2012 was identified (N = 22,685). Coding of the CPIC records was based on offenders’ release on their first term for two cohorts: one released in 2007/2008, and the second released in 2011/2012 (total N = 8441). Additional analyses examined the rates of recidivism by region, index offence, reoffence type, age, and citizenship.

The most commonly reported outcome reported internationally is based on reconviction rates within two years of release. Using this measure, the overall two year recidivism rate for the 2011-2012 cohort was 23%; the rate was 24% for men and 12% for women. Recidivism of Indigenous offenders was higher, 38% for Indigenous men and 20% for Indigenous women. The violent reoffending rate two year post-release in the total 2011-2012 cohort was 12%. These rates were lower than for offenders released in 2007-2008 and both cohorts had lower rates than those quoted in the previous report on recidivism of federal offenders released in 1996-1997 (Bonta et al., 2003) which found a two year reconviction rate of 41% for all offenders, and 16% for women. In the current study, results based on OMS data indicated that reconviction rates declined each year from 2007-2008 to 2011-2012. These outcomes provide an encouraging indication that reoffending of federal offenders has decreased steadily over time. Analysis of the crime severity on the admitting offence compared to the reoffence while offenders were under federal warrant demonstrated that 61% of recidivists reoffended with an offence of lesser severity.

Longer follow-up periods produced higher recidivism rates. Nearly 38% of all federal offenders released in 2011-2012 reoffended within 5 years and about 60% of Indigenous men reoffended in this time period. The most frequent reoffence category for men and women was ‘other non-violent’ and property offences. Offenders with a property or robbery index offence had the highest rates of reoffending and those with an index offence of robbery had the highest rates of violent reoffending. Younger adult offenders (under 25) had the highest rates of recidivism and violent recidivism, rates of violent recidivism were over six times than that of offenders over age 55.

There was regional variation in recidivism rates. While offenders were under warrant, the Prairies and Pacific regions had the highest rates of returns to federal custody with an offence, and the Ontario and Quebec regions, the lowest. Pacific had the highest rates of new warrants of committal across all follow-up periods.

A standardized measure of recidivism provides a benchmark by which to evaluate an agency’s effectiveness in facilitating crime reduction over time. Common agreement on key measures of recidivism would facilitate the examination of factors that contribute to differences in recidivism some of which could be related to policies and interventions that could be adopted more broadly.

• Acknowledgements
• Executive Summary
• List of Tables
• List of Figures
• Introduction
  • Measures of recidivism internationally and in Canada
• Method
  • Participants
  • Procedure/Analytic Approach
  • Measures/Material
• Results
  • Reoffending and violent reoffending rates (combined OMS returns with an offence and CPIC reconviction records)
  • Rates of return to federal custody
  • Rates of return to federal custody pre warrant expiry (revocation rates)
  • Rates of return to federal custody post-warrant expiry (new warrants of committal)
  • Rates of reoffending post-WED based on CPIC records
  • Recidivism Rates by Index Offence Type
  • Results by Crime Severity Index
  • Recidivism by Region of Release
    • Revocation Rates Pre-Warrant Expiry
    • Return to federal custody post warrant expiry (new warrants of committal) by region of release
  • Recidivism Results by Age of Release
  • Results by Self-Identified Ethnic Group
  • Results by Citizenship
• Discussion
  • Limitations/Considerations
  • Conclusions
• References
• Appendix A: Recidivism Study Coding Guide
• Appendix B: Additional Tables

• Table 1 Definitions of Recidivism and Follow-Up Periods
• Table 2 Released Offenders for the 2007/2008 to 2011/2012 Cohorts by Gender and Ethnic Group
• Table 3 Rates of Reoffending of Any Kind (Combined CPIC and Returns to Federal Custody with an Offence)
• Table 4 Rates of Violent Reoffending (Combined CPIC and Returns to Federal Custody with an Offence)
• Table 5 Revocation Rates by Fiscal Year of Release (N = 21,915)
• Table 6 Rates of Reoffending Post Warrant Expiry Resulting in a New Warrant of Committal (Return to Federal Custody (N = 22,038)
• Table 7 Rates of Reoffending Based on New Warrants of Committal (Returns to Federal Custody)
• Table 8 Rates of Reoffending Post WED Based on CPIC Records
• Table 9 Rates of Reoffending by Index Offence (Based on OMS and CPIC Records)
• Table 10 Rates of Violent Reoffending by Index Offence (Based on OMS and CPIC Records)
• Table 11 Mean Crime Severity Index Weights of Index and Reoffence by Offender Group
• Table 12 Rates of Reoffending by Region (Based on OMS and CPIC records)
• Table 13 Rates of Violent Reoffending by Region (Based on OMS and CPIC records)
• Table 14 Rates of Revocations by Region of Release*: Pre Warrant Expiry (N = 21,911)
• Table 15 New Warrant of Committals within Fixed Follow-Up Time Periods by Region of Release
• Table 16 Rate of Reoffending Post WED Based on CPIC Records by Region of Release
• Table 17 Rates of Reoffending by Age Bands (Combined OMS and CPIC Records)
• Table 18 Violent Reoffending Rates by Age of Release (Combined OMS and CPIC Records)
• Table 19 Rates of Reoffending Post Release by Ethnic Group (Combined OMS and CPIC Records)
• Table 20 Rates of Violent Reoffending Post Release by Ethnic Group (Combined OMS and CPIC Records)
• Table 21 Rates of Revocations by Canadian Citizenship Based on CPIC and OMS Records (N = 21,911)
• Table B1 Revocation Rates within Six Months (N = 21,766)
• Table B2 Revocations by Index Offence (N = 21,889)
• Table B3 Rates of Reoffending and Returning to Federal Custody (New Warrants of Committal) by Index Offence (N = 22,012)
• Table B4 Reoffending Rates Based on CPIC Records (Post WED) by Index Offence
• Table B5 Rate of Reconvictions Based on New Warrants of Committal within Fixed Follow-Up Time Periods by Index Offence
• Table B6 Revocations by Age Group at Release: OMS data. (N = 21,911)
• Table B7 Revocations with an Offence by Most Serious Reoffence Type (N = 21,911)
• Table B8 Rates of New Warrant of Committals by Most Serious Reoffence Type (N = 22,038)
• Table B9 Reoffending Rates by Reoffence Category (CPIC records) (n = 8,439)
• Table B10 Rates of Revocation by Self-Identified Ethnic Group: OMS Data (N=21,911)
• Table B11 Rates of New Warrants of Committal by Self-Identified Ethnic Group (N = 22,038)
• Table B12 Rates of New Warrants of Committal within Fixed Follow-Up Periods by Self-Identified Ethnic Group
• Table B13 Rates of Reconviction Based on CPIC Records (Post WED) for Various Follow-Up Periods by Ethnic Group
• Table B14 Cox Regression Models Predicting Revocations with an Offence by Indigenous Background and Age at Release for Men and Women
• Table B15 Cox Regression Models Predicting Revocations with an Offence by Indigenous Background, Age at Release and the Interaction between the Two for Men and Women

• Figure 1. Rates of any new reoffence 2 years post release (rate of violent reoffending 2 years post release in brackets)
• Figure 2. Revocation rates by fiscal year of release for all federal offenders
• Figure 3. Rates of any reoffending post WED based on CPIC records for two release cohorts: Two year fixed follow-up

Recidivism is commonly defined as an individual’s return to criminal behaviour after receiving a sanction or intervention for previous criminal behaviour (Saris et al., 2016). Recidivism rates serve as the key indicator of the impact of correctional interventions and sanctions, providing direction on the efficacy of correctional practices. Despite its importance to the criminal justice system, there is no consensus on how recidivism should be reported. Definitions, measurements and reporting practices vary across constituencies. A large scale systematic review of recidivism rates worldwide determined that the approach to reporting on recidivism is so variable that the data cannot be considered valid for international comparisons (Fazel & Wolf, 2015). In the last study examining recidivism rates of federal offenders in Canada, Bonta, Rugge, and Dauvergne (2003) provided a framework for the reporting of recidivism specifying that the following four factors should be considered: 1) the definition of return to crime and the data sources used to obtain this information, 2) the nature or type of recidivism, 3) the follow-up period, and 4) variation in the study sample. These recommendations are similar to those previously specified by a White Paper commissioned by the Council of Juvenile Justice Administrators in the US (Harris, Lockwood, & Mengers, 2009) and by Ministry of Justice in Netherlands (Wartna & Nijssen, 2000).

Common definitions of recidivism include new arrests, new convictions, and reincarceration, each providing important information and varying estimates of rates of offenders’ return to crime. For example, arrest data may provide the most accurate details regarding the date and time of the offence, but can overestimate recidivism rates by including individuals who were arrested, but not convicted, of an offence (Bonta et al., 2003; Payne, 2007). New convictions can result in incarceration, but may underestimate recidivism by excluding criminal behaviours where there was insufficient evidence to convict, or when the adjudication results in a plea bargain. Reincarceration may over-represent offenders who commit more serious crimes and have more extensive criminal histories (Bonta et al., 2003; Dunrose, Cooper, & Snyder, 2014; Payne, 2007).

Given that recidivism is usually operationalized by using official counts of criminal behaviour such as new arrests or new convictions, administrative data serve as the main source for this information. Some researchers argue that the use of administrative data may in itself underestimate rates of recidivism by excluding crimes that go undetected by law enforcement (Bonta et al., 2003). In Canada, metrics that examine only returns to federal custody before or after warrant expiry dates do not consider the possibility that a reoffence after warrant expiry could have earned a provincially, or territorially, administered sanction and therefore the calculations will result in an under-estimation of actual recidivism rates.

In the Correctional Service of Canada (CSC), research examining recidivism has typically used returns to federal custody as the key outcome measure. As pointed out by Bonta and colleagues (2003), this allows the Service to examine its impact on criminal behaviour during the period that the offender is under CSC’s direct responsibility. However, it can be argued that, if effective, the impact of programs and sanctions during the time offenders are under warrant should persist after the penalty expires. What is more, the Canadian public is less likely to make the distinction on whether the reoffence earns a provincial or federal sentence. Recently, the Auditor General called for CSC and the Public Safety Ministry to provide standardized statistics on the rates of recidivism that include not only returns to federal custody but also reoffending that results in provincial and territorial sanctions (Office of the Auditor General, 2018). This report is one step in meeting that obligation but in the longer term a more efficient way to integrate provincial and territorial sanctions into the production of ongoing recidivism rate estimates is required.

Various follow-up time frames will produce different results; the longer the follow-up period or time at risk, the higher the recidivism rate. Depending on how the administrative data are collected and from which source, a fixed follow-up period can distort measures of reoffending by restricting the sample to offenders who can be followed for the time period. For example, in Canada, federal offenders who are released late in their sentences (generally the higher risk offenders) may have shorter periods of supervision until their warrant expiry dates, and therefore may not be included in an approach that sets a one year fixed follow-up standard. In addition, a return to crime after the federal warrant expires that earns a provincial/territorial sentence would not be recorded using only estimates based on a return to federal custody, suppressing recidivism estimates. Finally, high rates of return to custody without an offence following release in CSC restricts the time revoked offenders can be considered ‘at risk’. Unless the period of reincarceration following a revocation is not considered in calculating the follow-up time, the resulting estimates on a one year fixed follow-up will also distort the overall measure of recidivism.

Various measures of recidivism and various follow-up periods have been used across agencies within Canada and internationally. In their systematic review of international recidivism rates, Fazel and Wolf (2015) observed that the definition of recidivism, the inclusion of some reoffences in the calculation (particularly whether they included fines or traffic violations), and the follow-up time varied to such an extent across studies that international comparisons were not possible. Similar observations were made in the international review conducted by Wartna & Nijssen (2006) in which they proposed a framework for a common measure of recidivism.

Table 1 provides a sample of various definitions and follow-up periods extracted from some seminal studies. Most agencies will make use of multiple definitions depending on the purpose of the study; however, the definition most commonly cited is a reconviction and the follow-up time period most frequently applied is within 2 years of release (Yukhnenko, Sridhar, & Fazel, 2019). This was the time period and definition employed in the previous study on recidivism of federal offenders completed by Public Safety in 2003 (Bonta et al., 2003).

The current study provides a comprehensive estimate of recidivism rates of federally sentenced offenders based on reconvictions that resulted in a return to federal custody or reconvictions that resulted in provincial or territorial sanctions. The study answers the following questions:

1. Using official records of reoffending based on Canadian Police Information Centre (CPIC) data combined with Offender Management System (OMS) reoffending data, what are the rates of reoffending of any kind and reoffending with a violent offence based on various follow-up periods for the following groups: all men, all women, all Indigenous offenders (First Nations, Métis, Inuit), all Indigenous men, all Indigenous women, Black offenders, South East Asian offenders?
2. Using OMS data and information from the Crime Severity Index what is the relative severity of the reoffence for recidivists compared to their index offence?
3. Using OMS data, what are the recidivism rates for the groups defined in (1) based on:
   • returns to federal custody for any reason;
   • return to federal custody with an offence (pre and post warrant expiry);
   • return to federal custody with a violent offence (including sexual) (pre and post warrant expiry);
   • return with a Schedule 1 sexual offence (pre and post warrant expiry);
   • reoffence type: e.g., return for non-violent offending including drug related crimes and property crimes or a violent offence (pre and post warrant expiry);
4. Using OMS data and combined OMS and CPIC data what are the recidivism rates and violent recidivism rates by region?
5. Using OMS data separately and combined OMS and CPIC data what are the recidivism rates and violent recidivism rate by index offence?
6. Using OMS data separately and combined OMS and CPIC data, what are the recidivism rates and violent recidivism rates based on age groups?
7. Using OMS data separately and combined OMS and CPIC data what are the recidivism rates and violent recidivism rate by self-reported ethnic group?
8. Using OMS and CPIC data separately what are the reoffending and violent reoffending rates by Canadian or non-Canadian citizenship?
9. What are the comparison recidivism rates for released Canadian offenders and non-Canadian offenders?

An annual release cohort^{Footnote 1} from the 2007/2008 through 2011/2012 fiscal years was identified for a total sample of 22,685 federally sentenced offenders. Table 2 shows the frequencies of self-identified ethnic groups for men and women. The designation of ethnic group is problematic and reflects a great deal of heterogeneity. We provided the analysis, however, to assist in determining whether some groups with higher recidivism rates may contain require more service in order to improve the gap in outcomes on their release.

CPIC coding was completed for all offenders in a subsample of this group that included all offenders in the 2007/2008 and 2011/2012 release cohorts (N = 8,441). Offenders were typically in their mid-30s at the time of release (median = 35); however, Indigenous men and women had a lower median age than non-Indigenous men and women (32 years versus 36 years).

Data available in the Offender Management System (OMS), the official electronic record within CSC, was collected for all offenders in the release cohort simultaneously. Revocations and offences between the first release on a sentence and the earliest of reoffence date, readmission date, warrant expiry date, data collection date, date of death, deportation, or extradition were used to calculate revocation and revocation with an offence outcomes.^{Footnote 2} If an offender had multiple sentences during the five fiscal year timeframe, the first was selected. The date of data collection for revocations and for new warrant of committals was February 25, 2018. Any new federal sentences beginning after offenders’ warrant expiry dates to the data collection date were identified for new warrant of committal outcomes.^{Footnote 3} Violent offences were defined as schedule 1 or homicide offences. Violent sexual offences were defined as sexual offences that were also schedule 1 offences.

Time at risk of any offence across revocation, new warrant of committal, and CPIC records of convictions began with a first releases from a sentence during the 2007/2008 and 2011/2012 fiscal years. The offenders were followed through their warrant expiry dates until CPIC records were collected December 22nd, 2017, or in some cases, the offender’s death or deportation. If an offender had a revocation without an offence during supervision, the time from readmission to warrant expiry date was subtracted from the follow-up time. Only offences dated between the first release and readmission contributed to revocation offences.^{Footnote 4} Records from the Canadian Police Information Centre (CPIC) were collected beginning December 22, 2017. CPIC is the central police database where Canada's law enforcement agencies access and input information. As Canada's only national law enforcement networking computer system, it ensures officers across the country access the same information. It is maintained by the Royal Canadian Mounted Police (RCMP). Calculations of time at risk were used to restrict the eligible offenders for fixed follow-up analyses. For example offenders with less than 2-years of time at risk from release to data collection were excluded from 2-year fixed follow-up analyses. This procedure disproportionately affected later release cohorts and longer fixed follow-up analyses.

CPIC coding was completed for all offenders in the 2007/2008 and 2011/2012 fiscal year release cohorts. Any convictions from an offender’s warrant expiry date to the date of the offender’s death, deportation, or extradition, or the data collection date of December 22, 2017 were coded. Two coders examined all English files and a third coder completed all French files (6% of CPIC records collected were only available in French). The first reconviction date was recorded. Reconviction types were categorized into homicide, sexual, robbery, assault, other violent, drug, property, other non-violent, and Long Term Supervision Order (LTSO) breaches. The number of convictions of each offence type between warrant expiry date and CPIC records collection were recorded with the date of the first conviction of each offence type. A total of 8,441 offender criminal records were coded. Inter-rater reliability was examined on 396 randomly selected English CPIC records and was deemed to be acceptable. A kappa value of 0.98 was calculated for the variable recording whether an offender had any post-warrant expiry offence on their CPIC record. We conducted an analysis to determine the agreement between OMS and CPIC on the index offences for offenders’ initial sentences. We found that index offences recorded on OMS were also available on CPIC for 99% of cases. This supports the reliability of both the CPIC records and the coding procedure. Rates of return to custody with an offence while under federal warrant were combined with the CPIC records post warrant expiry to produce a composite rate of overall reoffending.^{Footnote 5} This calculation was made including all offences within the "other non violent" category.

Crime severity was measured using the Crime Severity Index (CSI; Babyak, Campbell, Evra, & Franklin, 2013). The data were derived from the offence serious code in the offence table in OMS. The Crime Severity Index was designed by Statistics Canada through a collaboration with the police, and territorial justice partners and academics across the country. The CSI was developed to create a measure that would provide a meaningful indicator of change in police-reported crime from year to year, and enhance the comparability of crime statistics at the provincial, territorial and municipal level by taking into account the relative seriousness of each offence. Each type of offence is assigned a seriousness weight. The weights are derived from sentences handed down by courts in all provinces and territories with more serious crimes being assigned higher weights.

The specific weight for any given type of offence consists of two parts. The first component is the incarceration rate for that offence type. This is the proportion of people convicted of the offence who are sentenced to time in prison. The second component is the average (mean) length of the prison sentence, in days, for the specific type of offence. Some common index offences with CSI weights less than 100 are: impaired operation of motor vehicle, assault, assault with a weapon or causing bodily harm, cannabis production or trafficking. First- and second-degree murder have the highest CSI weights followed by manslaughter, importing or exporting heroin, and attempted murder. First degree murder for example is weighted 1700. Impaired operation causing death, robbery, sexual assault with a weapon or causing bodily harm have CSI weights in the 500s. Heroin trafficking, level 1 sexual assault, and breaking and entering have CSI weights near the median of 237. The seriousness of the reoffence was compared to that of the index offence for offenders who were returned on a new offence while under federal community supervision.

A spreadsheet was provided to each CPIC coder pre-populated with offenders’ Finger Print Service (FPS) identifying number, name, sentence commencement date of the initial federal sentence, warrant expiry date, and date of death, deportation or extradition. Coders completed columns indicating whether the index offence from the initial sentence was on the CPIC record, whether there was an offence following Warrant Expiry Date (WED), the most serious type of sentence across reoffences on the CPIC record, the number of charges of each offence type, the conviction date of the first offence for each type, and the date of death, deportation or extradition if indicated on the CPIC record. Sentence types in order of severity included custodial, conditional, suspended sentence, probation, a fine, a prohibition order, or other. Offence types included homicide, sexual offences, robbery, assault, other violent, drug, property, other non-violent, and breach of LTSO conditions. The coding guide can be found in Appendix A.

Ethnic group based designations are derived from the categories selected by offenders at intake. Offenders can select from several categories. The White, Metis, First Nations, Black and South East Asian groups were the most represented and their numbers therefore permitted analyses of their outcomes. This standard is compatible with the recommendations for censuses contained in the United Nations' "Principles and Recommendations for Population and Housing Censuses", Revision 2, 2008 and is the convention used by Statistics Canada in their latest census.

Tables 3 and 4 present the rates of reoffending and violent reoffending for the offender groups for two cohort periods overall and also controlling for time at risk by reporting outcomes for fixed follow-up periods. The fixed follow-up periods began at the date of the offenders’ release from federal custody.

Of particular interest are the rates of reconviction for the two year fixed follow-up, the rate most commonly referenced in the literature (Yukhnenko et al., 2019) and the estimate provided in the previous study on federal reoffender recidivism (Bonta et al., 2003).^{Footnote 6} Figure 1 demonstrates the two year follow-up recidivism rates for all reoffences and for violent reoffences for all offender groups over two release cohorts 2007-2008 and 2011-2012. The two year reoffending rate for all offenders in the 2011-2012 cohort was 23%; the rate for men was 24% and for women 12%. Rates of recidivism for Indigenous offenders were higher – 38% for Indigenous men and 20% for Indigenous women. There is a steady reduction in recidivism rates for both men and women over the two release cohorts. Rates increase with longer follow-up periods. For the 2011-2012 cohort, almost 38% of all federal offenders reoffended within five years of release and almost 60% of Indigenous men reoffended within this time period.

The two-year post release violent reconviction rate for all federal offenders in the latest cohort year was 12%. The rates for Indigenous men and women were higher than for non-Indigenous offenders at 22% and 11% respectively. The rate of violent reoffending increased with each year of follow-up with the five year rate of violent reoffending being just under 20%. We do, however, see a decrease in rates of violent recidivism for offenders released in the 2011-2012 compared to the 2007-2008 year.

Figure 1. Rates of any new reoffence 2 years post release (rate of violent reoffending 2 years post release in brackets)^{Footnote 7}

NIM: non-Indigenous men | IM: Indigenous men
NIW: non- Indigenous women | IW: Indigenous women

Rates of returns to federal custody, particularly the rates of return for a new offence, during the time that an agency is responsible for the offenders’ supervision, is a key marker of an agency’s success in promoting public safety. Ideally, revocation of a conditional release in the absence of a new offence can be viewed as a suppression strategy that deters potential reoffending. Offenders can be revoked without having committed an offence based on the parole officers’ observation that risk is escalating or due to violation of the conditions of release.

Table 5 provides the revocation rates including revocations for a reoffence and revocation with a violent reoffence by gender and Indigenous ancestry for each year of release. Figure 2 based on these results illustrates the gradual improvement in revocation rates over each year. With some exceptions, the pattern of improvement in results is evident across offender groups. Rates for women are more erratic because of their low rate of reoffending and smaller numbers. Rates of violent reoffending while offenders are under warrant were low, under 3% for men and under 1% for women, and rates of sexual offending while under warrant were well under 1% for all groups. In order to confirm that the decline in rates over time was not related to varying time at risk, we completed an analysis with a 6 month fixed follow up. Table B1 in Appendix B provides these results in which only revocations and revocations with an offence that occurred within 6 months of release are considered, and only offenders with at least 6 months from release to WED, deportation or death are included. The results confirmed a similar pattern to that in Table 5. In particular, the rates of revocation for any reason are sharply improved for the later cohort groups for women (from 21% among the 2007-2008 cohort to 11% for the later 2011-2012 cohort) as were rates of revocation with an offence for women (from 7% to 3%).

Offenders who returned to federal custody post-warrant expiry have reoffended after their warrants were completed and committed an offence that earned a sentence of two years or more. Table 6 shows the same decline in rates of reoffending and returning to federal custody with each cohort with the exception of sexual offending which remained under 1% across the cohort years for all groups except for Indigenous men.

Controlling for variable follow-up time across groups using fixed follow-up periods, Table 7 shows that the rate of reoffending increased the longer the follow-up period. The mean rate of return to federal custody with an offence within 5 years for the combined cohorts is 18%; the rate for women is 10%. The rate of return to federal custody with an offence for Indigenous men offenders is substantially higher over this time period at 27%. However, for all offender groups we see a gradual decline in rates of return until the 2011-2012 cohort where there was a slight uptick across most groups^{Footnote 8}.

Table 8 provides the rates of reoffending post WED-based on CPIC records. These reoffences could have resulted in either a provincial/territorial or federal sentence. Again, we see the gradual increase in recidivism with increased follow-up time. The sharpest increase with longer follow-up is for Indigenous women in the 2011-2012 cohort where there is an increase from 10% within one year of release to 44% after 5 years of release. Rates of reoffending declined at all fixed follow-up periods for the later cohort group providing additional evidence for the gradual improvement in correctional results over time. Figure 3 further illustrates this trend. We conducted an analysis of the types of violent reoffences committed by offenders post WED that were recorded on CPIC. The most frequent offence category among the violent offences was assault (58% of violent crimes on CPIC records). Within this category over half were for common assault Level 1. Other categories of violent reoffending included homicide (0.6%), sexual assault (3%), robbery (6%) and other violent offences (38%) of which the largest category was criminal harassment and uttering threats.

Although offenders are typically criminally versatile there are established patterns in reoffending rates among offenders convicted for certain offence categories. Tables 9 and 10 provide the overall reoffending and violent reoffending rates for offenders who were released on a federal sentence for various crime categories. Tables B2-B5 in Appendix B provide the results related to the analysis of reoffending by offence categories pre- and post-warrant expiry.

Across all groups, on all measures of recidivism (i.e., rates of returns to federal custody for any reason and returns for an offence, rates of return to federal custody with new warrants of committal and rates of return post-WED based on CPIC) the highest rates of reoffending were for offenders who had been serving a sentence for robbery or property offences. The lowest rate of reoffending was among offenders who had served a sentence for a sexual offence. The highest rates of violent reoffending were consistently for male offenders with an index offence for robbery and assault. The rate of recidivism was higher for Indigenous than non-Indigenous men for every index offence type, and at each time period with the exception of homicide-related offences which were higher for non-Indigenous men at one- and two-year follow up periods. The 5 year reoffending rate for Indigenous men released on robbery or property offences is approximately 76%. Rates of reoffending for Indigenous women with robbery or property offences are also high at over 60%.

Crime Severity Index (CSI) weights were available for the index offences of 22,646 of the 22,685 offenders released during the 2007/2008 to 2011/2012 fiscal years. The mean CSI weight was 467, and the median CSI was 237, with a standard deviation of 898. Table 11 compares the mean CSI weights of the index offence and the reoffence for offenders who were under a federal warrant at the time of the reoffence. As depicted in the table, 61% of offenders who reoffended while under warrant had a CSI score that was lower than their original index offence.

The following tables (Tables 12-16) show various measures of recidivism rates for each of the five regions over the 2007/2008 through 2011/2012 fiscal years. General recidivism rates (Table 12) for the total population post release for all time frames were usually highest for the Prairie region and lowest for the Quebec region. Rates for Indigenous men were more similar across the regions but the 2, 3, and 5 year follow up periods showed the highest rates of reoffending among Indigenous offenders in the Ontario region. The highest overall rates for violent reoffending were in the Prairie region and the lowest in Quebec and the Atlantic regions (Table 12). However, among non-Aboriginal offenders the highest rate of violent reoffending was in the Ontario region and among women offenders, the highest rate of reoffending among non-Indigenous women was in the Atlantic region.

The pre-WED return to custody outcomes demonstrated notable differences in the regional rates of offenders’ return to custody, revocations with an offence, as well as revocations with a Schedule 1 or homicide offence. Numbers were too small to meaningfully compare revocations due to Schedule 1 sexual offences. The region with the lowest median days to return (272), and the highest rate of return to custody (54%) and revocation with offence (17%) was the Prairie region; Quebec had the highest median days to return (366), lowest rate of any return (40%), and one of the lowest rate of revocation with offence (10%; Ontario was slightly lower). Indigenous men in the Prairie region had the highest overall rate of return prior to their warrant expiry date at 67%. The rate of violent reoffending across regions was under 4%; the rates of violent reoffending while under federal warrant was highest in the Pacific region (4%) and lowest in the Ontario region (2%).