Research Reports

Managing High-Risk Offenders: A Post-Detention Follow-Up 1995 N° R-39

by
Laurence L. Motiuk Raymond L. Belcourt
Research Division
Correctional Service of Canada
James Bonta
Policy Branch
Ministry Secretariat
January, 1995

TABLE OF CONTENTS

ACKNOWLEDGEMENTS

EXECUTIVE SUMMARY

LIST OF TABLES

I.  INTRODUCTION

II.  A GENERAL PROFILE OF DETENTION CASES

III.  POST-DETENTION FOLLOW-UP

Federal Re-admission Within One-Year

Extended Follow-up: New Convictions

IV.  CHARACTERISTICS OF THE RECIDIVISTS

Federal Re-admission:

Case Characteristics

Criminal History

SIR Risk Categories

Major Admitting Offence

General Recidivism:

Case Characteristics

Criminal History

SIR Risk Categories

Major Admitting Offence

Violent Recidivism:

Case Characteristics

Criminal History

SIR Risk Categories

Major Admitting Offence

Sexual Recidivism:

Case Characteristics

Criminal History

SIR Risk Categories

Major Admitting Offence

V.  RISK, TIME AT RISK AND OUTCOME

VI.  RETOOLING THE SIR

VII.  DISCUSSION

VIII.  REFERENCES

LIST OF TABLES

Table 1. Group Comparisons: Detention Cases and General Penitentiary Population

Table 2. Case Characteristics by Detention Grouping

Table 3. Criminal Conviction History by Detention Grouping

Table 4. Distribution of SIR Risk Categories by Detention Grouping

Table 5. Distribution of Federal Re-admissions by Detention Grouping

Table 6. Distribution of New Conviction Rates by Detention Grouping

Table 7. Case Characteristics of Federal Re-admissions by Detention Groupings

Table 8. Correlations: Case Characteristics and Federal Re-admission

Table 9. Distribution of Federal Re-admissions by SIR Risk Categories

Table 10. Distribution of Federal Re-admissions by Major Admitting Offence

Table 11. Case Characteristics of General Recidivists by Detention Grouping

Table 12. Partial Correlations: Case Characteristics and General Recidivism

Table 13. Partial Correlations: Criminal History and General Recidivism

Table 14. Distribution of General Recidivism Rates by SIR Risk Categories

Table 15. Distribution of General Recidivism Rates by Major Admitting Offence

Table 16. Case Characteristics of Violent Recidivists by Detention Grouping

Table 17. Partial Correlations: Case Characteristics and Violent Recidivism

Table 18. Partial Correlations: Criminal History and Violent Recidivism

Table 19. Distribution of Violent Recidivism Rates by SIR Risk Categories

Table 20. Distribution of Violent Recidivism Rates by Major Admitting Offence

Table 21. Case Characteristics of Sexual Recidivists by Detention Grouping

Table 22. Partial Correlations: Criminal History and Sexual Recidivism

Table 23. Distribution of Sexual Recidivism Rates by SIR Risk Categories

Table 24. Distribution of Sexual Recidivism Rates by Major Admitting Offence

Table 25. Partial Correlations: SIR, revised-SIR and Outcome Measures

ACKNOWLEDGEMENTS

We would like to express our sincere appreciation to all those who assisted us in the execution of this study. At National Headquarters, Shelley Brown and Kelley Blanchette provided considerable support, particularly with regards to the collection of follow-up data. Our thanks are also extended to Pierre Couturier and Ken Hazard of the National Parole Board who provided information on detention orders.

EXECUTIVE SUMMARY

Preventative detention during the period of statutory release (formerly mandatory supervision) is an option for managing potentially violent offenders serving a federal sentence. The 1986 passage of Bill C-67 provided a legal mandate for controlling the automatic release of potentially dangerous inmates before the expiry of their sentence. A previous research effort has been directed at evaluating the effectiveness and efficiency of the legislative provisions ^1,2. However, this study was essentially descriptive and did not deal with the characteristics of post-detention recidivists or the impact of preventative detention on criminal futures. This study explores more fully the issue of post-detention recidivism by introducing two statistical controls: level of risk for re-offending and time at risk in the community.

Historical information was collected from a variety of official sources (National Parole Board, Correctional Service of Canada, Royal Canadian Mounted Police) on all male federal offenders (807) who had detention orders applied to them before December 31, 1991. A comparison between detention cases and the general penitentiary population over the same time period (57,157) revealed the following:

• that the Ontario and Prairies regions had larger proportions of detained inmates relative to general inmates (about 31% versus 25%);
• Aboriginal offenders were over-represented among detainees (25.6% versus 12.8%);
• detained offenders were more likely than the general inmate population to have been convicted of a major admitting violent offence, particularly sexual offenses (48% versus 10%) and assault (16% versus 6%); and
• there was a substantially larger proportion of detained offenders serving sentences between two and four years than the general inmate population (76% versus 39%).

A follow-up sample was assembled of 424 detained offenders who had been released from custody for at least one-year. A further extension of the post-detention follow-up period yielded an average of 4.1 years since release (ranging from 2.2 to 7.8 years) for the study sample.

The follow-up sample was divided into three different groups for comparison purposes:

1. 148 offenders (35%) who had been detained but then had their detention orders 'lifted' before the end of their sentence;
2. 114 offenders (27%) who were 'fully detained' (held in custody for their entire sentence); and
3. 162 offenders (38%) who were detained until the end of their sentence after having their "one-chance" statutory release revoked.

Of the detained offenders who had their orders 'lifted', 41% were released with a residency order, 25% were placed on "one-chance" statutory release, and 34% were granted regular statutory release.

Statistical analyses revealed that average age at first adult conviction for the 'lifted', 'fully detained' and "one- chance" revoked offenders did not differ significantly (average age was 20.1, 21.9, and 20.4, respectively). However, the 'fully detained' population was found to be significantly older at first and current federal admission (average age was 29.4 and 33.4, respectively) than detainees who had their orders 'lifted' (average age was 26.1 and 30.2, respectively) and those detained after revocation of "one-chance" statutory release (average age was 26.5 and 30.9, respectively). As expected, the average amount of time served on detention by 'lifted' offenders was significantly shorter than for those detained until the end of sentence (average time detained was 1.0, 1.5 and 1.5 years, respectively).

An examination of criminal conviction history showed that the total number of criminal convictions for 'fully detained' offenders was substantially lower (an average of 14) than those whose orders had been 'lifted' or "one-chance" revoked (an average of 17 for each). As for type of conviction, 'lifted' and 'fully detained' offenders were more likely to have sex conviction histories (62.8% and 63.2%, respectively) than those detained after a "one-chance" statutory release revocation (56.2%). 'Fully detained' offenders were also less likely to have robbery convictions (35.1%) than 'lifted' and 'one- chance' revocation offenders (50.7% and 44.4%, respectively). Finally, offenders detained after a "one-chance" statutory release revocation were more likely to have assault conviction histories (64.8%) than 'lifted' and 'fully detained' offenders (59.5% and 56.1%, respectively).

Using official records, proxy Statistical Information on Recidivism (SIR) Scale³ scores were calculated for the follow-up sample. Three items were not scored due to insufficient information. Given that detained offenders are a likely group to be single, unemployed at the time of arrest, and have fewer than three dependents, it is unlikely that the scoring of these items would substantially alter risk categorizations. Interestingly, the SIR risk categorizations for half of the detained population ranged from "fair" to "very good" risk. This may be due, in part, to the large percentage of sex offenders in the detention population who have been found to perform considerably better on the SIR⁴. Nevertheless, it was found that the percentage of cases in the poorer risk categories was greater for offenders detained after a "one-chance" statutory release revocation (average SIR score =-5.3), as compared to offenders whose detention orders had been 'lifted' (average SIR score =-4.2) and 'fully detained' offenders (average SIR score -1.3). These differences between the three groups were statistically significant.

The follow-up sample also revealed that 83 (20%) detained offenders were re-admitted to federal custody within one-year of release into the community. 50 for technical violations and 33 for new offenses. In other words, the one-year federal re-admission rate for a new offence was 8.0%. Interestingly, just 6.2% of the 'fully detained' offenders and 7.4% of the "one-chance" revocation offenders returned to federal custody for any reason within one-year of release. On the other hand, the one-year federal re-admission rate for offenders who had their detention orders 'lifted' was 43.9%. A recent investigation found a comparable re-admission rate of 46.6% for statutory releases⁵. Among the 'lifted' offenders, the federal re-admission rate for those released on "one-chance" statutory release was 32.4%, on regular statutory release was 44%, and on statutory release with residency was 50.8%.

Also calculated for an extended (average 4.1 years) follow-up were general, violent and sexual recidivism rates for the detention population. Overall, about 60% of the "detention" offenders were convicted of a new criminal offence, nearly 40% for violent crimes and roughly 15% for sexual offenses. Of special note, statistical analyses revealed that general, violent and sexual recidivism rates across 'lifted', 'fully detained' and "one-chance" revocation groups did not differ significantly. Similarly, average times to conviction did not differ significantly across the three groups (averages = 1.7, 1.7 and 1.4 years, respectively).

A closer look at the nature of new convictions revealed that for about one-third of the offenders, the most serious offence was assault and for nearly one-fifth of the offenders it was sex offenses. Just three offenders, all from the "one-chance" revocation group, were convicted of murder (second degree) during the follow-up period. The remainder of the post-detention offenses was largely property-related. Perhaps the most striking finding was the relatively low rate of post-detention serious re-offending (such as murder, sexual assault, robbery). Considering this population was seen as one of the highest risk group of offenders within the system, a much higher percentage was expected.

Among those re-admitted to federal custody within one-year, there were no statistically meaningful differences in relation to age or amount of time served before/after detention across the groups of detained offenders. In an analysis conducted with each follow-up group, correlational analyses revealed that, for 'lifted' offenders, age at first and current federal admission; number of property and overall convictions, and SIR score risk categorizations were significantly related to federal re-admission within one year. Surprisingly, correlational analyses conducted on 'fully detained' offenders found no significant relationships between any of the variables measured and federal re-admission. However, for "one-chance" revocation offenders, age at first federal admission and number of assault and violent convictions were significantly related to federal re-admission.

With respect to general recidivism, significant differences were found across the three follow-up groups in relation to the amount of time served in custody before and after the detention decision. As expected, offenders who had their detention orders 'lifted' served more time in custody before the decision to detain was made and less time on detention than the other offenders. Partial correlation analyses (controlling for time at risk in the community) revealed that among those whose detention orders had been 'lifted', age at first adult conviction, first and current federal admission; number of robbery, property, drug and total convictions; and SIR scores and risk categorizations were significantly related to general recidivism. Similarly, 'fully detained' offenders had significant relationships between the age-related variables, number of robbery, assault, violent and total convictions, SIR scores (but not risk categorizations) and general recidivism. As for the "one-chance" revocation offenders, age-related variables, number of assault, property, drug, violent and total convictions, and SIR scores and risk categorizations were significantly related to general recidivism. No significant relationships were found in any of the detention group between time served on detention or type of major admitting offence and general recidivism.

As for violent recidivism, a significant difference was found in the amount of time served on detention across the three offender groups. As expected, offenders who had their detention orders 'lifted' served less time on detention than the other offenders. A series of partial correlational analyses (controlling for time at risk in the community) revealed that time served in custody prior to a detention decision and number of robbery, violent and total convictions were significantly related to violent recidivism for the 'lifted' offenders. Partial correlational analyses conducted on 'fully detained' offenders found significant relationships between the age-related variables, number of robbery, assault, violent and total convictions, and SIR scores and risk categorizations and violent recidivism. As for "one-chance" revocation offenders, age-related variables; number of assault and total convictions; and SIR scores and risk categorizations were significantly related to violent recidivism. Again, no significant relationships were found in any of the groups for time served on detention or type of major admitting offence and violent recidivism.

When looking at sexual recidivism, a significant difference was also found in the amount of time served on detention across the three groups. Once more, as expected, offenders who had their detention orders 'lifted' served less time on detention than the other offenders. A series of partial correlational analyses (controlling for time at risk in the community) revealed that for the number of sex and total convictions were significantly related to sexual recidivism for the 'lifted' offenders. Similarly, there were significant relationships between the number of sex and violent convictions and sexual recidivism for 'fully detained' offenders. As for "one-chance" revocation offenders, number of sex convictions was found to be significantly related to sexual recidivism. Curiously, for this group, the amount of time served on detention was significantly and positively associated with sexual recidivism. Notwithstanding the fact that there was a previous history of sexual offending in every case of sexual recidivism, the question necessarily arises as to whether this finding is spurious or indeed worthy of further exploration. At this time, the question remains outside the scope of the present investigation. Of special note, no significant relationships were found in any of the groups between age-related, SIR scores, type of major admitting offence and sexual recidivism.

Using analysis of covariance (a statistical technique which provides tests on means adjusted for covariates) we evaluated the general, violent and sexual recidivism rates across the three offender groups, while controlling for level of risk (SIR score and total number of convictions) and time at risk in the community. No significant differences were observed across the 'lifted', 'fully detained' and "one-chance" revocation groups with respect to any of the recidivism measures. This finding would suggest that amount of time served on detention had no effect on post-release recidivism.

A history of sex convictions was positively and significantly related to new sex convictions post-detention. While such a finding is expected, it runs counter to the manner in which the item is scored in the SIR scale. We decided to adjust the SIR total score by redefining previous convictions to include current sex offence(s), interpreting multiple counts of sex assault as multiple convictions and reversing the weight attached to this factor when present. Consequently, detained offenders who had more than one sex offence on their criminal record were scored in the direction of greater risk than offenders who had either none or only one sex conviction. After controlling for length of follow-up, the revised-SIR was found to be significantly related to post- detention general, violent and sexual recidivism. This suggests that a re-tooling of the SIR to fit the current offender population profile is warranted.

In sum, this study yielded important information on the characteristics of post-detention recidivists and the impact of preventative detention on future criminal behavior. As an option for managing potentially violent offenders until the end of their sentence, the selection of cases for detention could be improved by taking into consideration both volume and type of criminal conviction history, as well as previous exposure/response to the criminal justice system. Not surprisingly, particular attention to past sex offence history (such as number and variety) can improve the selection of sex offenders for preventative detention.

These results validate previous findings on the predictive value of criminal history and point to the consideration of other important case-based variables (such as attitudes, attachments, addictions, etc.) which could improve the selection of potentially violent offenders for preventative detention and provide useful targets for intervention. Given that time served on detention did not reduce the likelihood of violent recidivism and the majority have not re-offended in a serious manner, the need to continue to improve offender risk assessment and treatment procedures becomes evident. Of course, a challenging question remains. How do we encourage offenders to actively participate in programs during the extra time in prison and stay crime-free afterwards?

MANAGING HIGH-RISK OFFENDERS: A POST-DETENTION FOLLOW-UP

I.  INTRODUCTION

     The management of the violent offender presents one of the most serious challenges for the criminal justice system. The challenge entails ensuring the immediate safety of the public through incarceration as well as maximizing long-term safety through risk-reduction programming while the offender is under sentence. Legislative efforts to manage the violent offender have included various forms of preventative detention both at the sentencing stage and at the time of parole eligibility. In Canada, dangerous offender legislation has been in operation since 1977 that allows a court to impose a detention order in a penitentiary for an indeterminate period. More recently, provisions in the Corrections and Conditional Release Act (CCRA; 1992) have allowed for the setting of parole eligibility dates for violent offenders at one half of sentence rather than one third.

     Options for managing violent offenders while under sentence include statutory release (formerly mandatory supervision), and, the subject of this report, the use of detention provisions during the period of statutory release. The detention provisions of the CCRA (formerly the Parole Act) allow the National Parole Board to detain high-risk violent offenders beyond their statutory release date and up to their warrant expiry date. Referrals to the National Parole Board for possible detention are made by the Correctional Service of Canada. The criteria for referral excludes offenders serving life or indeterminate sentences and includes the following: offenders currently serving a term defined by the schedules in the CCRA; an offence which caused the death of, or serious harm (physical or psychological) to, another person; and reasonable grounds to believe that the offender is likely to commit, prior to sentence expiration, an offence causing death or serious harm to another person. When all these criteria are met, the offender must be referred to the National Parole Board for a hearing to consider whether or not a detention order should be imposed. However, if the only criteria met is that of reasonable grounds for the belief that an offender may re-offend in a serious manner, the offender is first referred to the Commissioner of the Correctional Service of Canada. If the Commissioner is satisfied that there may be grounds for detention, the offender is then referred to the Chairperson of the National Parole Board.

     The National Parole Board reviews detention referrals from the Commissioner as to offender's likelihood of committing another "serious harm" offence prior to their warrant expiry date. If the Board is satisfied that the offender may re-offend in a violent manner, then the offender is detained and is not eligible for another hearing until one year after the date of the detention order. Along with ability to detain an offender until warrant expiry, the Board has the further authority to impose a statutory release residency requirement in a community-based facility or to grant an offender a "one-chance" statutory release. If the offender violates supervision conditions while on this "one-chance" type of release and conditional release is revoked, regardless of the reason (technical violation or new offence), detention follows.

     The intent of the detention legislation was clearly to identify and manage the highest risk offenders in the federal correctional system. These are the offenders considered not just simply higher risks to re-offend but to re-offend, with serious consequences. Less serious offenders and/or offenders who pose a high likelihood of recidivism for non- violent crimes are to be managed by the parole and temporary absence processes. Options available to these offenders range from escorted temporary absences to full parole. To better understand the most serious federal offenders and consequently improve risk- reduction management strategies, we undertook a general review of detention cases since implementation of the detention provisions in 1986. In addition, we conducted two post-release follow-ups of detention cases to explore their rate of re-offending and the nature of new offenses. One follow-up involved 424 federally detained offenders who were at risk for a minimum of one year and where the outcome measure was return to federal custody. The second worked with the same sample, but with a 4.1 year follow-up any new conviction (federal or otherwise) as our recidivism measure. This report describes the results.

II.  A GENERAL PROFILE OF DETENTION CASES

     A review of both the National Parole Board's decision data base and the Correctional Service of Canada's offender information system yielded a total of 807 male offenders who had detention orders applied to them since 1986. Four female offenders were also detained, but due to their low frequency only the results for males are reported. Table 1 provides a general description of this group of offenders, along with a comparison to the general penitentiary population (57,157) over the same time period (July 1986 to December 1991).

     When comparing the regional proportions of detained and all penitentiary inmates, the Ontario (30.9% versus 25.6%) and Prairie (31.2% versus 22.3%) regions have greater proportions of detained inmates relative to all penitentiary inmates. This finding could be explained by the increasing number of federal sex offenders found in these two regions in recent years⁶. Aboriginal offenders were also found to be over-represented among detainees.

     Interestingly, the detention and general populations were roughly comparable as to the proportion of offenders serving their first federal term (51.5% and 53.5%, respectively). Moreover, the distribution by type of admission were similar with the majority being admitted for a new offence.

     Consistent with the referral criteria for detention, the vast majority of the offenders (86.5%) had a major admitting offence categorized as violent. Although 13.5% had a nonviolent major admitting offence, this does not mean that these offenders did not fit the criteria for detention. Two explanations are possible. First, and most likely, the term "major offence" is based on the offence with the longest sentence. Thus, an offender with a mix of violent and non- violent offenses, of which the non-violent offence produced the longest sentence, will have this offence coded as the "major offence." The second explanation is that a fuller picture of the offender's criminal history may come to light after admission, indicating the likelihood of another serious offence.

     Compared to the general penitentiary population, detained offenders were certainly more likely to have been convicted of a violent offence. In particular, major admitting sexual offenses were prevalent (48.9% versus 10.5%), as were assault/wounding (16.4% versus 6.3%), manslaughter (7.8% versus 3.6%), and attempt murder (3.5% versus 1.2%). Also noteworthy was the finding that detained offenders were serving relatively shorter sentences compared to the general population. Seventy-six percent of detained offenders had sentences between 2 and 4 years, compared to just 39% of the general population. Although this result may be explained by the high proportion of sex offenses (where sentence lengths are often relatively shorter), it is still leaves unexplained why these offenders are viewed as requiring detention when the courts treated them less severely than others.

Table 1.

Group Comparisons: Detention Cases and General Penitentiary Population

Characteristic	Detention Cases (%)	General Population (%)
Region:
Atlantic Quebec Ontario Praries Pacific	5.2 16.5 30.9 31.2 16.2	10.5 28.4 25.6 22.3 13.1
Race:
Caucasian Aboriginal Black Other	68.4 25.6 2.9 3.1	81.2 12.8 3.5 2.5
First Federal Term	51.5	53.5
Type of Admission:
Warrant of Committal Revocation-Technical Revocation-New Offence Other	77.2 13.1 7.4 2.2	70.1 16.6 8.1 5.2
Major Admitting Offence:
*Sexual Offence *Assault / Wounding *Robbery *Manslaughter Property *Attempt Murder *Kidnapping / Abduction Offensive Weapon Arson Drug Other *Violent Major Offence	48.9 16.4 8.2 7.8 5.9 3.5 1.6 1.6 1.5 0.7 3.9 86.5	10.5 6.3 21.2 3.6 27.7 1.2 0.7 1.4 0.6 9.9 16.9 49.2
Sentence Length:
2 to 4 years 4 to 10 years other	76.1 11.9 13.0	39.0 24.4 36.6

III.  POST-DETENTION FOLLOW-UP

     To ensure that the follow-up period of one year was constant, detention decisions made before December 31, 1991, were chosen as the cut-off point for our selection of the follow-up sample. This post-detention follow-up sample consisted of 424 offenders (52.5% of all detained inmates) who were released for at least one year. A comparison of the follow-up sample to the whole detention sample on the variables listed in Table 1 revealed only one difference. The follow-up sample, as expected, was serving shorter sentences. For example, offenders with sentences between 10 and 20 years comprised 5.1% of the follow-up sample, as compared to 10.5% of the general detention sample. By and large, however, the follow-up sample can be viewed as representative of all detention cases.

     The majority of these follow-up offenders were released at their warrant expiry (276 offenders or 65.1%). Of these, 114 (26.9%) were detained until their warrant expired and 162 (38.2%) were released at warrant expiry after having been revoked on "one chance" statutory release. These 276 offenders served as subjects for the post-detention follow-up. The remaining 148 offenders were also examined. These detained offenders were held at one point, but had their detention orders 'lifted' and were released, before warrant expiry, either on statutory release with residency (61 offenders), "one-chance" statutory release (37) or 'regular' statutory release (50).

     Table 2 presents selected background characteristics for offenders across the three follow-up groups. Statistical analyses revealed that average age at first adult conviction for the 'lifted', 'fully detained', and "one-chance" statutory release revocation offenders did not differ significantly. However, the 'fully detained' population was found to be significantly older at first federal admission and at current admission than the offenders whose "one-chance" statutory release was revoked. As expected, the average amount of time served on detention for 'lifted' offenders was shorter than for those detained until the end of their sentence.

Table 2.

Case Characteristics by Detention Grouping

	Detention ’lifted’ (n = 148)	Fully detained until warrant expiry (n = 114)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 162)	F	p
Age at first adult conviction :	M = 20.1 yrs (SD=6.5) R = 12 - 66	M = 21.9 yrs (SD=7.0) R = 15 - 57	M = 20.4 yrs (SD=6.0) R = 15 - 44	2.7	ns
Age at first federal admission:	M = 26.1 yrs (SD=8.5) R = 17 - 66	M = 29.4 yrs (SD=10.1) R = 17 - 61	M = 26.5 yrs (SD=8.7) R = 17 - 63	5.2	<.01
Age at current admission:	M = 30.2 yrs (SD=9.7) R = 17 - 67	M = 33.4 yrs (SD=10.7) R = 19 - 63	M = 30.9 yrs (SD=8.9) R = 19 - 63	3.7	< .05
Time served to decision:	M = 3.1 yrs	M = 2.5 yrs	M = 2.3 yrs	6.8	<.01
Time served after decision:	M = 1.0 yrs	M = 1.5 yrs	M = 1.4 yrs	17.7	<.001
Time served in custody:	M = 4.1 yrs	M = 4.0 yrs	M = 3.7 yrs	0.9	ns
Note: M = Mean; SD = Standard deviation; R = Range.

     Examination of criminal records found that 95.5% of the detained offenders had a history of convictions for violent offenses. Table 3 presents a breakdown of type of criminal convictions for the 'lifted', 'fully detained' and "one-chance" revocation offenders. Although the type of convictions was quite varied, sex convictions were most prevalent among 'lifted' and 'fully detained' offenders (62.8% and 63.2%, respectively), whereas a history of assault (64.8%) convictions was most prevalent among offenders whose "one-chance" statutory release had been revoked.

Table 3.

Criminal Conviction History by Detention Grouping

	Detention ’lifted’ (n = 148)	Fully detained until warrant expiry (n = 114)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 162)
Homicide:	16.2% M = 0.2 R = 0 - 2	13.2% M = 0.1 R = 0 - 2	15.4% M = 0.2 R = 0 - 2
Sex:	62.8% M = 1.9 R = 0 - 24	63.2% M = 2.1 R = 0 - 16	56.2% M = 1.5 R = 0 - 16
Robbery:	50.7% M = 1.5 R = 0 - 13	35.1% M = 0.8 R = 0 - 11	44.4% M = 1.2 R = 0 - 23
Assault:	59.5% M = 1.7 R = 0 - 13	56.1% M = 1.6 R = 0 - 9	64.8% M = 2.3 R = 0 - 25
Property:	83.8% M = 6.02 R = 0 - 43	80.7% M = 4.8 R = 0 - 39	78.4% M = 6.6 R = 0 - 73
Drug:	25.0% M = 0.6 R = 0 - 13	22.8% M = 0.4 R = 0 - 5	19.1% M = 0.4 R = 0 - 8
Violent:	96.6% M = 5.5 R = 0 - 25	93.0% M = 4.8 R = 0 - 17	96.3% M = 5.3 R = 0 - 27
Total:	M=17.4 MD=14 R = 1 - 58	M=14.4 MD=10 R = 1 - 62	M=17.7 MD=15 R = 1 - 101
Note: M = Mean; SD = Standard deviation; R = Range.

     The General Statistical Information on Recidivism (SIR) Scale reflects a list of 15 risk-related factors found to be significantly associated with the decision to grant or deny parole. Given that the majority of risk-related items in the SIR scale measure exposure to the criminal justice system (such as previous convictions, incarcerations and revocations), a proxy score can be derived from a systematic review of an offender's criminal record and birth date. At the time of the follow-up study, personal information on the employment and family background of detained offenders was not available. Consequently, three items contained in the SIR scale could not be scored: current marital status (+1 for "is married or has common law spouse at time of incarceration"), number of dependents (+2 for "had 3 or more dependents under one roof") and employment status at time of arrest (+1 for "was employed at time of arrest for current offence"). However, considering that detained offenders are a likely group to be single, unemployed and have fewer than three dependents at the time of arrest, these variables may be viewed as constants and it is unlikely that the scoring of these items would substantially alter risk categorizations.

     Table 4 presents a breakdown of the proxy SIR scale risk groupings for detained offenders. The five SIR risk categories are: "very good" (scores ranging from 6 to 27); "good" (scores ranging from 1 to 5); 'fair" (scores ranging from 0 to -4); "poor" (scores ranging -5 to -8); and "very poor" (scores ranging from -9 to -30).

     Although the overall SIR risk categories for one-half of the detained population ranged from "fair" to "very good", it was found that the percentage of cases in the poorer risk category was greater for the "one-chance" revocation offenders (Mean SIR=-5.3) than for the 'lifted' (Mean SIR=-4.2) or 'fully detained' offenders (Mean SIR-1.3). These differences were statistically significant (F=8.7, p<.01).

Table 4.

Distribution of SIR Risk Categories by Detention Grouping

Category	Detention ’lifted’ (n = 148)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Very good	13.6% (n = 20)	14.2% (n = 16)	12.4% (n = 20)
Good	10.9% (n = 16)	26.6% (n = 30)	9.9% (n = 16)
Fair	23.8% (n = 35)	25.7% (n = 29)	18.0% (n = 29)
Poor	20.4% (n = 30)	19.5% (n = 22)	19.3% (n = 31)
Very poor	31.3% (n = 46)	14.2% (n = 16)	40.4% (n = 65)
Total score	M = -4.2	M = -1.3	M = -5.3
Note: M = Mean.

FEDERAL RE-ADMISSION WITHIN ONE-YEAR

     Table 5 presents the rates of federal re-admission within one-year of detention. Within one-year, 83 (19.7%) of 424 released offenders were re-admitted to federal custody, 50 for technical violations (all within the 'lifted' group) and 33 for committing new offenses. In other words, the rate of federal re-admission for a new offence was 8.0%.

     Perhaps most striking is that just 6.2% of the 'fully detained' offenders returned to federal custody within one-year of release. Considering this population is viewed as one of the highest risk group of offenders within the system, we were expecting to find a higher return percentage. Similarly, the 7.4% return rate for "one-chance" revocation offenders is also low.

     The one-year federal re-admission rate for offenders who had their detention order 'lifted' was 43.9%. Among these offenders, the federal re-admission rate for those released on "one-chance" statutory release was 32.4%, on regular statutory release was 44%, and on statutory release with residency was 50.8%.

Table 5.

Distribution of Federal Re-admissions by Detention Grouping: One-year Follow-up

	Detention ’lifted’ (n = 148)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Re-admitted	43.9% (65)	6.2% (7)	7.4% (12)

EXTENDED FOLLOW-UP: NEW CONVICTIONS

     It is possible that the federal re-admission rate is an underestimate of the post-detention recidivism rate since other offenders may have been re-arrested for new offenses but either received a provincial sanction or are still awaiting trial. In addition, recent research on recidivism (especially with sex offenders) suggests that longer follow-ups may produce more accurate recidivism rates. Therefore, an examination of all releases, regardless as to whether they have been "at risk" for a minimum of one year, provides an opportunity to capture those who may have recidivated later on. A search of the RCMP's fingerprint service averaging 4.1 years since release (with a range of 2.2 to 7.8 years) yielded the conviction rates for each detention follow-up grouping (see Table 6). Three cases were untraceable during the extended follow-up period and were removed from this portion of the study.

     Table 6 illustrates that during the post-detention follow-up period, about 60% of the offenders were convicted of a new criminal offence, nearly 40% for violent crimes and roughly 15% for sexual offenses. Interestingly, offenders who had their detention orders 'lifted' and those who had their "one-chance" statutory release revoked had substantially higher rates of general and violent recidivism than offenders who were 'fully detained'. In fact, the "one- chance" revocation offenders were, on average, convicted sooner than the other two groups of offenders (Means=1.4, 1.7 and 1.7 years, respectively), although average times to conviction did not differ significantly.

     A closer look at the nature of new convictions revealed that for about one-third of the post- detention recidivists, the most serious offence was assault and for nearly one-fifth, the most serious offence was sex offenses. Just three offenders, all from the "one-chance" revocation group, were convicted of murder (second degree) during the follow-up period. The remainder of the post-detention offenses were property-related.

Table 6.

Distribution of New Conviction Rates by Detention Grouping

Conviction	Detention ’lifted’ (n = 148)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)	p
General:	66.0% (97)	51.3% (58)	60.3% (97)	ns
Violent:	44.9% (66)	34.5% (39)	43.5% (70)	ns
Sexual:	17.0% (25)	13.3% (15)	12.4% (20)	ns
Note: ns = non-significant.

IV.  CHARACTERISTICS OF THE RECIDIVISTS

FEDERAL RE-ADMISSION WITHIN ONE-YEAR

Case Characteristics.

     Table 7 presents analyses of variance on selected background characteristics across the follow-up groups for federal re-admissions. As Table 7 shows, there were no statistically meaningful differences in any of the age-related or amount of time served before/after detention decision variables among federal re-admissions.

Table 7.

Case Characteristics of Federal Re-admissions by Detention Grouping

Conviction	Detention ’lifted’ (n = 65)	Fully detained until warrant expiry (n = 7)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 12)	F	p
Age at first adult conviction :	M = 18.3 yrs	M = 18.9 yrs	M = 18.0 yrs	0.3	ns
Age at first federal admission:	M = 23.9 yrs	M = 26.9 yrs	M = 21.3 yrs	1.7	ns
Age at current admission:	M = 28.1 yrs	M = 30.4 yrs	M = 27.3 yrs	0.3	ns
Time served to decision:	M = 3.0 yrs	M = 2.0 yrs	M = 2.0 yrs	1.9	ns
Time served post decision:	M = 0.9 yrs	M = 1.0 yrs	M = 1.1 yrs	0.5	ns
Time served in custody:	M = 3.9 yrs	M = 3.0 yrs	M = 3.0 yrs	1.2	ns
Note: ns = non-significant.

Table 8 is a series of correlational analyses between various case characteristics and federal re-admission within one year. For 'lifted' offenders, age at first adult conviction, age at first federal admission and age at current admission were significantly correlated with federal re-admission within one year. These age-related variables were found to be unrelated to federal re- admission within one-year for 'fully detained' offenders and just age at first federal admission was significantly related to outcome for "one-chance" revocation offenders.

     Amount of time served before and after a detention decision, and total time served in custody were found to be statistically unrelated to federal re-admission for each follow-up group.

Table 8.

Correlations: Case Characteristics and Federal Re-admission Within One-year

	Detention ’lifted’ (n = 148)	Fully detained until warrant expiry (n = 114)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 162)
Age at first adult conviction :	r = -.26 (p < .001)	r = -.11 ( ns )	r = -.12 ( ns )
Age at first federal admission:	r = -.22 (p < .01)	r = -.07 ( ns )	r = -.16 (p < .04)
Age at current admission:	r = -.20 (p < .02)	r = -.07 ( ns )	r = -.12 ( ns )
Time served to decision:	r = -.03 ( ns )	r = -.06 ( ns )	r = -.06 ( ns )
Time served post decision:	r = -.13 ( ns )	r = -.14 ( ns )	r = -.15 ( ns)
Time served in custody:	r = -.07 ( ns )	r = -.09 ( ns )	r = -.09 ( ns )
Note: r = Pearson correlation coefficient; ns = non-significant

Criminal History.

     Correlational analyses were conducted between conviction history and federal re-admission within one year. Property (r = .18) and number of convictions (r = .18) were significantly related to federal re-admission within one-year for the 'lifted' offenders. Surprisingly, neither type or total number of convictions were found to be related to this outcome for 'fully detained' offenders. However, for those whose "one-chance" statutory release was revoked, assault (r = .35) and violent (r = .23) conviction histories were significantly related to federal re-admission.

SIR Risk Categories.

     Table 9 breaks down the various SIR risk groupings and presents the corresponding one-year federal re-admission rates. Chi-square analyses revealed statistically significant differences in post-detention recidivism rates among the SIR risk groupings for 'lifted' offenders (X²=12.0, df=4, p<.02). However, there were no statistically meaningful differences among the SIR risk groupings in relation to federal re-admission within one-year for 'fully detained' offenders.

Table 9.

Distribution of Federal Re-admissions by SIR Risk Category

Category	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Very good	15.0% (3/20)	0.0% (0/16)	0.0% (0/20)
Good	31.3% (5/16)	3.3% (1/30)	6.3% (1/16)
Fair	45.7% (16/35)	3.5% (1/29)	0.0% (0/29)
Poor	43.3% (13/30)	9.1% (2/22)	9.7% (3/31)
Very poor	58.7% (27/46)	18.8% (3/16)	12.3% (8/65)
Base rate	43.5%	6.2%	7.5%

Major Admitting Offence.

     The detained offenders were categorized as to whether their major admitting offence was violent and if it was sexual in nature. Please note that there is overlap in the sex offender and violent offender groups which accounts for totals exceeding that actual number of returns to federal corrections. Table 10 presents release outcome for the follow-up sample with respect to the offender grouping, whether the major offence was violent or sexual, and rate of re-admission. For each of the groups, no significant differences were found in federal re-admission rates between offenders with violent and non-violent major admitting offences. Similarly, no significant differences emerged between offenders whose major admitting offence were sexual and non-sexual in nature.

Table 10.

Distribution of Federal Re-admission by Major Admitting Offence

Category	Detention ’lifted’ (n = 147)			Fully detained until warrant expiry (n = 113)			Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex
Released:	(122)	(26)	(61)	(100)	(14)	(63)	(144)	(18)	(66)
Readmitted:	41.8 (51)	53.9 (14)	39.3 (24)	6.0 (6)	7.1 (1)	6.3 (4)	7.6 (11)	5.6 (1)	4.5 (3)
Note: Viol. = Violent; Non-viol. = Non-violent.

GENERAL RECIDIVISM

Case Characteristics.

     Analyses of variance for selected background characteristics of general recidivism across the offender groups are presented in Table 11. Significant differences were found across the groups in relation to the amount of time served before and after the detention decision. As expected, general recidivists whose detention was 'lifted' served more time prior to detention decision and less time detained than the other offenders.

Table 11.

Case Characteristics of General Recidivists by Detention Grouping

	Detention ’lifted’ (n = 97)	Fully detained until warrant expiry (n = 58)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 97)	F	p
Age at first adult conviction:	M = 19.1 yrs	M = 20.2 yrs	M = 19.1 yrs	1.4	ns
Age at first federal admission:	M = 24.0 yrs	M = 26.4 yrs	M = 24.7 yrs	2.3	ns
Age at current admission:	M = 28.6 yrs	M = 30.6 yrs	M = 29.6 yrs	1.3	ns
Time served to decision:	M = 2.8 yrs	M = 2.3 yrs	M = 2.1 yrs	3.4	<.03
Time served post decision:	M = 1.0 yrs	M = 1.4 yrs	M = 1.4 yrs	8.3	<.01
Time served in custody:	M = 3.8 yrs	M = 3.7 yrs	M = 3.7 yrs	0.4	ns
Note: M = Mean; ns = non-significant.

     Partial correlations controlling for length of follow-up between various case characteristics and new conviction during the post-detention follow-up period are set out in Table 12. For all three groups, age at first adult conviction, age at first federal admission and age at current admission were found to be significantly related to post-detention general recidivism. As Table 12 shows, age-related variables were found to be inversely related to general recidivism. This means that the older an offender was at the time, the less likely they were to be convicted of a new crime post-detention. None of the other temporal variables predicted general recidivism. That is, length of imprisonment was unrelated to general recidivism.

Table 12.

Partial Correlations: Case Characteristics and General Recidivism

	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Age at first adult conviction:	r = -.21 (p < .02)	r = -.23 (p < .05)	r = -.29 (p < .001)
Age at first federal admission:	r = -.30 (p < .001)	r = -.28 (p < .01)	r = -.26 (p < .001)
Age at current admission:	r = -.20 (p < .02)	r = -.25 (p < .01)	r = -.20 (p < .01)
Time served to decision:	r = -.16 (ns)	r = .01 (ns)	r = -.13 (ns)
Time served post decision:	r = -.06 (ns)	r = -.04 (ns)	r = -.07 (ns)
Time served in custody:	r = -.12 (ns)	r = -.01 (ns)	r = -.12 (ns)
Note: r = Pearson correlation coefficient; ns = non-significant

Criminal History.

     In moving toward a more meaningful understanding of the relationship between criminal conviction history and post-detention general recidivism, Table 13 presents correlations for the various types of convictions. It can be seen that robbery, property and drug convictions were significantly correlated with the post-detention general recidivism of 'lifted' offenders. However, a history of robbery and assault convictions were significantly related to the post-detention recidivism of 'fully detained' offenders while a history of assault, property and drug convictions were significantly related to the post- detention recidivism of "one-chance" revocation offenders.

Table 13.

Partial Correlations: Criminal History and General Recidivism

Type	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Homicide:	r = -.08 (ns)	r =.07 (ns)	r = -.03 (ns)
Sex:	r =.02 (ns)	r = -.01 (ns)	r = -.04 (ns)
Robbery:	r =.17 (p < .04)	r =.19 (p < .05)	r =.03 (ns)
Assault:	r =.05 (ns)	r =.27 (p < .01)	r =.24 (p < .01)
Property:	r =.25 (p < .01)	r =.04 (ns)	r =.17 (p < .03)
Drug:	r =.23 (p < .01)	r =.04 (ns)	r =.21 (p < .01)
Violent:	r =.14 (ns)	r =.26 (p < .01)	r =.18 (p < .01)
Total:	r =.29 (p < .001)	r =.23 (p < .05)	r =.24 (p < .01)
Note: r = Pearson correlation coefficient; ns = non-significant.

SIR Risk Categories.

     With respect to the SIR scale, correlational analyses controlling for length of follow-up revealed that total SIR scores predicted post-detention general recidivism for all detention cases whose orders were 'lifted' (r = -.33, p<.001), where the offender was 'fully detained' (r = -.25, p<.01) and where the offender's "one-chance" statutory release was revoked (r = -.43, p<.001).

     Table 14 breaks down the SIR risk groupings and post-detention general recidivism rates. Chi- square analyses revealed statistically significant differences in post-detention recidivism rates among the SIR risk groupings for the 'lifted' (X²=22.2, df=4, p<.001) and "one-chance" revocation offenders (X²=32.0, df=4, p<.001). However, there were no statistically meaningful differences achieved among the SIR risk groupings in relation to general recidivism for the 'fully detained' offenders.

     Although the chi-squares indicate statistically significant differences among risk categories within each grouping for general recidivism, it is clear from the table that each risk category is not contributing equal weight. Therefore, a note of caution is warranted. Better discrimination may be achieved by collapsing risk categories. For 'lifted' cases, we have three risk categories: 'very good', 'good + fair' and 'poor + very poor'. For 'fully detained' cases, we have two risk categories: 'very good' versus the rest. Finally, for the "one-chance" revocation group, we have four risk categories: 'very good', 'good + fair', poor', and 'very poor'.

Table 14.

Distribution of General Recidivism Rates by SIR Risk Categories

Category	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Very good	30.0% (6/20)	18.8% (3/16)	15.0% (3/20)
Good	56.3% (9/16)	50.0% (15/30)	43.8% (7/16)
Fair	57.1% (20/35)	65.5% (19/29)	48.3% (14/29)
Poor	86.7% (26/30)	54.6% (12/22)	67.7% (21/31)
Very poor	78.3% (36/46)	56.3% (9/16)	80.0% (52/65)
Base rate	66.0%	51.3%	60.3%

Major Admitting Offence.

     Table 15 presents general recidivism rates for the follow-up sample with respect to whether the major admitting offence was violent or sexual in nature. No significant differences were found in general recidivism rates between offenders with violent and nonviolent major admitting offences. Furthermore, no significant differences in general recidivism rates were evidenced between offenders whose major admitting offences were sexual and non-sexual in nature.

Table 15.

Distribution of General Recidivism Rates by Major Admitting Offence

Category	Detention ’lifted’ (n = 147)			Fully detained until warrant expiry (n = 113)			Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex
Released:	(121)	(26)	(60)	(99)	(14)	(63)	(143)	(18)	(65)
Convicted:	65.3 (79)	69.2 (18)	63.3 (38)	48.5 (48)	71.4 (10)	46.0 (29)	59.4 (85)	66.7 (18)	50.8 (33)
Note: Viol. = Violent; Non-viol. = Non-violent.

VIOLENT RECIDIVISM

Case Characteristics.

     Table 16 presents analyses of variance for selected background characteristics of violent recidivists across the offender groups for the extended follow-up. As expected, a significant difference was found in relation to the amount of time served on detention by the 'lifted' offenders relative to offenders detained until the end of their sentence.

Table 16.

Case Characteristics of Violent Recidivists by Detention Grouping

	Detention ’lifted’ (n = 66)	Fully detained until warrant expiry (n = 39)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 70)	F	p
Age at first adult conviction:	M = 19.6 yrs	M = 19.6 yrs	M = 19.2 yrs	0.2	ns
Age at first federal admission:	M = 24.2 yrs	M = 25.2 yrs	M = 24.8 yrs	0.4	ns
Age at current admission:	M = 28.4 yrs	M = 29.9 yrs	M = 29.3 yrs	0.5	ns
Time served to decision:	M = 2.6 yrs	M = 2.3 yrs	M = 2.1 yrs	1.7	<.03
Time served post decision:	M = 0.9 yrs	M = 1.3 yrs	M = 1.4 yrs	7.7	<.01
Time served in custody:	M = 3.5 yrs	M = 3.7 yrs	M = 3.4 yrs	0.1	ns
Note: M = Mean; ns = non-significant.

     Table 17 sets partial correlations (controlling for length of follow-up) between various case characteristics and violent recidivism. As Table 17 shows, age at first adult conviction, age at first federal admission and age at current admission were found to be inversely related to violent recidivism for the 'fully detained' and "one-chance" revocation groups. In particular, this means that the older an offender was at the time of admission, the less likely they were to be convicted of a violent offence post-detention. For the most part, none of the other variables predicted violent recidivism. It is noteworthy that length of imprisonment (time served in custody) was not related to violent recidivism. It appears that detention has little deterrent effect for this group of offenders. The only exception was for 'lifted' where the amount of time served to prior detention decision was inversely related to violent recidivism. This finding may, in part, be explained by the additional incapacitation experienced by offenders who were returned to federal custody for technical violations of conditional release. Such a situation necessarily confounds the amount of time available to re-offend post-detention. In any event, no statistically significant relationship was found between amount of time detained and violent recidivism.

Table 17.

Partial Correlations: Case Characteristics and Violent Recidivism

	Detention ’lifted’ (n = 66)	Fully detained until warrant expiry (n = 39)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 70)
Age at first adult conviction:	r = -.06 ( ns )	r = -.23 (p < .02)	r = -.19 (p < .02)
Age at first federal admission:	r = -.16 ( ns )	r = -.29 (p < .01)	r = -.18 (p < .03)
Age at current admission:	r = -.14 ( ns )	r = -.22 (p < .02)	r = -.19 (p < .02)
Time served to decision:	r = -.17 (p < .05)	r = -.03 ( ns )	r = -.10 ( ns )
Time served post decision:	r = -.00 ( ns )	r = -.09 ( ns )	r = -.04 ( ns )
Time served in custody:	r = -.15 ( ns )	r = -.05 ( ns )	r = -.09 ( ns )
Note: r = Pearson correlation coefficient; ns = non-significant.

Criminal History.

     Table 18 presents partial correlations (controlling for length of follow-up) for the relationship between criminal conviction history and violent recidivism. A history of violent convictions was significantly related to post- detention violent recidivism for all three offender groups. Interestingly, a history of robbery convictions was significantly related to post-detention violent recidivism for both 'lifted' and 'fully detained' offenders but not for "one-chance" revocation offenders. On the other hand, a history of assault convictions was significantly related to post-detention violent recidivism for the 'fully detained' and "one-chance" revocation offenders.

Table 18.

Partial Correlations: Criminal History and Violent Recidivism

Type	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Homicide:	r = -.02 ( ns )	r = -.04 ( ns )	r = -.04 ( ns )
Sex:	r =.13 ( ns )	r =.03 ( ns )	r =.03 ( ns )
Robbery:	r =.19 (p < .03)	r =.21 (p < .03)	r = -.01 ( ns )
Assault:	r =.12 ( ns )	r =.34 (p < .001)	r =.35 (p < .001)
Property:	r =.13 ( ns )	r =.01 ( ns )	r =.03 ( ns )
Drug:	r =.14 ( ns )	r =.07 ( ns )	r =.04 ( ns )
Violent:	r =.27 (p < .001)	r =.32 (p < .001)	r =.27 (p < .001)
Total:	r =.24 (p < .01)	r =.26 (p < .01)	r = .15 ( ns )
Note: r = Pearson correlation coefficient; ns = non-significant.

SIR Risk Categories.

     As for the SIR scale, partial correlations controlling for length of follow-up showed that total SIR scores predicted post-detention violent recidivism for both 'fully detained' (r = -.27, p<.001) and "one-chance" revocation (r = -.35, p<.001) offenders. However, the SIR was found to be unrelated to post-detention violent recidivism for 'lifted' offenders.

     Table 19 presents the distribution of SIR risk groupings with respect to post-detention rates of general recidivism. Chi-square analyses revealed statistically significant differences in post-detention violent recidivism rates among the SIR risk groupings for 'fully detained' (X²=14.1, df=4, p<.01) and "one-chance" revocation offenders (X²=24.9, df=4, p<.001). However, there were no significant differences among the various SIR risk categories with respect to violent recidivism for the 'lifted' offenders.

     While statistically significant differences among risk categories within each grouping emerged for violent recidivism, it is clear from the table that each risk category is not contributing equal weight. Again, better discrimination may be achieved by collapsing risk categories. For 'lifted' offenders, we have three risk categories: 'very good', 'good + fair', and 'poor + very poor'. For 'fully detained' offenders, we have three risk categories: 'very good', 'good', and the rest. Finally, for "one-chance" revocation offenders, we have four risk categories: 'very good', 'good + fair', 'poor', and 'very poor'.

Table 19.

Distribution of Violent Recidivism Rates by SIR Risk Categories

Risk Category	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Very good	25.0% (5/20)	0.0% (0/16)	10.0% (2/20)
Good	43.8% (7/16)	26.7% (8/30)	31.3% (5/16)
Fair	40.0% (14/35)	51.7% (15/29)	24.1% (7/29)
Poor	60.0% (18/30)	40.9% (9/22)	48.4% (15/31)
Very poor	47.8% (22/46)	43.8% (7/16)	63.1% (41/65)
Base rate	44.9%	34.5%	43.5%

Major Admitting Offence.

     Table 20 displays violent recidivism rates with respect to whether the major admitting offence was violent or sexual in nature. As before, no significant differences were found in violent recidivism rates between offenders with major admitting offences which were either violent (versus non-violent) or sexual (versus non-sexual) in nature.

Table 20.

Distribution of Violent Recidivism Rates by Major Admitting Offence

Category	Detention ’lifted’ (n = 147)			Fully detained until warrant expiry (n = 113)			Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex
Released:	(121)	(26)	(60)	(99)	(14)	(63)	(143)	(18)	(65)
Convicted:	45.5 (55)	42.3 (11)	43.3 (26)	31.3 (31)	57.1 (8)	27.0 (17)	42.7 (61)	50.0 (9)	38.5 (25)
Note: Viol. = Violent; Non-viol. = Non-violent.

SEXUAL RECIDIVISM

     In examining post-detention sexual recidivism, it was found that among the 'lifted, offenders who had recidivated with a sex offence, the average time to conviction was 2.0 years (a range of 20 days to 4.4 years). The average time to conviction for a sexual offence post-detention for "fully detained" offenders was also 2.0 years (a range of 84 days to 5.1 years), while it was just was 1.8 years (a range of 80 days to 4.5 years) for the "one-chance" revocation offenders.

Case Characteristics.

     Table 21 presents analyses of variance for selected background characteristics of sexual recidivists across detention groupings. Again, significant differences were found in relation to the amount of time served after detention decision for sexual recidivists. As expected, the amount of time served on detention was shorter for 'lifted' offenders as compared to the other two groupings.

Table 21.

Case Characteristics of Sexual Recidivists by Detention Grouping

	Detention ’lifted’ (n = 25)	Fully detained until warrant expiry (n = 15)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 20)	F	p
Age at first adult conviction:	M = 20.4 yrs	M = 20.2 yrs	M = 21.7 yrs	0.4	ns
Age at first federal admission:	M = 25.1 yrs	M = 27.1 yrs	M = 25.3 yrs	0.3	ns
Age at current admission:	M = 29.2 yrs	M = 31.0 yrs	M = 31.2 yrs	0.2	ns
Time served to decision:	M =2.7 yrs	M =2.4 yrs	M =2.4 yrs	0.2	ns
Time served post decision:	M =0.9 yrs	M =1.4 yrs	M =1.7 yrs	7.0	<.01
Time served in custody:	M =3.5 yrs	M =3.8 yrs	M =4.1 yrs	0.3	ns
Note: M = Mean; ns = non-significant.

     For the most part, none of the offender characteristics (age at first adult conviction, age at first federal admission, age at admission, time served until/after decision, total time served) predicted post-detention sexual recidivism. The only exception was for "one-chance" revocation offenders. Correlational analysis controlling for length of follow-up revealed that, for these cases, amount of time served after detention decision was significantly and positively related to sexual recidivism (r = .20).

Criminal History.

     Table 22 sets out the correlations between criminal conviction history and sexual recidivism (controlling for length of follow-up). The most significant correlate of post-detention sexual recidivism was previous sex convictions. In fact, there was a previous history of sexual offending for all "one-chance" revocation offenders who were subsequently convicted for a new sexual offence. It should also be mentioned that not all 'lifted', 'fully detained' or "one-chance" revocation sex offenders re-offended sexually.

Table 22.

Partial Correlations: Criminal History and Sexual Recidivism

Type	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Homicide:	r = -.11 ( ns )	r = -.15 ( ns )	r =.02 ( ns )
Sex:	r =.54 (p < .001)	r =.23 (p < .02)	r =.23 (p < .01)
Robbery:	r =.02 ( ns )	r =.07 ( ns )	r = -.05 ( ns )
Assault:	r = -.06 ( ns )	r = -.03 ( ns )	r = -.01 ( ns )
Property:	r =.00 ( ns )	r =.05 ( ns )	r = -.07 ( ns)
Drug:	r =.15 ( ns )	r = -.05 ( ns )	r = -.09 ( ns )
Violent:	r =.39 (p < .001)	r =.22 (p < .02)	r =.11 ( ns )
Total:	r =.12 ( ns )	r =.12 ( ns )	r = -.07 ( ns )
Note: r = Pearson correlation coefficient; ns = non-significant.

SIR Risk Categories.

     An examination of partial correlations controlling for length of follow-up showed that total SIR scores were unrelated to post-detention sexual recidivism. Moreover, chi-square analyses revealed that there were no statistically meaningful differences achieved among the various SIR risk groupings in relation to sexual recidivism (see Table 23).

Table 23.

Distribution of Sexual Recidivism Rates by SIR Risk Categories

Risk Category	Detention ’lifted’ (n = 147)	Fully detained until warrant expiry (n = 113)	Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
Very good	10.0% (2/20)	0.0% (0/16)	10.0% (2/20)
Good	25.0% (4/16)	13.3% (4/30)	18.8% (3/16)
Fair	11.4% (4/35)	24.1% (7/29)	6.9% (2/29)
Poor	16.7% (5/30)	9.1% (2/22)	16.1% (5/31)
Very poor	21.7% (10/46)	12.5% (2/16)	12.3% (8/65)
Base rate	17.0%	13.3%	12.4%

Major Admitting Offence.

     Table 24 presents the distribution of sexual recidivism rates by type of major admitting offence across the various offender groups. There were no significant differences between type of major admitting offence and post- detention sexual recidivism.

Table 24.

Distribution of Sexual Recidivism Rates by Major Admitting Offence

	Detention ’lifted’ (n = 147)			Fully detained until warrant expiry (n = 113)			Revoked on "one- chance" statutory release, then detained until warrant expiry (n = 161)
	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex	Viol.	Non-viol.	Sex
Released:	(121)	(26)	(60)	(99)	(14)	(63)	(143)	(18)	(65)
Convicted:	45.5 (55)	42.3 (11)	43.3 (26)	15.1 (15)	0.0 (0)	17.5 (11)	11.9 (17)	16.7 (3)	18.5 (12)
Note: Viol. = Violent; Non-viol. = Non-violent.

V.  RISK, TIME AT RISK AND OUTCOME

     A series of analyses of covariance (a statistical technique which provides tests on means adjusted for covariates) were conducted to evaluate whether or not there were any significant differences in general, violent or sexual recidivism rates across the three follow-up groups (controlling for level of risk and time at risk in the community). Risk level was established using the proxy SIR scores and total volume of criminal convictions. Time at risk in community was derived from the amount of time which lapsed between date of release from federal custody and date of new conviction. No significant differences were observed across the 'lifted', 'fully detained' and "one-chance" revocation groups with respect to any of the three recidivism measures, after controlling for offender risk level (with both measures run separately) and time at risk. This finding would suggest that time served on detention had little or no effect on post-release recidivism.

VI.  RETOOLING THE SIR

     In order to improve the SIR scale's predictive validity in relation to sexual offending, we explored the feasibility of adjusting the scoring of Item 13, previous convictions: sexual offence(s). Curiously, this item is scored in a direction which is counter intuitive. More specifically, an offender who has only one previous sex conviction is scored more negatively (such as -4), while an individual with either current multiple counts of sexual assault or a repetitive history of sexual offending is scored in the positive direction (such as 0). Notwithstanding that there was no predictive value in previous sex convictions at the time of SIR scale construction, this can be explained by the nature of the sample (offenders released in 1970-1972) used to develop the SIR. This population would not be very representative of more recent federal offender populations.

     As we had found, a history of sex convictions was positively and significantly related to new sex convictions post-detention. While such a finding is in the expected direction, it runs counter to how Item 13 in the SIR is scored. We decided to adjust the SIR total score by redefining previous convictions to include current sex offence(s), interpreting multiple counts of sex assault as multiple convictions and reversing the weight attached to this factor when present. Consequently, offenders who had more than one sex conviction on their criminal record were scored in the direction of greater risk than offenders who had either none or one sex conviction.

     Table 25 presents the partial correlations (controlling for length of follow-up) with regard to the relationship between SIR, revised-SIR and the various outcome measures. For detention cases, the revised-SIR was significantly related to post-detention general, violent and sexual recidivism.

Table 25.

Partial Correlations: SIR, revised-SIR and Outcome Measures

New Conviction	SIR (N = 421)	revised-SIR (N = 421)
General:	r = -.36 (p < .001)	r = -.31 (p < .001)
Violent:	r = -.27 (p < .001)	r = -.27 (p < .001)
Sexual:	r = - .04 ( ns )	r = -.20 (p < .001)
Note: r = Pearson correlation coefficient; ns = non-significant.

VII.  DISCUSSION

     The results of the post-detention follow-up study validated previous findings regarding the predictive value of criminal history and the utility of actuarial devices. Post-detention recidivists were differentiated by the number and variety of criminal convictions as well as previous exposure to the criminal justice system. The following summarizes the major findings:

1. Three different groupings emerged: offenders who had their detention orders 'lifted' before the end of sentence; who were 'fully detained' for their entire sentence; and who were detained until the end of their sentence after "one-chance" statutory release revoked.
2. There was a relatively low return rate to federal custody for any reason within one year of release.
3. An extended follow-up found that 60% of the detained offenders were convicted of a new criminal offence, nearly 40% for violent crimes and roughly 15% for sexual offenses.
4. General, violent and sexual recidivism rates across 'lifted', 'fully detained' and "one-chance" statutory release revocation groups did not differ significantly.
5. Average times to conviction did not differ significantly across the three offender groups.
6. A consistent relationship exists between age, criminal conviction history (number and variety), exposure/response to the criminal justice system and criminal futures.
7. The study reconfirmed the SIR scale's ability to predict general recidivism.
8. Re-scoring the "previous sexual conviction(s)" item in the SIR resulted in better levels of precision in predicting sexual recidivism.
9. Amount of time detained had no effect on post-release recidivism.

     The post-detention follow-up study yielded important information on the characteristics of recidivists and the impact of preventative detention on future criminal behavior. As an option for managing potentially violent offenders until the end of their sentence, the practice of selecting cases for detention could be improved by considering both volume and type of criminal conviction history as well as previous exposure/response to the criminal justice system. Not surprisingly, particular attention to past sex offence history (such as number and variety) can improve the selection of sex offenders for preventative detention.

     These results validate previous findings regarding the predictive value of criminal history and point to the consideration of other important case-based variables (such as attitudes, attachments, addictions, etc.) which could improve the selection of potentially violent offenders for preventative detention and provide useful targets for intervention. Given that time served on detention did not reduce the likelihood of violent recidivism and the majority of detained offenders have not re-offended in a serious manner, the need to continue to improve offender risk assessment and treatment procedures becomes evident. Of course, a challenging question remains. How do we encourage offenders to actively participate in programs during the extra time in prison and stay crime-free afterwards?

VII.  REFERENCES

1. Correctional Service of Canada. (1990). Evaluation of the Detention Provisions of Bill C-67. Ottawa.
2. Corbeil, R. C., Faure, S. D., & Lemaire, D. (1990). The evaluation of Bill C-67, an act to amend the parole act and the penitentiary act. The Canadian Journal of Program Evaluation, 6, 131-156.
3. Nuffield, J. (1982). Parole decision-making in Canada: Research towards decision guidelines. Ottawa.
4. Motiuk, L. L., & Porporino, F. (1993). An examination of sex offender case histories in federal corrections. R-30. Ottawa: Correctional Service Canada.
5. Nouwens, T., Motiuk, L., & Boe, R. (1993). So you want to know the recidivism rate. Forum on Corrections Research, 5, 22-26.
6. Motiuk, L. L. and Deurloo, B. (1993). A statistical profile of Canadian federal sex offenders. Correctional Service Canada.