Correctional Service Canada
Symbol of the Government of Canada

Common menu bar links

Research Reports

Warning This Web page has been archived on the Web.

Rates of Recidivism for Women Offenders

2008 No R-192

Renée Gobeil 
Meredith Robeson Barrett
Correctional Service Canada
September, 2007

ACKNOWLEDGEMENTS

Appreciation is extended to Kelley Blanchette for her methodological and editorial suggestions.

EXECUTIVE SUMMARY

To date, researchers in the area of Canadian adult women's recidivism have employed varying methodologies and have reported rates of recidivism ranging from 3.7 to 46.9% (as summarized in Blanchette, 2001). The present study aimed to add precision to these estimates by using an explicitly defined methodology to provide an up-to-date measure of recidivism rates for federally sentenced women. The use of a procedure similar to that employed in a recent study (Bonta, Rugge, & Dauvergne, 2003) ensured a common language across these studies, allowing for meaningful comparisons.

Recidivism was defined in three ways: 1) any revocation of conditional release (due to a technical violation, new charge, or new offence) occurring within two years of release; 2) any reconviction occurring within two years of release; and, 3) any reconviction for a violent offence occurring within two years of release. The study sample was two cohorts representing all federally sentenced adult women in Canada released on day parole, full parole, statutory release, or sentence expiration / warrant expiry between April 1, 2002 and March 31, 2003 ( n = 333) and between April 1, 2003 and March 31, 2004 ( n = 326).

In considering revocations of conditional release for any reason , a rate of return to custody of 37.2% was found for women released in 2002-03, and a rate of 38.0% was found for those released in 2003-04. Rates of conviction for new offences were 28.2% and 29.8% respectively, while conviction rates for violent offences specifically were 4.2 % and 5.2% respectively. The rates of new conviction were consistent with those found using a similar methodology by Bonta and colleagues (2003) for three cohorts of women, though rates of violent reconviction were slightly lower in the present study.

When data were disaggregated by ethnicity, index offence, and type of release, findings were in line with previous research and with expectations. Aboriginal women were more likely than their non-Aboriginal counterparts to have their conditional release revoked and to receive a new conviction. However, there was no difference in the likelihood of receiving a new violent conviction. Women whose index offences included a violent act were more likely to receive a new conviction for a violent offence. Finally, women who had been granted discretionary release (day or full parole) were less likely to receive a new conviction than were those released on statutory release or at warrant expiry.

TABLE OF CONTENTS

LIST OF TABLES

LIST OF APPENDICES

INTRODUCTION

Examining changing rates of recidivism – that is, of return to criminal behaviour – allows interpretation of the impacts of policy and procedural shifts in correctional practice. The considerable variability in reported rates of recidivism, however, complicates this endeavour, most notably with research pertaining to women offenders. The smaller number of offenders in this group, along with changes in women's incarceration patterns in recent years (Research Branch, 2006) adds supplementary challenges in reaching conclusions regarding recidivism rates for women. This study provides an up-to-date benchmark of the rates of recidivism (including revocation of conditional release for any reason, reconviction for a new offence, and reconviction for a new violent offence) for Canadian federally sentenced women offenders.

Women Offenders' Recidivism Rates

As women represent a small proportion of incarcerated offenders, they tend to receive less research attention relative to male offenders. This general trend applies to recidivism research as well. Indeed, several investigators have reported mainly aggregated recidivism rates (i.e., across genders) and have presented gender-specific data only in small, relatively superficial sections (e.g., Langan & Levin, 2002) or in appendices (e.g., Cuppleditch & Evans, 2005).

Those few studies which have focused specifically on women offenders (either exclusively or in tandem with men offenders) have reported a wide range of rates of recidivism. One review summarized recidivism rates of Canadian adult women offenders reported in a variety of studies (Blanchette, 2001); rates of reconviction for new offences ranged from 3.7% to 46.9%. More recent investigations also report results within this range: 16% (Verbrugge, Nunes, Johnson, & Taylor, 2002) and 20.7% (Law, 2004). Such a broad range of rates, however, can be confusing and of limited utility. When estimates change drastically from year to year or from study to study, those who use such estimates (e.g., National Parole Board members, parole officers, community program officers) are challenged to correctly anticipate recidivism. A considerable portion of this variability, however, is directly attributable to measurement issues.

Measurement Issues

Measurement in recidivism studies represents a number of specific challenges. Likely the most important methodological issue, leading directly to differences in estimates of recidivism, is the operationalization of the term. ‘Recidivism' has variously been defined in previous investigations as return to custody for any reason, including technical violations (e.g., Verbrugge et al., 2002), re-arrest (e.g., Benda, 2005), reconviction (e.g., Law, 2004), and re-incarceration (e.g., Deschenes, Owen, & Crow, 2006). Often, more than one definition is used in the same study. When using the first definition – that is, any return to custody – it must be recalled that a return to custody does not suggest, in and of itself, a poor post-release outcome. Indeed, a revocation may represent very appropriate risk management strategies enacted in order to prevent future offending.

Studies involving Canadian offenders have been further complicated by the existence of multiple correctional jurisdictions (federal and various provincial / territorial). Some researchers have defined recidivism as only federal re-incarceration while others have used both federal and provincial re-incarceration. While there is a tendency to use return to federal custody as a measure of recidivism, possibly due to the relative ease with which these data can be obtained, doing so fails to acknowledge that offenders may have been convicted for new offences (resulting in a return to provincial or territorial custody) not captured in federal records.

A further measurement issue is the definition of the population of interest. As mentioned, in Canada , there exist both provincial and federal correctional jurisdictions based on length of sentence. 1 Some researchers have focused on persons released from provincial / territorial institutions (e.g., Rettinger, 1998) while others have focused on those released from federal institutions (e.g., Bonta, Rugge, & Dauvergne, 2003). In the latter group, some researchers have further narrowed their sample by examining only the recidivism rates of first-time federal offenders (Belcourt, Nouwens, & Lefebvre, 1993). As antisocial behaviour is learned (e.g., Andrews & Bonta, 2006) and repeat offenders therefore tend to persist in such behaviour, this method provides an underestimate of the general rate of recidivism.

There is also considerable variability in recidivism rates based on length of follow-up period. Published studies have been based on periods from one year (e.g., O'Brien & Bates, 2005) to as long as 18 years (Putkonen, Komulainen, Virkkunen, Eronen & Lönnqvist, 2003). In choosing a follow-up period, long-range behavioural prediction must be balanced with the desire to produce temporally relevant estimates; many researchers accomplish this goal by using a two-year follow-up period (e.g., Blanchette, 1996; Bonta et al., 2003; Kershaw, Goodman, & White, 1999). Indeed, in one study of women offenders' recidivism involving an 18 year follow-up, the researchers found that of those offenders who did recidivate, 80% did so within the first two years (Putkonen et al., 2003). This calls into question the efficiency of using extensive follow-up periods.

Clearly, decisions regarding methodological issues such as the operationalization of recidivism, definition of the population of interest, and length of follow up can have important impacts on the recidivism rates reported. Given this, consideration must be given to how best to examine recidivism. The “best” examination strategy, however, is fluid in definition, in that research questions, access to data, and operational and procedural realities all combine to influence the optimal research strategy and design of recidivism studies.

Two important guidelines emerge from this review, however. First, measurement decisions must be clearly explained and methodologies must be made explicit. Second, where possible, there must be consistency in the measurement strategies used to examine recidivism across studies. Such consistency would allow greater confidence in jointly considering the results of separate investigations.

One recently published study was admirable in its approach to measurement and methodological issues. Bonta and his colleagues (2003) developed a strong study design, and they clearly reported the measurement-related decisions made in their study, as well as the rationale for each decision. They examined recidivism – defined in two ways: reconviction for new offences generally and for violent offences specifically during a fixed two-year follow up – for three cohorts of Canadian adult federally sentenced men and women released in the three fiscal years from 1994-95 to 1996-97.

Study Rationale

Given that the most recently released individuals in the cohorts used by Bonta and colleagues (2003) comprised data from a decade ago, it seemed appropriate to update the recidivism rates of women offenders reported by these researchers. In order to ensure comparability of findings, the procedures used in the present study were similar to those used by Bonta and colleagues. The examination of more recent data, however, provides an up-to-date benchmark against which to assess any changes in rates of recidivism in the female offender population.

STUDY DESIGN

A retrospective longitudinal design was used to examine recidivism rates among Canadian federally sentenced women offenders.

Definition of Recidivism

Recidivism was defined variously as 1) any revocation of conditional release (i.e., for technical reasons or due to a new charge or offence) within two years of release from a federal institution, 2 2) any new conviction within two years, and 3) any new conviction for a violent offence within two years. It is important to acknowledge that in some cases, an indication of not having recidivated during the two-year follow-period may reflect that the offender had a limited time period during which she could do so. For example, those women returned to federal custody due to revocations or suspensions for technical violations would have had significantly less opportunity to commit new offences or new violent offences during the two-year follow up period. It is also important to note that when a woman received a new conviction after the termination of her sentence (but still within the two year follow-up period) she was assessed as having a new conviction but not as having her conditional release revoked.

Data

Release

Offenders are frequently released more than once on a given federal sentence (e.g., returned to custody due to suspensions or technical revocations and later re-released). Within the current study, the last release during the fiscal year was used as the commencement of the offender's follow-up period. The use of this date ensured comparability with Bonta and colleagues' (2003) previous findings.

Return to Custody

Returns to federal custody due to revocations (either for technical reasons or due to the commission of a new offence) were examined using Offender Management System data. These data were accessed electronically and any revocation occurring during the two-year follow up period was recorded.

Canadian Police Information Centre data were then used to record new convictions. Though there are typically six to twelve month delays before new convictions are actually recorded in these files, a one-year ‘cushion' was left between the end of the second cohort's follow-up period and the beginning of data collection/ coding . In other words, the end of the second follow-up period was one year prior to the date of data collection; this procedure provided sufficient time for new convictions to be recorded, thereby minimizing the likelihood of missing conviction data. New offences were recorded, along with their dates, and were coded as being violent (Schedule 1 or murder) or non-violent. In cases where a woman was convicted for more than one offence simultaneously, only the most serious offence (i.e., the one with the most serious / longest disposition) was recorded.

New convictions occurring within 60 days of release were verified in the Offender Management System to assess whether the conviction corresponded to an offence occurring prior to or after release. 3 These cases (one offender from each cohort) were removed from the analyses both to be consistent with the methodology used in the previous study and because discounting of only this conviction (i.e., the verification of files for subsequent convictions) would mean that the women had decreased opportunities to re-offend due to time in custody.

Deaths and Deportations

Deaths and deportations occurring within the two year follow-up period were examined individually. No offenders were found to be deported during the follow-up period, while three offenders (<1%) from cohort 1 and two offenders (<1%) from cohort 2 died during this time. The cases of offenders who recidivated prior to their death were retained in the samples (one case from each cohort), but offenders who died before having recidivated were removed (two cases from cohort 1 and one case from cohort 2). This procedure was used as it was impossible to ascertain whether these women would have recidivated had they remained in the sample for the full two years.

Sample

The final sample included all federally sentenced adult women released on day parole, full parole, statutory release, or sentence expiration / warrant expiry during the study period. Women released between April 1, 2002 and March 31, 2003 comprised cohort 1 ( n = 333), while those released between April 1, 2003 and March 31, 2004 comprised cohort 2 ( n = 326). 4 These numbers reflect the removal of cases corresponding to convictions for charges for offences which occurred prior to or during incarceration, as well as todeath (i.e., three cases from cohort 1 and two cases from cohort 2; outlined above).

Analyses

Recidivism Rates

Proportions of women recidivating by the end of the follow up period were calculated separately for each cohort and for each of the three definitions of recidivism. These proportions were also calculated separately for Aboriginal vs. non-Aboriginal women, for women whose index offences were violent vs. non-violent, and for women receiving discretionary vs. non-discretionary release. The length of time which elapsed prior to revocation or reconviction was also compared based on ethnicity, index offence, and type of release. For these analyses, only the first reconviction (either violent or non-violent) was used, as offenders returned to custody for non-violent offences would have more limited opportunity to commit a subsequent violent offence than would those remaining in the community. This could impact the length of time elapsed prior to a new conviction for a violent offence, resulting in inappropriate comparisons.

Inter-rater Reliability

As the Canadian Police Information Centre data were coded by hand, a second researcher coded ten percent of cases from each cohort. Using the kappa coefficient for categorical variables and intra-class correlations for continuous variables, inter-rater reliability coefficients were calculated for ten variables. 5 These were found to be very high, ranging from .92 to 1.00, with an average of .99.

RESULTS

Sample Characteristics

The demographic characteristics of the women comprising the research sample were examined ( nCOHORT 1 = 333; nCOHORT 2 = 326). The two cohorts were similar in composition. Women were typically about 35 years old on release ( MCOHORT 1 = 34.6; MCOHORT 2 = 35.1), and, as shown in Table 1, were mainly single, separated, divorced, or widowed. About a quarter of the women in each cohort were of Aboriginal ethnicity.

Table 1. Sample Demographic Characteristics
  Cohort 1 (2002-03) Cohort 2 (2003-04)
Demographic Characteristic n / 333 % n / 326 %
Ethnicity        
Caucasian 196 59 186 57
African Canadian 37 11 29 9
Aboriginal 82 25 91 28
Other / Unknown 18 5 20 6
Marital Status        
Married / Common-law 120 36 129 40
Single / Separated / Divorced / Widowed 209 63 190 58
Other / Unknown 4 1 7 2

Almost half of the women were convicted for at least one violent offence (see Table 2), with the most common offences in this category being assault and robbery. Among those with non-violent offences, women relatively frequently had convictions for property and drug offences. A considerable proportion of the women also had convictions for other non-violent offences, including administration of justice offences.

1 The federal correctional system administers sentences of two years or more.

2 As mentioned earlier, this type of return to custody may more accurately represent appropriate risk management than risk escalation. However, statistics on return to custody are important to population forecasts and ensure comparability of the present results with those of jurisdictions that use this indicator.

3 In rare cases, the offence may have occurred prior to the sentence of interest and charges / convictions brought forward.

4 These sample sizes are larger than the approximately 225 women per cohort in the study conducted by Bonta and colleagues (2003). The difference is likely attributable to the fact that there have been increases in both the overall numbers of federally sentenced women and in the proportion of these women serving short sentences (and therefore being released sooner) since the previous study (Research Branch, 2006).

5 The remaining variables were not assessed for inter-rater reliability as there were no cases with corresponding data for these variables in the subsample of cases coded by the second researcher (e.g., date of death or deportation).

Table 2. Sample Index Offences
  Cohort 1 (2002-03) Cohort 2 (2003-04)
Offence Type n / 333 % n / 326 %
Violent Offences 160 48 161 49
Homicide 15 5 11 3
Attempted Murder 4 1 0 0
Assault 79 24 76 23
Robbery 53 16 51 16
Kidnapping / Confinement 10 3 12 4
Sexual Assault 9 3 20 6
Arson 10 3 2 1
Utter Threats 30 9 25 8
Weapon Offence 28 8 29 9
Non-violent Offences 287 86 284 87
Drug Offence 104 31 102 31
Property 129 39 101 31
Fraud 40 12 45 14
Other 234 70 239 73

Finally, slightly more than half of the women received discretionary release (i.e., day parole or full parole; nCOHORT 1 = 185 [56%]; nCOHORT 2 = 191 [59%]). The remainder ( nCOHORT 1 = 148 [44%]; nCOHORT 2 = 135 [41%]) were released on statutory release, or at their warrant expiry dates. As offenders can be released more than once during their federal sentence, these releases do not necessarily represent each woman's first release on the current sentence.

Rate of Return to Custody

The rates of return to custody for women in each cohort are presented in Table 3. Of women released from federal incarceration in 2002-03 and 2003-04, less than two-in-five had their conditional release revoked during the two year follow-up. Slightly more than a quarter of the women were convicted for a new offence generally (i.e., either non-violent or violent), and about one-in-twenty was convicted for a new violent offence. Rates of revocation and reconviction were very similar across the cohorts. Rates of violent reconviction specifically differed somewhat across cohorts, though this was likely a result of the very low base rate of such reconvictions.

Table 3. Rates of Return to Custody in a Two Year Follow-up
  Cohort 1 (2002-03) Cohort 2 (2003-04)
Return to Custody n / 333 % n / 326 %
Any Revocation 124 37.2 124 38.0
Reconviction 94 28.2 94 28.8
Violent Reconviction 14 4.2 17 5.2

The offences for which the women were reconvicted were recorded; these are presented in Table 4. As explained previously, in cases where a woman was simultaneously convicted for more than one offence, only the most serious was noted. As can be seen, the pattern of reconvictions differed considerably from the pattern of original convictions. More than half of the most serious first reconvictions were property offences or other non-violent offences, with the bulk of the latter being comprised of administration of justice offences such as failure to appear. Violent offences were less frequent, and, notably, there were no reconvictions for homicide, attempted murder, kidnapping / confinement, or crimes of a sexual nature recorded for either cohort.

The dispositions received were also less serious, with only about three-quarters of the reconvicted women receiving custodial sentences ( nCOHORT 1 = 68 [72%]; nCOHORT 2 = 76 [81%]). Moreover, sentences were typically relatively short ( MCOHORT 1 = 6.4 months; MCOHORT 2 = 7.6 months), with less than a tenth, overall, of the reconvicted women being sentenced to two years or more ( nCOHORT 1 = 6 [6%]; nCOHORT 2 = 10 [11%]).

Table 4. Most Serious Offences (First Reconviction)
  Cohort 1 (2002-03) Cohort 2 (2003-04)
Offence Type n / 94 % n / 94 %
Violent Offences 14 13.8 16 17.0
Assault 8 8.5 9 9.6
Robbery 2 2.1 7 7.5
Arson 3 3.2 0 0
Utter Threats 1 1.0 0 0
Non-violent Offences 80 86.2 78 83.0
Drug Offence 15 16.0 15 16.0
Property 31 33.0 24 25.5
Fraud 7 7.5 10 10.6
Other 27 28.7 29 30.8

Note. As this table summarizes only first reconvictions, the rate of violent reconviction is slightly lower than that indicated in Table 3.

Rates of return to custody were also examined for differences according to a number of variables. First, rates of return to custody were examined by ethnicity. As can be seen in Table 5, Aboriginal women had significantly more revocations and reconvictions, generally, than did non-Aboriginal women, though not more violent reconvictions.

Table 5. Rates of Return to Custody (%) by Aboriginal Ethnicity
Return to Custody Non-Aboriginal Aboriginal χ2
Any Revocation      
Cohort 1 (2002-03) 33.1 50.0 7.58**
Cohort 2 (2003-04) 34.5 47.3 4.52*
Reconviction      
Cohort 1 (2002-03) 23.1 43.9 13.19***
Cohort 2 (2003-04) 25.1 38.5 5.70*
Violent Reconviction      
Cohort 1 (2002-03) 4.0 4.9 0.12
Cohort 2 (2003-04) 3.8 8.8 3.27

Note. Non-Aboriginal offenders: nCOHORT 1 = 251; nCOHORT 2 = 235. Aboriginal offenders: nCOHORT 1 = 82; nCOHORT 2 = 91.

* p < .05. ** p < .01. *** p < .001.

Next, women whose index offences were violent were found to be reconvicted more frequently for violent offences than were women with convictions for non-violent offences only, though rates of reconviction generally were not different across groups. The relationship between index offence and revocation was less clear, with a small but significant difference detected in the first cohort, and with non-violent offenders having a lower rate of return to custody. In the second cohort, women convicted of non-violent offences had non-significantly higher rates of return to custody. Together, these results demonstrate that previously violent offenders were more likely to subsequently receive violent reconvictions, but were not more likely to receive other types of reconvictions nor consistently more likely to be revoked.

Table 6. Rates of Return to Custody (%) by Index Offence
Return to Custody Violent Index Offence Non-violent Index Offence χ2
Any Revocation      
Cohort 1 (2002-03) 43.8 31.2 5.59*
Cohort 2 (2003-04) 36.0 40.0 0.55
Reconviction      
Cohort 1 (2002-03) 31.9 24.9 2.02
Cohort 2 (2003-04) 31.1 26.7 0.77
Violent Reconviction      
Cohort 1 (2002-03) 8.1 0.6 11.76***
Cohort 2 (2003-04) 9.3 1.2 10.83***

Note. Violent: nCOHORT 1 = 173; nCOHORT 2 = 165. Non-violent only: nCOHORT 1 = 160; nCOHORT 2 = 161.

* p < .05. *** p < .001.

In considering the type of release received by the offenders, results demonstrated that offenders on discretionary release were between two and three times less likely to receive new convictions than were offenders on non-discretionary release. Those liberated at statutory release or warrant expiry were also more likely to be convicted of violent offences. The difference between the proportion of women in the discretionary release group who were returned due to a revocation and who were returned due to a reconviction suggests that technical revocations (i.e., due to a breach of conditions) were quite common among this group. The opposite pattern of difference was found for the women in the non-discretionary release group; follow-up investigations revealed that this was attributable to these women being convicted for new offences after the expiration of their sentences. On consideration, this is not surprising, since women who do not receive discretionary release typically reach their warrant expiry dates sooner after release than do their counterparts released on parole. This also explains why no differences were found between the two groups in rates of revocation. Specifically, the different periods under federal jurisdiction in the community may mask variation; women who were not on discretionary release would presumably have a shorter time period during which they could have their release revoked for technical reasons.

Table 7. Rates of Return to Custody (%) by Release Type
Return to Custody Non-discretionary Discretionary χ2
Any Revocation      
Cohort 1 (2002-03) 36.5 37.8 0.06
Cohort 2 (2003-04) 32.6 41.9 2.90
Reconviction      
Cohort 1 (2002-03) 46.6 13.5 44.49****
Cohort 2 (2003-04) 43.7 18.3 24.83****
Violent Reconviction      
Cohort 1 (2002-03) 8.8 0.5 13.87***
Cohort 2 (2003-04) 9.6 2.1 9.09**

Note. Non-discretionary release: nCOHORT 1 = 148; nCOHORT 2 = 135. Discretionary release: nCOHORT 1 = 185; nCOHORT 2 = 191.

** p < .01. *** p < .001. **** p < .0001.

Rates of reconvictions occurring prior to and after each offender's warrant expiry date were also examined. 6 As can be seen in Table 8, more reconvictions occurred after warrant expiry, though the pattern was not the same for both cohorts. Notably, these differences were not compared statistically due to differing time periods prior to and after warrant expiry. While more than twice as many reconvictions occurred after warrant expiry relative to before in cohort 1, this difference was much less marked in cohort 2. For violent reconvictions specifically, slightly more reconvictions occurred after warrant expiry in both cohorts; again, however, the low base rates in this category makes this difference difficult to interpret with confidence.

Table 8. Rates of Return to Custody (%) by Warrant Expiry Date
Return to Custody Pre-WED Post-WED
Reconviction    
Cohort 1 (2002-03) 9.9 18.8
Cohort 2 (2003-04) 14.0 15.3
Violent Reconviction    
Cohort 1 (2002-03) 1.9 2.5
Cohort 2 (2003-04) 1.9 3.4

Note. N COHORT 1 = 325; N COHORT 2 = 321 (excludes cases with missing WED information).

A series of follow up analyses were conducted to examine the length of time elapsed in the community prior to revocation or reconviction. Time until violent reconvictions was not examined as offenders reconvicted for non-violent offences had more limited opportunities to commit violent offences than did their counterparts in the community; this difference may have influenced findings. Results demonstrated that there was no difference in this length of time according to Aboriginal ethnicity, type of release, or index offence (see Appendix B). There was a non-significant trend, however, for those who did not receive discretionary release to spend shorter periods of time in the community prior to revocation or reconviction than their counterparts who were released on parole.

6 These calculations were not completed for revocations as offenders past the expiration of their sentences are no longer under supervision and therefore can no longer have their release revoked for technical reasons, thereby rendering the comparison inappropriate.

CONCLUSION

The present study provided a specific and up-to-date indication of the rates of recidivism for Canadian federally sentenced women offenders. After consideration of the advantages and disadvantages of various methodological factors, three definitions of recidivism were selected for examination: revocation of conditional release, conviction for any new offence, and conviction for a new violent offence. The files of all women released in 2003-03 and 2003-04 were examined retrospectively for periods of two years from their dates of release.

In considering revocations of conditional release occurring within two years of release, a rate of return to custody of 37.2% was found for women released in 2002-03, and a rate of 38.0% was found for women released in 2003-04. Rates of conviction for new offences for these cohorts were of 28.2 and 28.8% respectively, while those for violent offences were of 4.2 and 5.2% respectively. In the earlier examination conducted by Bonta and colleagues (2003), rates of new conviction and of new violent conviction ranged from 23.0 to 30.0% and 6.7 to 7.4% respectively for federally sentenced women released in fiscal years 1994-95 to 1996-97. This contrast demonstrates that rates of reconviction have not changed since the earlier study, while rates of violent reconviction have decreased. Consideration of new violent reconvictions, however, must be tentative, as very low base rates of violent re-offending among women means that minor differences, in terms of actual numbers, can result in seemingly impressive changes in proportions. Further exploration of women's recidivism rates in later years will determine whether these small differences demonstrate a changing trend in patterns of recidivism or whether they are simply artifacts of sample size.

When rates of return to custody were considered by ethnicity, index offence, or type of release, findings were consistent with expectations. First, Aboriginal women were more likely than their non-Aboriginal counterparts to have their conditional release revoked and to receive a new conviction (though not to receive a new violent conviction). This is in step with the results reported both by Bonta and his colleagues (Bonta et al., 2003; Bonta, Lipinski, & Martin, 1992) and by others ( Jackson , 1989) . Second, women whose index offence was violent in nature were more likely than their counterparts with only non-violent index offences to be convicted of a subsequent violent offence, suggesting that violence may be a learned behaviour for this group of women. Finally, women receiving discretionary release were less likely to be reconvicted, both generally and for violent offences specifically, than were women not receiving discretionary release. This result is not surprising, since the decision to grant discretionary release is a result of perceived risk of re-offending (National Parole Board, 2007). These findings support the National Parole Board's current risk assessment techniques, suggesting they are appropriately discriminating between women likely and unlikely to recidivate. The results also suggest that community parole officers are ably managing offenders in the community. Offenders receiving discretionary release are typically supervised by a parole officer for longer periods than their counterparts who are not granted discretionary release, and these results indicate that supervision may be contributing to a decrease in anti-social behaviour. From this point of view, this pattern of results is particularly striking; given the longer supervision periods of those in the former group, it could be expected that staff would have greater opportunity to become aware of non-compliance with conditions and revoke release. That this was not the case lends further support to the contention that community supervision contributes to decreased risk.

Though these results offer a standard against which to assess the impacts of policy and procedural shifts in correctional practice, it is important to acknowledge a number of limitations. First, changes in the federal offender population may result in changes in rates of recidivism. Specifically, trends over the last decade indicate that by and large, a greater proportion of women admitted to federal custody are assessed as high risk and as having high levels of criminogenic needs (Research Branch, 2006). If these trends continue, it is likely that rates of recidivism will also change. Second, it is possible that the present definitions may not be suitable in all contexts. For example, it would be inappropriate to compare these rates with those of re-arrest, re-incarceration, and self-reported criminal activity, though these also have advantages as definitions of recidivism. However, the present study was explicit in its definitions and methodology, and therefore offers a clear, unambiguous, and up-to-date indication of rates of recidivism of federally sentenced women offenders in Canada . Moreover, the similarities between this study and that of Bonta and colleagues (2003) allowed for comparisons over time; continued investigations using similar methodologies will provide for a more comprehensive investigation of changing trends.

REFERENCES

Andrews, D.A. & Bonta, J. (2006). The psychology of criminal conduct (4th ed.). Cincinnati, OH: Anderson Publishing Company.

Benda, B.B. (2005). Gender differences in life-course theory of recidivism: A survival analysis. International Journal of Offender Therapy and Comparative Criminology, 49, 328-342.

Blanchette, K. (1996). The relationship between criminal history, mental disorder, and recidivism among federally sentenced female offenders. Unpublished Masters thesis, Carleton University, Ottawa, ON.

Blanchette, K. (2001). Classifying female offenders for effective intervention: Application of the case-based principles of risk and need. Unpublished comprehensive paper, Carleton University, Ottawa, ON.

Bonta, J., Lipinski, S., & Martin, M. (1992). The characteristics of Aboriginal recidivists. Canadian Journal of Criminology, 34, 517-521.

Bonta, J., Rugge, T., & Dauvergne, M. (2003). The reconviction rate of federal offenders. Ottawa, ON: Solicitor General Canada.

Cuppleditch, L. & Evans, W. (2005). Re-offending of adults: Results from the 2002 cohort. Home Office Statistical Bulletin (December 2005) London: Home Office.

Deschenes, E.P., Owen, B., & Crow, J. (2006). Final report: Recidivism among female prisoners: Secondary analysis of the 1994 BJS recidivism data set. Washington, DC: United States Department of Justice.

Jackson, M. (1989). Locking up Natives in Canada. University of British Columbia Law Review, 23, 215-300.

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

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

Law, M. (2004). Federally sentenced women in the community: Dynamic risk predictors. Forum on Corrections Research, 16, 18-20.

National Parole Board (2007). Core values. Retrieved June 15, 2007 from http://www.npb-cnlc.gc.ca/about/corev_e.htm

O'Brien, P. & Bates, R. (2005). Women's post-release experiences in the U.S.: Recidivism and re-entry. International Journal of Prisoner Health, 1, 207-221.

Putkonen, H., Komulainen, E.J., Virkkunen, M., Eronen, M., & Lönnqvist, J. (2003). Risk of repeat offending among violent female offenders with psychotic and personality disorders. American Journal of Psychiatry, 160, 947-951.

Research Branch (2006). The changing federal offender population: Profiles and forecasts, 2006. Ottawa, ON: Correctional Service Canada.

Rettinger, J.L. (1998). A recidivism follow-up study investigating risk and need within a sample of provincially sentenced women. Unpublished doctoral dissertation, Carleton University, Ottawa, ON.

Verbrugge, P., Nunes, K., Johnson, S., & Taylor, K. (2002). Predictors of revocation of conditional release among substance abusing women offenders. Ottawa, ON: Correctional Service of Canada.

Appendix A. CPIC Coding Guide

Cohort: ⟩ 1 ⟩ 2

FPS: ________________ Name: ____________________________________

from cohort information sheet from cohort information sheet

1. Release date: _________________________________________________

from cohort information sheet (yyyy/mm/dd)

2. Follow up end date: _________________________________________________

Release date plus two years (yyyy/mm/dd)

Consider only convictions occurring from Release date to Follow up end date:

3. Did a new conviction occur? ⟩ No ⟩ Yes

4. Offence date: ____________________ (yyyy/mm/dd)

5. Type of offence (most serious offence):

If there are multiple offences, code for the one with the most serious disposition. If two offences receive the same disposition, then code the offence which causes most harm against a person.

Violent offences:

  • Homicide
  • Attempted murder
  • Sexual offence
  • Assault
  • Robbery
  • Kidnap / Confine
  • Weapon offence
  • Utter threats
  • Arson
  • Other: ________________________

Non-violent offences:

  • Property (e.g., theft, B&E)
  • Drug offence
  • Impaired driving
  • Fraud
  • Admin. of justice
  • Other: _____________________

6. Disposition for most serious offence:

Custodial sentence: ______ year ______ month _______day

Probation: ______ year ______ month _______day

Fine: $ _______

If the offence identified above was non-violent, continue reviewing convictions occurring from Release date to Follow up end date. Otherwise, proceed to question 11.

7. Did a new violent offence occur? ⟩ No ⟩ Yes

8. Offence date: ____________________ (yyyy/mm/dd)

9.Type of offence (most serious offence):

If there are multiple offences, code for the one with the most serious disposition. If two offences receive the same disposition, then code the offence which causes most harm against a person.

  • Homicide
  • Attempted murder
  • Sexual offence
  • Assault
  • Robbery
  • Kidnap / Confine
  • Weapon offence
  • Utter threats
  • Arson
  • Other: _______________________

10. Disposition for most violent offence:

Custodial sentence: ______ year ______ month _______day

Probation: ______ year ______ month _______day

Fine: $ _______

Consider the time period from Release date to Follow up end date:

11. Did the offender die during this period? ⟩ No ⟩ Yes

12. Was the offender deported during this period? ⟩ No ⟩ Yes

13. Date of death / deportation: ____________________ (yyyy/mm/dd)

Appendix B. Length of Time in Community Prior to Return to Custody

Length of Time Prior to Return to Custody (in days)
  Cohort 1 (2002-03) Cohort 2 (2003-04)
Return to Custody Yes n Yes M No n No M t Yes n Yes M No n No M t
Aboriginal Ethnicity
Any Revocation 41 218 83 240 0.69 43 224 81 269 1.30
Reconviction 36 386 58 337 1.05 35 335 59 353 0.39
Violent Index Offence
Any Revocation 70 225 54 242 0.54 58 256 66 251 0.14
Reconviction 51 323 43 395 1.59 50 383 44 304 1.78
Discretionary Release
Any Revocation 70 247 54 213 1.14 80 268 44 227 0.23
Reconviction 25 406 69 338 1.34 35 360 59 338 0.48

Note. Analyses include only those offenders who were revoked or reconvicted.

Non-Aboriginal offenders: NCOHORT 1 = 251; NCOHORT 2 = 235. Aboriginal offenders: NCOHORT 1 = 82; NCOHORT 2 = 91.

Violent index offences: NCOHORT 1 = 173; NCOHORT 2 = 165. Non-violent index offences only: NCOHORT 1 = 160; NCOHORT 2 = 161.

Non-discretionary release: NCOHORT 1 = 148; NCOHORT 2 = 135. Discretionary release: NCOHORT 1 = 185; NCOHORT 2 = 191.

All results non-significant.