The Reconviction Rate of Federal Offenders 2003-02

PDF (48 KB)

Table of Contents

James Bonta, Tanya Rugge, Mia Dauvergne
Solicitor General Canada

In consultation with:
Portfolio Corrections Statistics Committee
Chair: Robert B. Cormier


The authors would like to thank the members of the Portfolio Corrections Statistics Committee for their assistance and guidance in the formulation of the project. Our indebtedness is extended to the students who worked diligently with the criminal history records (Ian Broom, Jennifer Cooney, Michelle Gringler, Jonathan Huber, Jennifer Lavoie, Amy Yuille and Victor Yan) and to Kimberly Smallshaw who conducted the Offender Management System searches. We also thank Suzanne Wallace-Capretta who helped us with various parts of the data coding and the construction of the original database and Karl Hanson for his advice and work on the statistical adjustment required for the second and third release cohorts. Finally, we extend our appreciation to Mike Hayden who extracted the offender demographic and offence data from the Offender Management System of the Correctional Service of Canada and patiently reviewed the data to insure its integrity.

Executive Summary

The recidivism of offenders is of interest to the public and an important indicator of the impact of criminal justice interventions. Measuring recidivism, however, is a complex matter. The various measures that are used have their advantages and disadvantages. For example, successful completions of day parole is helpful for assessing the National Parole Board's release decisions but the time period is relatively short and does not include criminal behaviour past the period of supervision.

The present study was conducted as part of the mandate of the Solicitor General Portfolio Corrections Statistics Committee to provide the general public and professionals basic statistical information on corrections and conditional release. In this study, recidivism was defined as any new conviction for an offence committed within two years of release from prison. The study samples included all releases (except for releases on temporary passes) from federal penitentiaries during the three fiscal years 1994/95, 1995/96 and 1996/97.

The reconviction rate for the first fiscal year release cohort was 44.0%, 42.8% for the second release cohort and 40.6% for the third cohort. These reconviction rates were comparable to other rates reported internationally and from other Canadian studies using a similar methodology. Nonviolent reconvictions accounted for the majority of the reconvictions. The violent reconviction rate was much lower; approximately 13% for all three release cohorts and the sexual offence reconviction rate was very low (0.7% to 1.7%). Consistent with previous research findings, Aboriginal offenders showed higher reconviction rates than non-Aboriginal offenders and the male reconviction rate was higher than for women. However, for both Aboriginal offenders and women the reconviction rates declined steadily over the three years of the study period. Finally, the majority of reconvictions occurred after the expiry of sentence when the offender was no longer under supervision in the community.


Recidivism, generally defined as a return to crime, is perhaps one of the most important indicators of the impact of criminal justice interventions. There are, however, certainly many other measures that help us understand the operation and worth of our efforts to deal with crime. Police clearance rates, court caseloads and the cost of incarceration are but a few examples of the types of information needed to understand and hopefully improve the different components of the criminal justice system. Recidivism, however, provides information that has value for all facets of the criminal justice system. Knowing the recidivism of offenders is important for police, the courts, crime prevention and, most importantly, for corrections. The understanding of offender recidivism within a correctional context is the focus of this report.

The major mandate of corrections is to administer the sentences of the courts with the overall goal to enhance public safety. To attain the latter objective, offenders must be reliably differentiated with respect to their risk of recidivism. Offenders do not pose equal risks to re-offend and recognizing these differences is important for conditional release decisions and the delivery of interventions that reduce the likelihood of offenders committing new crimes. Regardless of our interest, the impact of release decisions or the effectiveness of correctional interventions, a reliable and valid measure of recidivism is needed. Other criminal justice indicators may be useful, but it is the reduction of criminal behaviour that is the primary concern of many correctional efforts. Today, there is an abundance of research and studies attesting to the merit of using recidivism to improve correctional decision-making and interventions. What is also clear from this large literature is that measuring recidivism is no simple matter.

The Measurement of Recidivism

There is no commonly accepted measure of recidivism. The various measures used have many advantages, but each also has disadvantages. There are four major factors (and a number of more minor considerations) influencing the measurement of recidivism. First, there is the definition of a return to crime. How do we know that someone has re-offended? Should we define recidivism as a police arrest, a new conviction or a return to prison? None of these indicators is perfect. An arrest may identify individuals who did not commit a crime and were later released by the police, the charges dismissed by the courts or found not guilty at trial. Convictions do not take into account the results of plea-bargaining and insufficient evidence to convict. Reincarceration typically includes the more serious crimes and over-represents offenders with lengthy criminal histories. Some observers argue that "official" measures of crime are misleading because they miss the many crimes that escape police detection but we know occur as evidenced in victimization surveys.

Table 1 presents the results from three American studies that illustrate what is typically found when different measures of recidivism are used. All three studies reported a follow-up of three years. Depending on the measure, we can have very different numbers. New arrests tend to show the highest rates while incarceration measures yield the lowest rates.

Table 1. Recidivism by Type of Measure (3 year follow-up)

% Recidivated





Beck (1989)




Corbo (1992)




Langan & Levin (2002)




Related to the definition of recidivism is the nature or type of recidivism. Many times we are interested in more than whether or not a crime has occurred. The seriousness of the new crime is often an important interest in studies of recidivism. Violent crimes are relatively infrequent and when recidivism is concerned with violence, we find low rates of violent recidivism. Sexual recidivism, a subset of violent recidivism, is more infrequent even among known sex offenders. For example, Hanson and Bussière's (1998) review of 61 sex offender studies found a general recidivism rate of 36.3% (four and one-half year follow-up) and 13.4% for sexual recidivism. Bonta, Harman, Hann and Cormier (1996) reported a general recidivism rate (reincarceration after three years) of 48.7% and a violence recidivism rate of 18.6% when the definition of violence included robbery and 9.8% when robbery was excluded.

A third factor to consider in the measurement of recidivism is the length of follow-up. The studies displayed in Table 2 used convictions as the measure of recidivism. Convictions are a commonly used measure in published research articles. Convictions appear to provide a reasonable estimate of recidivism without the extreme results found with measures such as arrest and incarceration. In general, the longer the follow-up, the higher the recidivism rate (see Table 2). Although the two-year recidivism rate hovers in the 40% to 50% range, there is significant variation. For example, Cormier (1981) reported a 48.9% rate and Hoffman and Stone-Meierhoefer (1980) found a recidivism rate of 25.7%. How can studies that use the same measure of recidivism and the same follow-up period find such disparate recidivism rates? Some of this variation can be explained by differences in the sample of offenders, the fourth major factor influencing the measurement of recidivism.

Table 2. Recidivism (Convictions) and Length of Follow-up


Follow-up (yr)

Recidivism (%)

Jones (1991)







Gendreau & Leipciger (1978)




Cormier (1981)




Kershaw (1999)




Hoffman & Stone-Meierhoefer (1980)










Corbo (1992)




Beck (1989)




Different offender samples will produce varying recidivism rates. For example, the differences in recidivism rates found by Cormier (1981) and Hoffman and Stone-Meierhoefer (1980) may be traced to the sample studied. Cormier (1981) analyzed the recidivism rate of Canadian federal offenders released from Joyceville Institution, a medium security penitentiary. Hoffman and Stone-Meierhoefer (1980) studied inmates released from U.S. federal prisons. Although both studies followed federal prisoners, the U.S. Bureau of Prisons has responsibility for offenders convicted of federal crimes with drug offenders representing the largest proportion of offenders in U.S. federal prisons. Table 3 presents a sampling of studies that show the variation in recidivism rates as a function of the type of sample.  

Table 3. Two-Year Inmate Recidivism (Convictions) Rates and Type of Sample


Recidivism (%)

Gendreau & Leipciger (1978) Provincial/Canada


Cormier (1981) Federal/Canada


Kershaw (1999) Inmates/UK


Hoffman & Stone-Meierhoefer (1980) Federal/US


Jones (1991) State/US


Beck (1989) State/US


From this brief summary of the issues surrounding the measurement of recidivism, the challenge for any correctional organization is to select a measure and follow-up period from a number of options available. Adopting one measure with one methodology yields three important benefits. First, it allows the Department of the Solicitor General to give the public an easily understood estimate of the federal offender's "return to crime" that is independent of the particular interest of the agencies of the Solicitor General. For example, re-arrests may be more important for the operational planning of the Royal Canadian Mounted Police (RCMP) and less important for the Correctional Service of Canada where re-admission to a federal penitentiary is more important. For the National Parole Board (NPB), outcomes while on some form of conditional release are important. Therefore, for example, unsuccessful completions while on day parole (approximately 20%) are not comparable to the reconvictions rates reported in this study. Parole outcomes are dependent on release decisions made by NPB and usually cover much shorter follow-up time periods. Second, a common definition of recidivism would permit the construction of year-to-year trends in recidivism. Finally, a uniform measure of recidivism can be used as a standard for measuring the effectiveness of correctional programs. Presently, program evaluators use different measures of recidivism outcomes making it difficult to compare the results across programs. The challenge of selecting a measure of recidivism must be addressed while recognizing that no single measure meets everyone's needs and that changes in recidivism over time may be due to many factors other than the performance of the correctional system (e.g., unemployment rates, socio-demographic variation).

Making Choices for the Assessment of Recidivism

The Portfolio Corrections Statistics Committee, as part of its mandate to produce the "Corrections and Conditional Release Statistical Overview" (CCRSO), undertook to adopt a measure of recidivism that would be used as a common measure by the Department and the agencies (Correctional Service of Canada and National Parole Board) within the Portfolio of the Solicitor General. The long-range plan is to include year-by-year recidivism results in the CCRSO. Thus, the immediate decision that the Committee faced was to choose the first release cohort of federal inmates that would mark the beginning of what would become part of a series of data points in a recidivism trend line. Releases during the fiscal year 1994/95 formed the initial release cohort and subsequent year releases will form the basis for future reports on the recidivism of federal offenders. The reasons for beginning with this release year were threefold. First, 1994/95 followed a major restructuring of the Offender Management System (OMS), the computerized database of federal offenders. Selecting an earlier year would have encountered inaccuracies and missing information evident in earlier versions of OMS. Second, the year chosen post-dated the Corrections and Conditional Release Act (1992) and therefore avoided possible confusions over changes in definitions introduced by the Act. Third, given the six to 12 month delay often encountered in recording new offences in the RCMP Criminal History records, the elapsed time between release and follow-up ensured a reasonably comprehensive record.

Next there was the more substantive challenge of selecting a measure of recidivism. As outlined earlier, no one measure is free from bias. Although there has been a tradition in recidivism research with federal offenders to use a return to federal custody as an outcome measure, the Committee decided to use reconviction during the period of release as the measure of recidivism. For most Canadian citizens, the distinction between federal and provincial corrections and the relevance of a return to custody for the federal system is immaterial. Defining recidivism as a new conviction recognizes the federal government's responsibility to address the criminal behaviour of federal offenders even when that behaviour does not impact on the federal correctional system. In addition, the Committee decided to examine both general recidivism and violent recidivism. Sexual recidivism was not demarcated from violent recidivism because of the very low base rates that would have made interpretation of future trends extremely problematic.

In Canada, conviction information is available from two sources: court records and the RCMP's Finger Print Service (commonly referred to as FPS criminal history records). Accessing court records on thousands of offenders released across the country was neither practical nor cost-efficient. FPS records, on the other hand, provide a national database of criminal convictions that is easily accessible. There are, however, some limitations in using FPS records as a source of information on reconvictions. One potential problem is that offenders with summary convictions may not be fingerprinted and recorded on the RCMP's database. This was not seen as a problem for federal offenders with lengthy and serious criminal histories but it could present a problem in measuring reconvictions in provincial and territorial jurisdictions where there are more offenders who have committed less serious crime. Thus, the methodology used in this study may not necessarily apply to a provincial or territorial jurisdiction.

Finally, there was the issue of the length of follow-up after release from penitentiaries. A two-year follow-up was finally decided upon. This time period was considered sufficiently long to permit a reasonable estimate of low base rate, violent crimes. Another reason for the two-year period was that approximately half of the release cohort would have progressed past the expiry of their sentence, thereby permitting an assessment of reconviction when they were no longer under supervision. In summary, recidivism was defined as any new conviction for an offence committed within two years of release from prison.


The present report deals with three release cohorts of federal offenders. Releases included full parole, day parole, statutory releases and releases at expiration of sentence. Offenders released on temporary passes (escorted or non-escorted) were not included. The first cohort consisted of all federal offenders released from penitentiaries during the 1994/95 fiscal year (April 1, 1994 to March 31, 1995). The second and third cohorts were samples of all releases during the 1995/96 and 1996/97 fiscal years (the sampling procedure for the second and third cohorts is described shortly). The last release during the particular fiscal year was chosen to minimize the number of prison re-admissions due to revocations for breach of conditions.

For the first cohort, the total number of releases during 1994/95 was 7,343 and there were 7,399 and 7,259 releases in the following two years. As the coding of reconvictions is a very resource intensive task, a sample of all releases for the following years was selected. All releases of female offenders (n = 224 and n = 232) and Aboriginal male offenders (n = 933 and n = 1,063) were included in the 1995/96 and 1996/97 cohorts. However, for non-Aboriginal male offenders a sample was randomly selected to give us a maximum error rate of two per cent (see the Technical Coding Manual for the specific formula used to calculate the sample size). In order to ensure an error rate not exceeding two per cent, and anticipating a loss of subjects when doing the criminal history follow-up, we sampled approximately 2,400 non-Aboriginal male cases for each of the two later years.

RCMP criminal history records were used to record new convictions and the type of new offence. Compared to using the Offender Management System (OMS) of the Correctional Service of Canada (CSC) to collect reconviction data, RCMP records have the advantage of recording offences that do not result in a return to federal custody (e.g., offences resulting in fines or provincial sentences). Although the RCMP records have a typical six to 12 month delay before some new convictions are actually recorded, the records were not requested until nearly four years after release of the inmates. Therefore, the records were considered relatively comprehensive.

Criminal history information was unavailable for 62 offenders from the 1994/95 cohort, 24 offenders from the 1995/96 cohort and 49 offenders from the 1996/97 cohort. Missing information was due to a variety of factors (e.g., incorrect names and FPS numbers in either the CSC or RCMP databases preventing proper matches). In addition to recording whether or not a new conviction occurred, information such as the type of new offence and the court disposition was noted (a Technical Coding Manual provides further detail on the various coding rules used in this study). Offences were grouped into two general categories: 1) non-violent, and 2) violent. The violent category included all crimes against the person, including sexual offences. In the case of multiple offences, the most serious offence dictated the assignment to categories. The most serious offence was defined as the offence with the most severe disposition, generally measured by sentence length. When two offences received the same disposition (e.g., a four year sentence for each crime), then a crime against person was given priority.

Deaths occurring during the follow-up period required special treatment. If an offender died before the end of the follow-up and showed no evidence of reconviction before his/her death, then the offender was removed from the sample. However, if the offender recidivated or remained offence-free and died after the follow-up period then the offender remained in the analyses. After removal of cases of death that occurred prior to the two-year follow-up and the reconviction event, 7,207 offenders remained in the 1994/95 release cohort, 3,505 offenders in the 1995/96 cohort and 3,602 offenders in the 1996/97 cohort. Information on gender and race was unavailable for eight cases in the 1995/96 cohort further reducing the sample of this cohort to 3,497. These three samples served as the bases for the analyses described in the report.

As described earlier, the construction of the second and third release cohorts involved a sampling of non-Aboriginal male offenders and not a complete representation of all releases during 1995/96 and 1996/97. Therefore, in calculating the reconviction rate it was necessary to make an adjustment for the sampling of non-Aboriginal releases (Formula 2 in Technical Coding Manual). We also calculated the 95% confidence interval (CI) for the second and third release cohorts because it did not represent all releases. The CI gives the lower and upper limits where the true population average would fall 19 times out of 20. Estimating the CI required the calculation of the variance for the stratified sample using Formula 3 (Technical Coding Manual).

Finally, once the RCMP criminal history information was coded, the data files were merged with CSC data files that contained some personal demographic and criminal history information (the Technical Coding Manual provides a complete listing of variables). The CSC data was drawn from the OMS system and allowed some comparisons in reconviction rates along gender and ethnicity.

Data Integrity

Sometimes inmates released from prisons are charged with crimes that have occurred prior to release. New evidence may come to the attention of the police or crimes are committed while incarcerated or before the period of incarceration. Therefore, it is important in studies of recidivism to ensure that old crimes are not included in the measure of recidivism. To minimize this threat to our measure of reconviction, all convictions that occurred within 60 days of release were further investigated. Using OMS, cases were reviewed to verify whether the conviction recorded on the RCMP's Criminal History record actually occurred after release from prison. For example, in the 1994/95 release cohort there was 324 cases of convictions recorded within 60 days of release. Of these cases, 4.6% involved crimes that occurred prior to release. Thus, these crimes were not counted as new reconvictions.

Research assistants under supervision of senior research officers conducted the coding of criminal history records. The research assistants were given training in interpreting the records and the operation of the computerized data entry program (SPSS). Coding inter-rater agreement by three research assistants was conducted on a random sample of 151 cases selected from the 1995/96 cohort and 202 cases from the 1996/97 cohort. Nineteen variables were chosen for evaluation. The inter-rater agreement ranged from 87% (most serious offence at reconviction) to 100% (has subject re-offended since release date). The overall inter-rater agreement was 97%.


An overview of the characteristics of the inmates released during the three years studied is shown in Table 4. The vast majority of releases in both cohorts were male and most had committed violent offences (Schedule 1). The average sentence length was 1,813 days (5.0 years) for the first cohort. For the second and third cohorts, average sentence lengths were 1,572 days (4.3 years) and 1,507 days (4.1 years) respectively. Offenders serving indeterminate sentences were very infrequent in all three cohorts (approximately 2% of releases).

Our measure of recidivism was a new conviction within two years of release. The reconviction rates for the three release cohorts are shown in Table 5. For the first release cohort, 44.0% of the released offenders were reconvicted within two years. For the second release cohort (1995/96), the estimated reconviction rate was 42.8% and it was 40.6% for the third cohort. Comparing the first cohort's reconviction rate (44.0) with the 95% confidence interval of the second cohort (41.5 - 44.1) shows overlap indicating that the reconviction rate remained unchanged in the second cohort. However, the 40.6% reconviction rate for the third cohort was lower than the rate for the 1995/96 cohort (the confidence interval did not overlap with the first cohort's reconviction rate). Similarly, the non-violent reconviction rate for the third cohort was also lower than the rate for the first cohort. There was no change in the violent reconviction rates over the three years. New sexual offences resulting in a conviction were infrequent. Forty-nine inmates or 0.7% of the sample recidivated with a sexual offence in the 1994/95 cohort, 1.0 % (n = 36) in the 1995/96 cohort, and 1.7% (n = 61) in the 1996/97 cohort.

Table 4. Personal-Demographic and Criminal Histories of the 94/95, 95/96 and 96/97 Release Cohorts

Per cent/Mean


(n = 7207)

(n = 3497)

(n = 3602)

Gender: Male








Race: Non-Aboriginal








Age at release (years)




Marital Status: Single












Sentence length (days)*




Most Serious Offence (MSO):      
























Schedule 1 Offence




Schedule 2 Offence




Indeterminate Sentence




* Inmates serving indeterminate sentences are excluded.


Numbers may vary due to missing information. For the 1994/95 cohort, sentence length and MSO was coded for a sample of releases (n = 881). For the 1995/96 and 1996/97 cohorts, proportions for gender and race vary due to complete sampling of women and Aboriginal offenders.

Schedule 1 comprises sexual offences and other violent crimes excluding first and second degree murder. Schedule 2 comprises serious drug offences, or conspiracy to commit serious drug offences (Corrections and Conditional Release Act).

For MSO, the category "other" includes crimes such as property violence and crimes against public order and public morals.

Table 5. Reconviction Rates for Federal Offenders (CI = Confidence Interval

% Reconvicted (CI)

Type of Reconviction

1994/95 cohort

1995/96 cohort

1996/97 cohort

Any Reconviction


(41.5 - 44.1)

(39.3 - 41.9)



(28.6 - 31.0)

(26.4 - 28.8)

Any Violent


(12.2 - 13.9)

(12.2 - 13.8)

1994/95 Cohort = 7,207; 1995/96 Cohort = 7,399. Rates based on a sample of 3,497.
1996/97 Cohort = 7,259. Rates for 1996/97 based on a sample of 3,629. For the 1995/96 and 1996/97 cohorts, the number reconvicted is an estimate.

Table 6 shows the recidivism rates as a function of race and gender. Only male offenders were compared along the variable race; there were only 35 female Aboriginal inmates in the 1994/95 cohort, 37 in the 1995/96 cohort and 40 in the last year. Aboriginal male offenders demonstrated higher reconviction rates across all categories and all years. The higher reconviction rate for male Aboriginal offenders may be partly traced to their elevated risk for re-offending. For example, male Aboriginal inmates were less likely to be released on full parole than non-Aboriginal male inmates (6.9% vs. 12.9% in the 1994/95 cohort, 7.7% vs. 12.9% for the 1995/96 cohort and 9.1% vs. 12.6% for the 1996/97 cohort). Although the Statistical Information on Recidivism (SIR) risk scale is not given routinely to Aboriginal offenders, it is administered to some. In the 1994/95 cohort, only 62 Aboriginal offenders were administered the SIR scale and the scores for Aboriginal offenders were no different from the non-Aboriginal offenders (-5.4 vs. -3.6, t = 1.44, p > .05). However, in the 1995/96 and 1996/97 cohorts where more Aboriginal inmates were administered the SIR scale (n = 147 and n = 304), they demonstrated statistically significant poorer risk scores than non-Aboriginal male inmates (-6.62 vs. -2.23; t = 4.92, p < .001 and -5.48 vs. -1.56; t = 6.09, p < .001). On a year-to-year basis, the male Aboriginal offenders did demonstrate slight decreases in the various reconviction rates. The male non-Aboriginal offenders showed statistically significant change only for reconvictions and non-violent reconvictions from the first year to the third year.

Table 6. Percent Reconvicted by Race: Males Only (CI)
Type of Reconvicted



Any Reconviction:    
1994/95 Cohort



1995/96 Cohort

41.2 (39.6 - 42.8)


1996/97 Cohort

39.1 (37.9 - 40.4)


1994/95 Cohort



1995/96 Cohort

29.1 (27.7 - 30.5)


1996/97 Cohort

27.1 (25.9 - 28.3)


Any Violent:    
1994/95 Cohort



1995/96 Cohort

12.1 (11.1 - 13.1)


1996/97 Cohort

12.0 (11.2 - 12.8)


Notes: 1994/95 Non-Aboriginal = 6,018; Aboriginal = 961. 1995/96 Non-Aboriginal = 6,242 (sample = 2,362); Aboriginal = 919. 1996/97 Non-Aboriginal = 5,964 (sample = 2,334); Aboriginal = 1,046.

Men showed higher reconviction rates than women across all categories (any, non-violent and violent; Table 7). For men in the third cohort, there was a statistically significant decrease in general and non-violent reconvictions compared to the 1994/95 cohort. There were no differences in violent reconvictions for men across the years. Women also showed similar trends (i.e., decreases in general and non-violent reconvictions and no trend in violent reconvictions). Extreme caution must be used in interpreting the violent reconviction rates for female offenders as the samples were very small (there were 14 women who were reconvicted of a violent offence in the 1994/95 cohort, 16 in the 1995/96 cohort and 15 in the 1996/97 cohort).

Table 7. Percent Reconvicted by Gender (CI)
Type of Reconviction



Any Reconviction:    
1994/95 Cohort



1995/96 Cohort

43.1 (41.7 - 44.4)


1996/97 Cohort

41.2 (39.9 - 42.5)


1994/95 Cohort



1995/96 Cohort

30.0 (28.8 - 31.2)


1996/97 Cohort

28.0 (26.8 - 29.2)


Any Violent:    
1994/95 Cohort



1995/96 Cohort

13.2 (12.3 - 14.1)


1996/97 Cohort

13.2 (12.3 - 14.1)


Notes: 1994/95 Cohort: males = 6,997; females = 210.
1995/96 Cohort: total male releases = 7,171, male sample = 3,281; females = 216.
1996/97 Cohort: males = 7,027, male sample = 3,397; females = 222.
Rates adjusted to account for sampling of male non-Aboriginal offenders.

The majority of the offenders in the release cohorts were under supervision prior to their warrant expiry date (WED). With a two-year follow-up, released inmates could recidivate while under federal supervision or after the end of their sentence. Table 8 displays the occurrence of the reconviction event in relation to expiration of sentence. More than half of the reconvictions occurred after the supervision period.

Table 8. Percent Reconvicted Pre- and Post-Warrant Expiry (CI)
Type of Reconviction



Any Reconviction:    
1994/95 Cohort



1995/96 Cohort

18.5 (17.4 - 19.6)

24.7 (23.6 - 25.9)

1996/97 Cohort

15.4 (14.5 - 16.4)

25.7 (24.5 - 26.8)

1994/95 Cohort



1995/96 Cohort

13.2 (12.3 - 14.1)

16.9 (15.9 - 17.9)

1996/97 Cohort

11.3 (10.5 - 12.1)

16.8 (15.9 - 17.8)

Any Violent:    
1994/95 Cohort



1995/96 Cohort

5.3 (4.7 - 5.9)

7.8 (7.1 - 8.5)

1996/97 Cohort

4.2 (3.7 - 4.7)

8.8 (8.2 - 9.6)

Cohort 94/95 = 7,207 (excludes indeterminate sentences and missing WED dates).
Cohort 95/96 = 3,462 (excludes indeterminate sentences and missing WED dates).
Cohort 96/97 = 3,522 (excludes indeterminate sentences and missing WED dates).
Rates estimated for sampling of non-Aboriginal offenders.
CI = 95% confidence interval.


The major goal of the present study was to derive a standard measure of recidivism for use by the Portfolio of the Solicitor General. After weighing the advantages and disadvantages of different measures of recidivism, the Committee chose a new conviction for an offence committed within two years as the most acceptable measure from the choices available. Using this measure, we found a general reconviction rate of 44.0% for inmates released in 1994-95 and estimated rates of 42.8% and 40.6% for releases in 1995-96 and 1996/97. Violent reconvictions were less frequent (13.8%, 13.1% and 13.0% for each respective cohort) and most reconvictions occurred after expiration of sentence and supervision.

Reconviction is by no means a perfect measure of recidivism that captures all new crimes committed by offenders. Many crimes probably go undetected. In some of our analyses, we could not even be certain whether the crime was committed before expiry of sentence and the period of supervision or after sentence completion. To be certain, a review of court and police files in the provinces and territories would have been required. Balancing the advantages and disadvantages of various measures of recidivism was a choice of selecting the most advantageous measure and it was a difficult choice for the Committee. However, a choice had to be made as the public deserves a uniformly reported measure of recidivism rather than the confusing range of statistics presently offered. We hope that by outlining the limitations of the present methodology and the reasons for choosing reconviction as our measure of recidivism we give a common language to the correctional agencies of the federal government.

In closing, a final note of caution is warranted. It may be tempting to view reconviction as a major performance indicator for the federal correctional system. However, this would be misleading. As noted in the introduction, many factors can determine the recidivism rate. In this study, for example, we found that the differences in reconviction rates for Aboriginal and non-Aboriginal offenders may be partly due to differences in risk characteristics between the two groups. However, other factors over which the prison system has no control can also account for the differences (e.g., situational factors in the community, socio-demographic and economic conditions). Furthermore, it would be a mistake to exclude other indices of performance (e.g., employment and educational achievement, family stability, sobriety) in favour of one measure such as reconviction. Interventions designed to re-integrate inmates into the community need to be evaluated by multiple levels of outcome. Recidivism, or reconviction in the present case, is but one performance indicator among many.


Beck, A. J. (1989). Recidivism of prisoners released in 1983. Bureau of Justice Statistics Special Report (April). Washington, DC: U.S. Department of Justice.

Bonta, J., Harman, W. G., Hann, R. G., & Cormier, R. B. (1996). The prediction of recidivism among federally sentenced offenders: A re-validation of the SIR scale. Canadian Journal of Criminology, 38, 61-79.

Corbo, C. A. (1992). Release outcome - 1984: A follow-up study. Trenton, NJ: New Jersey Department of Corrections.

Cormier, R. B. (1981). Canadian recidivism index. Canadian Journal of Criminology, 23, 103-104.

Gendreau, P., & Leipciger, M. (1978). The development of a recidivism measure and its application in Ontario. Canadian Journal of Criminology, 20, 3-17.

Hanson, R. K., & Bussière, M. T. (1998). Predicting relapse: A meta-analysis of sexual offender recidivism studies. Journal of Consulting and Clinical Psychology, 66, 348-362.

Jones, P. R. (1991). The risk of recidivism: Evaluating the public safety implications of a community corrections program. Journal of Criminal Justice, 19, 49-66.

Kershaw, C. (1999). Reconvictions of offenders sentenced or discharged from prison in 1994, England and Wales. Home Office Statistical Bulletin (April 19). London: Home Office.

Langan, P. A., & Levin, D. J. (2002). Recidivism of prisoners released in 1994. Washington, DC: Bureau of Justice Statistics

Date modified: