Abstract
A sample of 2,190 sex offenders seen between 1966 and 2009 was compared on lifetime sexual and all offending, using charges, convictions, court appearances, and self-report as criteria. Of these various criteria, between 47.4% and 81.1% reoffended. Canadian child abuse reporting laws, which came into effect in the 1980s, were associated with increased charges and convictions for offenders, who victimized children, and with a reduction in their longer term reoffense rates. Immigration and population mobility, use of aliases, study follow-up time, and self-reported undetected sex crimes influenced reoffense rates. Results indicate that sex offenders continued to have short prison sentences and/or spend little or no time incarcerated during the latter part of the 20th century.
Criminal history has an important bearing on correctional planning, risk predictions, which are now common in actuarial measures, and in evaluating sex offender treatment success. In a 25-year follow-up study of lifetime sex offender criminal history, Langevin et al. (2004) found that repeat sexual offenses were more frequent than indicated from studies in the literature that usually involves short-term follow ups. They examined a number of reoffending indices in 320 male sex offenders and 31 nonsex offenders seen for psychological assessment between 1966 and 1974. Using lifetime charges, convictions, or court appearances as criteria, they found that three in five men reoffended at some time during their lives and this increased to four in five when all offenses and undetected sex crimes were included. The typical known criminal career spanned almost two decades, indicating that sexual reoffending remained a problem over a significant part of the offenders’ adult lives.
The first purpose of the present study was to extend the 2004 study from 1966 to 2009 to show that long-term follow up of untreated sex offenders will be associated with a significantly higher rate of offending than typically reported in short-term follow-up studies. A second purpose was to examine a number of other factors that may influence criminal history measures. In particular, the study evaluates the contribution of three social changes, which are expected to have had a significant influence on criminal activity measures since the 1970s, namely, Canadian child abuse reporting laws, increased immigration, and population mobility.
Criminal History of Sex Offenders Reported in the Literature
Most of the information on lifetime criminal history of sex offenders comes from recidivism studies and about half of these are treatment studies. Extensive reviews of sexual reoffending studies and their problems are available and only summaries will be offered here (see Cottle, Lee, & Heilbrun, 2001; Furby, Weinrott, & Blackshaw, 1989; Hanson, Bourgon, Helmus, & Hodgson, 2009; Hanson & Morton-Bourgon, 2004; Langevin et al., 2004). Hanson and Morton-Bourgon’s meta-analysis is noteworthy for examining 95 studies on 31,216 sex offenders, with study samples varying from 12 to 5,000, with a mean of 475 cases. They reported that the studies were usually short in duration, averaging less than 6 years. This may be one factor in the relatively low reported reoffense rates of 13% for sex offenses and 36% for offenses in general. Furby et al. (1989) reported that of 42 studies they examined, reoffending varied from 0% to 88%. In a long-term follow-up study, Soothill and Gibbens (1978) found 33% of reconvictions for sexual offenses occurred after 10-year follow up and even 15% after 24 years at large. As noted previously, Langevin et al. (2004) found, in a Canadian sample of sex offenders, that at least three out of five reoffended up to 25 years after their initial offense, indicating that most existing follow-up studies are too short in duration to offer an accurate estimate of lifetime criminal activity and reoffense rates. Because the average criminal career of sex offenders in the Canadian sample spanned almost two decades (Langevin et al., 2004), follow-up studies of at least 20 years or more are expected to provide the most complete assessment of lifetime criminal activity. There have been no studies of sexual reoffending with more than 20-year follow ups reported in the literature since the Langevin et al. study, and issues reviewed in that study appear to be pertinent at present.
Problems in Evaluating Criminal History
Providing reliable and valid measures of criminal history is surprisingly complex and the existing literature is often considered limited in estimating known offenders’ actual lifetime crimes (Furby et al., 1989). The following factors may have a negative effect in attempts to evaluate criminal activity, generally artificially reducing repeat offense rates.
Length of Follow Up
Most studies have follow-up periods that are far too short to evaluate accurate adult lifetime repeat offending. As noted previously, Hanson and Morton-Bourgon (2004) in a meta-analysis found that studies averaged 3- to 6-year follow up, whereas significant numbers of charges and convictions are common after 10 years and even after 20 years at large (Langevin et al., 2004; Prentky, Lee, Knight, & Cerce, 1997; Soothill & Gibbens, 1978). As current studies increase their follow-up times, results on criminal activity approach those from the few long-term studies available. For example, Hanson, Broom, and Stephenson (2004) presented an interesting comparison of 403 treated and 321 untreated sex offenders, followed up for 12 years. Although the groups did not differ significantly, reoffense rates were 21% to 22% for sex crimes, 42% to 45% for violent crimes, and 56% to 60% for general reoffending. The present study examines length of follow up as a factor in lifetime criminal history.
Excluding Juvenile Criminal Records
A number of recent studies on juvenile offenders suggest that incomplete records in part account for the low crime rates reported in the literature (Allan, Allan, Marshall, & Kraszian, 2003; Caldwell, 2010; Fortune & Lambie, 2006; Vandiver, 2006). Caldwell (2010) conducted a meta-analysis of 63 studies that examined 11,219 juvenile sex offenders with an average follow up of 59.4 months. Typical of studies on adult sex offenders with short follow up was the low reoffense rates of 7.1% for sex offenses and 43.4% for other offenses. Noteworthy, however, was the difference in studies that examined juvenile records versus those with adult records only. Those with juvenile records showed four times the reoffense rates of those that relied only on adult records. Allan et al. (2003), who examined reoffending in 326 juvenile male sex offenders for follow ups of 5 to 13 years, found similar results, with 70% having new sexual offenses. A problem in our Canadian research is that juvenile records are typically not available without a court order, so even if researchers wanted to obtain official juvenile records, they may be prevented from doing so. As a consequence, their reoffense rates may be artificially reduced. The present study includes an examination of juvenile records.
Charges Versus Convictions and Court Appearances
Using different criteria, that is, charge versus conviction, is also problematic as the latter may artificially reduce the reoffense rates. Hood, Shute, Feilzer, and Wilcox (2002) were critical of using reconviction rates, a criterion used in 84% of recidivism studies (Hanson & Bussiere, 1998; Hanson & Morton-Bourgon, 2004), as they may underestimate actual crime rates. They noted that it is difficult to secure convictions for sex crimes and arrest rate/charges may be more informative. Rates of charges and conviction are examined in this study to evaluate the Hood et al. hypothesis.
Sexual Versus “Nonsexual Offenses”
A further problem is that charges other than those labeled as sexual offenses may not be considered sexual in nature, and general reoffending is typically more common than sex reoffending in past research. Although researchers conducting sex offender therapy studies may object to the inclusion of nonsex offenses as indices of treatment failure, it may be the case that sexual offenses are plea bargained to violent or other offenses. For example, an offender may acknowledge a common assault, but not a sexual assault, and agree to plead guilty to the former, eliminating the necessity of a trial. Furthermore, a rape-murder may appear only as a homicide, masking the sexual nature of the crime (see Chan & Heide, 2009). At times, offenders attempting sexual assaults may be caught before carrying out the actual offense and they may only be charged with break and enter or causing a disturbance, but may acknowledge to the examining clinician that they had intended to carry out a sex crime. Further complicating the issue, these men often have a criminal history of property offenses that are motivated by financial gain and not necessarily by sexual interest. Moreover, some property offenses may be sexually motivated, for example, an offender breaks into a home and steals only women’s underwear (see Schlesinger & Revitch, 1999). These crimes are often difficult to define as sexual and it was beyond the scope of the present study to do so. Nevertheless, such features of “nonsexual crimes” indicate that a broader range of charges should be examined in addition to those labeled as sexual offenses, which, in themselves, may underestimate the true incidence of sexual crimes. Therefore, sexual and all charges, convictions, and court appearances are examined in this report.
Undetected Crimes
A number of researchers have expressed concern that reported reoffense rates may be underestimates because they do not include undetected/unreported crimes (Cottle et al., 2001; Furby et al., 1989; Hanson & Bussiere, 1998; Hood et al., 2002). Langevin et al. (2004) reported that 88.3% of their sample would be sex offense recidivists if they had been caught, an increase of 27.2% over sex offense conviction rates. Undetected/unreported crimes are examined again in the present study.
First-Time Versus Repeat Offenders
It may be argued that inclusion of offenders, who are recidivists when first seen, may artificially inflate crime rates, as individuals who previously were charged and did not recidivate were not included in the sample. Previous reoffense rates for groups, such as pedophiles in the Langevin et al. (2004) study, were so high (89.8%-94.1%) as to preclude such an explanation, and results suggest that in sampling recidivists and first-time offenders, one is seeing offenders at different times in their criminal careers. However, in this study, first-time and repeat offenders were examined separately to explore these two competing hypotheses.
Official Databases Used
A further concern for Canadian studies is the use of the “gold standard” Royal Canadian Mounted Police (RCMP) database, which misses a substantial number of crimes, especially sex crimes. Hanson et al. (2004) noted an increased reliance on national databases, such as the RCMP’s, with 53% using them, although it is unclear how many studies used the RCMP database alone. Langevin et al. (2004) found that the widely used RCMP national criminal database was incomplete because it did not always include provincial records. Noteworthy, in the sample seen from 1966 to 1974, if the offender had served time in a federal prison (sentences of 2 years or more), the RCMP had records for 80.3% of the cases. However, if the offenders only served time in provincial prisons (sentences less than 2 years) or were sentenced to probation, the RCMP only had 47.5% of the records. This result suggests that provincial institutions are not always reporting cases to the RCMP. It is especially a concern for criminal history research because a 2006 report by Public Safety and Emergency Preparedness Canada found that only 4% of male criminals were sentenced to 2 years or more incarceration and, 87.7% of offenders had sentences of 6 months or less. Significantly, in the Langevin et al. (2004) study of sex offenders, the RCMP database missed 643 convictions, of which 599 (93.2%) were for labeled sexual offenses. Thus, using only the RCMP databases, sex reoffending may be significantly and artificially underestimated. Hanson (2006) suggested that older files for cases, such as those in the Langevin, Curnoe, et al. (2004) study, are inactive and records may not be retained by the RCMP. Thus, it may be argued that one may expect more complete records from recent cases and a loss of information in long-term follow-up using only this database. This hypothesis was examined in the present study. In addition, the RCMP will not release juvenile records to researchers without a court order, as noted previously, adding a further limitation of using this database alone.
Type of Sample
Finally, reoffense rates may show different results depending on the sample used, that is, federal versus provincial prison, general prison samples versus psychiatric samples, or treated versus untreated offenders. Reports on treatment may be biased by selection of cases by prison or hospital staff. There are few reports on acceptance and completion of treatment among sex offenders, although dropouts are often noted. Langevin and Curnoe (2006) reviewed existing studies and, in their own study of 778 sex offenders, found that only 50.6% expressed a wish for treatment, 42% actually attended therapy sessions, and only 13.6% completed even a short course of therapy. A number of factors were associated with treatment desire and completion, but noteworthy here is that, repeat offenders were significantly more likely than first-time offenders to desire treatment and to attend therapy but were no more likely to complete it.
Child Abuse Reporting Laws
Arguably the most important change in the past half century, which may have influenced Canadian criminal activity and reoffense rates, has been the child abuse reporting laws, introduced in the 1980s, requiring professionals and others to report to the Children’s Aid Society the names of any children suspected of having been physically or sexually abused. One may expect the change in legislation to have led to an increase in charges involving child victims. This hypothesis is examined in the present study. Although benefiting children, the reporting laws may have had a second unexpected effect and been instrumental in changing the candor of offenders with treatment providers and clinical assessors. In examining the proportion of offenders who admitted to their current offense at the time of assessment and those who acknowledged having a deviant sexual preference, Langevin and Curnoe (2006) found that there was a significant reduction from the 1960s to the 2000s with a change from 89.9% to 50.4% acknowledging to examining clinician their current offense and from 64.5% to 11.7% admitting to a deviant sexual preference. Thus, one also may expect a corresponding decline in self-reported and undetected offenses over the same time period, forcing greater reliance on official statistics, arrests, and convictions. This hypothesis also is examined in the present study.
Immigration
Another dimension of change over the time period of the present study has been the increased number of immigrants to Canada, who may influence criminal history and reoffending measures, as they may be subject to deportation and consequently their criminal history lost. Statistics Canada (2006b) reported that in 1991 a total of 16.1% of the census population was foreign born and this increased to 18.4% by 2001. Toronto and Ontario, from where the present sample primarily was drawn, had the highest proportion of foreign-born individuals of any city or province, at 43.7% and 26.8%, respectively. Statistics Canada reported that immigration in 2001 was the highest in 70 years. This change in population composition may be expected by chance to be associated currently with more foreign-born individuals in the criminal population, who may be subject to deportation. Current Canadian law requires that criminals, who are not citizens of Canada and have a sentence of 2 years or more, are automatically subject to deportation and it cannot be appealed. This would especially bias federal RCMP statistics, which are more likely to have a record of criminals with sentences of 2 years or more (see Langevin et al., 2004). The inclusion of immigrants in a follow-up study could artificially reduce the reoffense rates. In the present report, this hypothesis is examined and foreign- versus Canadian-born cases are considered separately.
Population Mobility
A further issue is population mobility. Particularly in the Toronto area, where most of the cases in this study were residing, it is easy to travel in 1 or 2 hour from one city to another and to the United States, where courts may not communicate with each other or with Canadian provincial and federal record keepers. Statistics Canada (2006b) reported that, in considering a period of 5 years or more, 44.9% of Torontonians and 41.3% of Ontarians moved residence. Canadians may also travel to other provinces for employment opportunities. Of the movers in Ontario, 22% moved out of the province or out of the country. As provinces do not routinely share criminal history information, such movers among criminals especially may be lost to follow up, artificially reducing reoffense rates. In the present study, a separate record was maintained for individuals who were known to be movers, used aliases, and who committed crimes in more than one city and in more than one province. Unfortunately, examination of foreign criminal records was beyond the scope of the present study.
Hypotheses of the Present Study
In sum, the hypotheses for the present study are as follows:
Hypothesis 1: As in other long-term follow-up studies, relatively higher rates of reoffending, than reported for short-term studies, would be found when lifetime criminal history is measured by charges, convictions, or court appearances.
Hypothesis 2: It is expected that the child-reporting laws introduced in the 1980s would lead to an increase in charges for offenders of children versus offenders of adults.
Hypothesis 3: It is further expected that increased immigration to Canada and population mobility would artificially reduce reoffense rates.
Hypothesis 4: Inclusion of self-reported undetected crimes was expected to increase crime rates but to be less useful at present than in the past when offenders were more candid with examining clinicians.
Hypothesis 5: When first-time offenders at initial assessment are considered separately, it is expected that they would not differ from repeat offenders in criminal activity, when important confounding variables such as length of follow up, time at large, and age are taken into account.
Hypothesis 6: Use of only RCMP records is expected to reduce crime rates, especially for sex offenses, due to a lack of communication between provincial and federal record keepers, although older records may be discontinued or inactive.
Method
Research Participants
A sample of 2,190 men referred to a psychiatric hospital and private forensic clinic for forensic psychiatric/psychological assessment and/or treatment by the court, police, probation and parole services, defense lawyers, or other mental health professionals between 1966 and 2009 was the subject of this study. All sex offenders reported by Langevin et al. (2004) were also included. Cases were classified into groups based on their known current and lifetime criminal charges. The men were classified in terms of the age and gender of victims as well as the type of sexual activity, that is, sexual touching or intercourse, versus courtship disorders such as exhibiting, frottage, and so on. The groups consisted of
61 men charged with possession of child pornography and who did not have any known sexual contact with children
704 incest offenders:
559 against girls
95 against boys
50 against both boys and girls
649 extrafamilial child sexual abusers:
279 against boys
297 against girls
73 against both boys and girls
141 exclusive genital exhibitionists
42 miscellaneous courtship disorders (combinations of voyeurism, exhibitionism, and/or toucheurism/frottage involving adult female victims, see Freund, Scher, and Hucker, 1983)
340 sexual aggressives against adult females (charged with sexual assault or rape of an adult female over 16 years of age)
51 men who sexually assaulted both adults and children
68 mixed gender dysphoric and transvestitic men:
22 heterosexual transgender
28 homosexual transgender
1 bisexual transgender
15 heterosexual transvestites
2 homosexual transvestites
99 miscellaneous sex offenders, including
65 who lacked sufficient information to classify them into the previous groups, and offender groups too small for independent analysis
20 sexual aggressives against men and/or both men and women
7 fetishists
5 sadists
2 Klinefelter’s syndrome
35 sexually polymorphous men, who offended against both males and females and against both adults and children and engaged in a wide range of sexually anomalous behavior
“Child” versus “adult” represents individuals below 16 years of age versus 16 and older and is in accord with current Canadian law, which defines the age of consent as 16. The gender group is infrequently studied for criminal behavior, and it was included here because of the sample’s criminal history, some of which included sexual and violent offenses. Transvestites often have gender dysphoria issues later in life so they were grouped with transgender cases (see Bentler & Prince, 1970; Steiner, Sanders, & Langevin, 1985, for further details).
Procedure and Measures Used
An attempt was made to obtain the lifetime criminal history of men in the sample from their first contact with the criminal justice system to the last known charges in 2009. The first arrest, charge, or conviction was used as the index marker event, and all subsequent charges or convictions were considered instances of reoffending. Two sources of data were used: (a) The RCMP records of national criminal offenses was examined in 1994 and again in 2000 and (b) Hospital records from original assessments and any updates on those assessments were also examined. The RCMP records include sentencing dates, location, charges and counts, length of sentence and probation, as well as charges withdrawn, dismissed, stayed, or acquitted. The hospital records often included provincial and federal criminal records, probation and presentence reports, reports from other hospitals, juvenile criminal records, and offenders’ self-reports, providing overlapping information with the RCMP records. A separate record was kept for the total number of charges identified by the RCMP and the hospital and the offenders themselves for comparison, but overlapping information was counted only once in the analyses. Hospital records contained juvenile criminal histories obtained from official sources as well as from offenders’ self-reports, circumventing the problem of obtaining court orders for such information.
Criminal history was divided into (a) sexual, including charges such as sexual assault, rape, indecent assault, and sexual interference and including sexually motivated homicide; (b) violent offenses that were not indicated as sexual, such as common assault, wounding, and homicide; (c) substance-related charges such as driving while ability impaired, possession of narcotics, and trafficking; (d) property offenses, such as break and enter, fraud, and possession of stolen property; and (e) procedural offenses, such as failure to comply with a court order, unlawfully at large, or violation of parole. In the present study, only sexual and all offenses (the latter including sexual offenses) are reported. A separate record was kept for the number of arrests, convictions, and charges dismissed, withdrawn, acquitted, or that did not come to trial. Total time served and total probation time were recorded when available. Because multiple charges related to a single incident and/or victim may be levied at one time, the number of separate court appearances related to different charges was recorded, independently of the number of charges, providing a different index of the number of incidences of criminal activity over time. This was done separately for sexual and all offenses.
Over 500 variables as well as standard psychological tests were scored for the database. Time span of criminal charges from first to last known charge, the times at large for potential to commit offenses, and the total time of sentences served, regardless of type of charge, were examined for this study. A record was also kept of the number of offenders who disclosed committing sexual offenses that went undetected and/or unreported. Demographic information included age, education, grade failures, placement in special education classes, country of birth, whether the offender lived in more than one city or more than one province, the number of cities and provinces where crimes were committed, and use of aliases. Results were analyzed with SPSS-17.0. Multivariate and univariate analyses of variance and regression analyses, t tests, and likelihood ratios were employed. All the data were not available on every case, and variables with sample sizes differing from the total are noted in the results. The study was approved by the ethics committees of the Clarke Institute of Psychiatry and University of Toronto.
Results
Demographics
There was considerable variation in age of the study participants with a mean of 35.6 years (SD = 12.8) at the time they first were assessed and an age range from 12 to 84 years. Although the mean age differences were statistically significant among groups (F = 20.78, df = 13, 2169, p < .001), there was considerable overlap in the distributions of group ages. The sample averaged 57.2 years of age (SD = 14.4, range = 15-107) at the termination of the study with 26.9% more than 65 years of age and 5.6% more than 80. Death records were not obtained, but a total of 26 were known to be deceased by 2009. The total sample has been studied for an average of 21.6 years (SD = 11.2, range = 0-49). Only 6.3% had follow-up periods of less than 5 years, and 50% were examined for 20 years or more, 24.5% for 30 years or more, and 4.9% for 40 years or more by the termination of the study.
The overall sample averaged 11.3 years of education (SD = 3.8, range = 0-27). Most study participants were high school dropouts (55.6%) and 50.6% had failed a grade in school, usually at the public school level, and 37.9% were in special education classes (vs. 2%-3% for the province in general; see Government of Canada, 2006 and Statistics Canada, 2006a, for more details on provincial statistics). A total of 17.5% had attended university and 16.3% community colleges. Again group differences in education were statistically significant (F = 9.14, df = 13, 2076, p < .001) with considerable overlap of group distributions.
A total of 83.6% of the sample were outpatients, 9.7% inpatients, and 6.7% both. Sexual disorder diagnoses were assigned for 86.7% of the sample and, typical of forensic and prison samples, 9.6% were considered psychotic at the time of assessment, comparable with Canadian correctional samples with 8.9% psychotic or schizophrenic (Wormith & McKeague, 1996). A total of 62 were dangerous offender applicants and 38 had murdered or critically injured their victims, indicating that overall the sample was sexually disordered but generally did not suffer from a Diagnostic and Statistical Manual (DSM) Axis I mental disorder, and most offenders were not considered highly dangerous. Most men were untreated or did not complete treatment in our setting.
Crime Rates
The 2,190 men were charged with a total of 20,659 crimes. They were convicted of 14,551 (70.4%) crimes and 6,108 crimes were dismissed/withdrawn or the offender was acquitted. There were a total of 8,547 labeled sex offenses with 5,918 (69.2%) convictions. For all these men known to have committed some sexual offense, only 41.4% of charges were labeled sex offenses and, of these, 40.7% were convictions for a sexual offense, indicating that the offenders had a more general criminal lifestyle and/or sex offenses were plea bargained to other types of offenses.
Moreover, the total of all sentences for the crimes was generally low with an overall mean of 2.7 years, but 48.7% of the offenders were not sentenced to any time in prison and an additional 23.1% had total sentences of less than 2 years for all their crimes, reflecting sentencing practices in Canada generally (Public Safety and Emergency Preparedness Canada, 2006.). Only 8.4% of offenders had total sentences of more than 10 years. The results are similar for probation time with 45.6% having none and an additional 41.1% with less than 5 years probation in total for all crimes. Only 3.1% spent more than 10 years on probation.
Reoffense Rates
Over the span of their criminal careers, the offenders had appeared in court a mean of 1.7 times (SD = 1.8, range = 0-19) for sex offense charges and 4.6 times (SD = 5.9, range = 0-52) for all offenses. Only 4% had appeared five or more times for sexual offenses and 25% for all offenses. Supporting the first hypothesis of this study, Table 1 shows that the overall reoffense rates for the six measures were generally high and ranged from averages of 52.2% to 81.1%, except for court appearances for sexual offenses at 35%. However, the latter increased to 47.4% when only offenders with at least 20-years follow up are considered (Table 1).
Percentage Reoffense Rates Among Sexual Offender Groups Using Charges, Convictions, and Court Appearances as Criteria
Note: Group differences for all reoffense variables were statistically significant. For sex offense charges, LR = 155.35; convictions, LR = 143.11; and court appearances, LR = 153.31; df = 13 and p < .001 in each case. For all charges, LR = 111.66; convictions, LR = 100.42; and court appearances, LR = 126.22; df = 13, and p < .001 in each case. All results remained statistically significant when cases for 20 or more years follow-up are considered separately.
The subgroups of offenders differed significantly on every measure. The groups with child victims (all extrafamilial child sexual abuser groups, incest offenders of both boys and girls, and mixed adult/child groups) had the highest reoffense rates. The incest offenders of both boys and girls had the highest percentage of repeat offenders of any group for sexual charges and convictions and were second highest on all charges and convictions. The gender group had the least number of repeat offenders for sexual and all offenses. Reoffense rates based on court appearances were generally somewhat lower than indicated by charges and convictions and especially noteworthy for the child pornography and incest offender of girls groups; reoffending based on court appearances was substantially less than indicated by charges and convictions.
Factors Confounding Reoffense Data
Tables 2 and 3 show a number of significant factors, which may influence reoffense results.
Factors Potentially Influencing Reoffending Data in Five Decades
Note: Changes over decades were significant for foreign versus Canadian born, n = 1,981, LR = 20.99, df = 4, p < .001; for movers, n = 2,086, LR = 320.94, df = 8, p < .001. For mean number of cities where crimes were committed, n = 1,372, F = 19.93, df = 4, 1365, p < .001; and number of provinces, n = 1,374, F = 9.21, df = 4, 1367, p < .001.
More Factors Potentially Influencing Reoffending Data
Note: For reporting of undetected sex crimes, n = 1,852, LR = 560.55, df = 4, p < .001; for one or more known aliases versus none, n = 1,443, LR = 36.19, df = 4, p < .001. For duration of criminal history, n = 1,281, F = 6.25, df = 4, 1274, p < .001; and for duration of follow-up from time first assessed to 2009, n = 2,187, F = 9979.68, df = 4, 2181, p < .001.
Immigration
Overall, 22.3% of the offenders were foreign born and may have been subject to deportation and thus, future criminal record information would be lost. The number of foreign-born cases is slightly higher compared with Canadian-born cases generally at 18.4%, but lower for Toronto at 43.7% and Ontario at 26.8%. There was a statistically significant increase in the numbers of foreign-born offenders from the 1960s at 18.6% to 24.1% in the 2000s (likelihood ratio [LR] = 20.99, df = 4, p < .001), reflecting changes in the population at large. It is not known how many were deported as a result of their criminal activities, but 30.4% had sentences of 2 years or more and thus, currently would be subject to deportation, if they had not been granted Canadian citizenship.
There was no significant difference in sex offender group membership among Canadian-born versus foreign-born cases (LR = 15.89, df = 13, ns). However, foreign-born offenders were significantly different on a number of pertinent factors, which might influence reoffense rates. They were older than Canadian-born offenders (M = 38.3 years, SD = 13.7 vs. 35.1, SD = 12.5; t = 4.39, df = 661.73, p < .001), better educated (M = 12.1 years, SD = 4.4 vs. 11.2, SD = 3.7; t = 3.75, df = 596.43, p < .001), had shorter known criminal careers (M = 13.9 years, SD = 9.9 vs. 16.8, SD = 10.6; t = 3.84, df = 1,192, p < .001), and shorter follow-up times (M = 19.3 years, SD = 10.7 vs. 21.5, SD = 11.4; t = 3.76, df = 750.90, p < .001). The two groups differed significantly on all reoffense measures, except sex offense charges, with foreign-born men scoring lower on every measure. When follow-up time was set to 20 years or more, reoffending differed only on court appearances for total offenses. Foreign-born offenders again scored lower at 68.2% repeat offenders versus 76.7% for the Canadian-born men (LR = 5.36, df = 1, p < .05).
Population Mobility
Information on group mobility was available only on 42.7% of the sample, so known movers were compared against those known not to be movers and the “do-not-know” category. The number of movers has also increased over the decades from 13.1% in the 1960s to 38.1% in the 2000s, although the number of cases without this information is large and results should be considered with caution. Moving too would add to problems of reliable reoffense data.
There were significant differences in mobility, with 52.5% of foreign-born men versus 29.3% of Canadian-born men moving residence (LR = 82.79, df = 2, p < .001). Thus, foreign-born cases were somewhat closer to the Toronto and Ontario population norms than Canadian-born offenders (44.9% and 41.3%, respectively, moved residence; Statistics Canada, 2006b). Crimes were committed in an average of 2 cities (SD = 1.8, range = 1-15, n = 1,372) and 1.2 provinces (SD = 0.5, range = 1-5, n = 1,374). Most crimes (57.9%) were committed in one city and one province (85.8%), but the 14.1% who committed crimes in other provinces is noteworthy.
When movers are compared with nonmovers on reoffense measures, only court appearances for sex offenses and total offenses were significant. Movers were more likely to show both more sexual reoffending (33.2% vs. 25.2%, LR = 4.04, df = 1, p < .05) and total reoffending (65.4% vs. 56.3%, LR = 4.67, df = 1, p < .05).
Undetected and Unreported Crimes
Table 3 shows that the number of offenders disclosing undetected/unreported sex crimes to examining mental health workers has declined significantly from 78.3% in the 1960s to 7.8% in the 2000s, rendering the inclusion of such undetected offenses in 2000 of marginal utility compared with the 1960s and 1970s. For example, although including undetected crime as one incident of sexual reoffending increased the rate for charges by 26.3% in the 1960s, it changed only by 1.5% in the 2000s. Results were similar for convictions and court appearances for all offenses.
Use of Aliases
Aliases (n = 1,275) were used by 9.4% of offenders, ranging from 0 to 7 names. Of these, 6.9% used one, 1.9% used two, and 0.6% used three or more aliases. The number of cases using aliases had declined significantly from 17.3% in the 1960s to 1.2% in the 2000s. When offenders using no alias versus those using any alias are compared with reoffending, the latter scored higher on every measure, except sex offense charges and convictions. The alias group, more often than those not using aliases, appeared in court for sex offense (47.0% vs. 37.7%) and for all charges (88.1% vs. 68.2%). The men using aliases offended significantly in more number of cities (M = 3.5, SD = 2.7 vs. 1.9, SD = 1.6; t = 5.91, df = 121.63, p < .001) and more provinces (M = 1.5, SD = 0.8 vs. 1.2, SD = 0.5; t = 4.50, df = 121.55, p < .001) than those men not using aliases.
Study Follow-Up Times
As may be expected, the mean of follow-up time to study reoffending in the groups changed significantly from the 1960s at 41.1 years to only 5.1 years on an average in the 2000s. Noteworthy, the average span of criminal activity had changed considerably less, albeit significantly over the decade, with criminal history spanning an average of 19 years in the 1960s and 15.4 years in the 2000s. To examine a possible role for follow-up time in the reoffense rates, cases were examined separately with more than 20 years’ follow-up time. All group differences remained statistically significant. The overall reoffense rate for sex and all charges and convictions changed only by 1% to 2%. However, sexual reoffending court appearances increased from 35% to 47.4% for cases with at least 20-year follow up (Table 1). For all reoffending, the corresponding results changed from 63.8% to 74.5%. Thus, it seems that length of follow up played some role in the results for court appearances. Results were essentially unchanged by considering time at large (minimum of 20 years examined), excluding men known to be deceased and men above 80 years of age and, separately, above 65 years of age at the end of the study (maximum change in reoffense rate, 3%).
Potential Role of Child Abuse Reporting Laws
To examine the possible role of reporting laws enacted in the 1980s, offenders were grouped into exclusive offenders of children (all extrafamilial child sexual abusers and incest groups) versus exclusive offenders of adults (exhibitionists, sexual aggressives, and mixed courtship disorders). A 4 × 2 MANOVA was performed with decade and offender group (adult vs. child victims) as independent variables and the six reoffending measures as dependent variables. MANOVA was significant for both main effects (decade and offender group) and their interaction. To examine the role of reporting laws, rates for the two offender groups were compared within each decade using likelihood ratios. Table 4 shows that there was no significant difference between the groups in charges for sexual reoffending until the 1990s when those victimizing children were more often charged again than those victimizing adults (73.6% vs. 57.3%, LR = 12.53, df = 1, p < .001). This pattern of results was also significant in the 2000s (63.5% and 46.3%, LR = 5.15, df = 1, p < .05). Results were also reflected in sex offense convictions: for the 1990s (52.1% vs. 41.9%, LR = 4.26, df = 1, p < .05) and 2000s (58.0% vs. 35.8%, LR = 8.30, df = 1, p < .01). Note also that the offenders of children had significantly more convictions in the 1960s as well. Results for court appearances related to sex offenses, however, were not concordant with those for charges and convictions. Although the two groups did not differ in court appearances in the 1960s and 1970s, in the 1980s offenders of children had significantly fewer court appearances than offenders of adults (37.8% vs. 50.3%, LR = 6.61, df = 1, p < .01). This pattern of results continued into the 1990s (22.4% vs. 35.8%, LR = 9.22, df = 1, p < .01) but not in the 2000s when the groups did not differ significantly again (18.2% vs. 15.4%, LR = 0.23, df = 1, ns). Results for total offenses (Table 5) indicate that the two groups did not differ until the 1980s when they were significantly different on each measure. The groups only differed on court appearances in the 1990s and only on charges and convictions in the 2000s for total offenses.
A Comparison of Sexual Reoffense Rates Among Offenders Against Children Versus Adults by Decade
Note: For the 4 × 2 MANOVA for decades by adult versus child offenders with all reoffense measures in Tables 5 and 6 as dependent variables: λ = 0.893, F = 8.64, df = 24, 6305.08, p < .001 for decade; λ = 0.980, F = 6.26, df = 6, 1807, p < .001 for adult versus child offender groups; and λ = 0.959, F = 3.19, df = 24, 6,305.08, p < .001 for the interaction of decade by offender group. Univariate results (not presented) were significant for all reoffense measures for decade and either the main effect or interactions were significant for child versus adult groups and the interaction of offender group by decade.
p < .05. **p < .01. ***p < .001 for likelihood ratios comparing child versus adult offenders within the decade.
A Comparison of All Reoffense Rates Among Offenders Against Children Versus Adults by Decade
p < .01. ***p < .001 for likelihood ratios comparing child versus adult offenders within each decade.
Comparing Factors Confounding Reoffense Rates
To evaluate the relative contribution of each confounding factor to reoffense rates, a stepwise regression analysis was performed for the six repeat offending measures separately with the following independent variables: age at first admission and at the end of the study, education, grade failures, special education placement, ethnicity, movers, number of cities and provinces where crimes were committed, use of aliases, study follow-up time, span of criminal career, time at large, undetected crime, and child versus adult offenders (as reported in Table 4). Pairwise deletion of missing data was used.
Ethnicity, movers, use of aliases, and grade failures did not enter any analysis as statistically significant. Independent variables for sex offense charges and convictions explained less than 5% of the reoffending variance and those for total charges explained only 8.1%. These will not be considered further. Follow-up time, number of cities where crimes were committed, and education were significant for sex reoffense court appearances and the three total reoffense measures. For sex offense court appearances, four variables entered the regression equation, in the following order:
Follow-up time
Number of cities, where crimes were committed
Undetected crime
Special education placement
Total R = .36 (13.3% association)
For all convictions:
Number of cities
Education
Follow-up time
Undetected crimes
R = .36 (13.2% association)
For all court appearances:
Number of cities
Education
Time at large
Age at end of the study
Span of criminal career
Follow-up time
R = .49 (24.2% association)
Availability of RCMP Records
There were RCMP records for 65.1% of the cases (n = 1,752, Table 6). The RCMP records missed 2,142 convictions for this sample and 1,366 (63.8%) were for sexual offenses. Information on sentence length was available for 1,429 cases. If the offender served time in a federal prison, that is, had a sentence of 2 years or more, the RCMP files were significantly better and had records for 90.5% of the cases (Table 6). When the offenders only served time in provincial prison or were sentenced only to probation, the RCMP had records for 70.1%. A 2 × 4 ANOVA, with sentence length (less than 2 years vs. 2 years or more) and decades as independent variables and RCMP record availability as dependent variable, showed a significant effect for sentence length (F = 71.87, df = 1, 1,421, p < .001) and decade of admission (F = 21.37, df = 3, 1,421, p < .001) and a nonsignificant interaction effect of decade by sentence length (F = 2.23, df = 3, 1,421, ns). The RCMP records were most complete in the 1980s and least complete in the 1960s, offering some support for Hanson’s (2006) claim about inactive files. However, in every decade, the RCMP had significantly fewer records if there was no sentence exceeding 2 years, also supporting the hypothesis that provincial record keepers do not always communicate with the RCMP.
Available RCMP data by sentence length and decade
Note: RCMP = Royal Canadian Mounted Police. Comparing availability of records in general (column 3), LR = 101.04, df = 3, p < .001. For 2 × 4 ANOVA with sentence length and decade as independent variables and availability of RCMP records as dependent variable: F = 71.87, df = 1, 1421, p < .001 for sentence length; F = 21.37, df = 3, 1421, p < .001 for decade; and F = 2.23, df = 3, 1421, ns for the interaction effect. RCMP records were not systematically collected over the 2000s and files for the 2000s were not included in the above analysis. Missing data for sentence length (n = 1,752 vs. 1,429) accounts for difference total records versus those by decade and sentence length.
Repeat Versus First-Time Offenders
Because some men were already repeat offenders when initially assessed by the authors, they might bias the overall reoffense rates. Table 7 shows that first-time offenders had lower reoffense rates than the total sample on every measure, even when only cases with at least 20 years follow up are considered separately. Although sex reoffense charges and conviction rates were 8.3% and 9.2% lower for first-time offenders than for the total sample, court appearances were 18.4% lower. The disparity in total reoffending was greater: 15.2% for charges, 20.9% for convictions, and 35.4% for court appearances. However, the first-time offenders also differ from repeat offenders on a number of pertinent measures that significantly influenced the results. Repeat offenders were statistically and significantly younger than the first-time offenders at the time of initial assessment (34.70 vs. 36.73, t = 3.65, p < .001) and, over their life span, were studied longer (22.95 vs. 19.87 years, t = 6.43, p < .001), had less time at large (29.75 vs. 36.09 years, t = 10.12, p < .001), had a longer criminal career (17.49 vs. 11.26 years, t = 9.75, p < .001), and more often reported undetected crimes (30.7% vs. 24.3%, LR = 9.55, p < .002). These factors were entered as covariates into a multivariate analysis of covariance with the independent measure being repeat versus first-time offenders and the six dependent measures: charges, convictions, and court appearances for each sexual and all offenses. Differences between the repeat and first-time offenders were not significant for sexual offenses but were for total offenses. Total reoffense rates, however, was high for all measures and the differences between the first-time and repeat offenders were small for charges (1.1%) and court appearances (6%) but more substantial for convictions (15.3%). Results support the hypothesis that the total sample is examining individuals at different times in their criminal careers and, with long enough follow up and correction for confounding variables, differences at initial assessment between first-time and repeat offenders in reoffense rates are small for total crimes and are not statistically significant for sex crimes.
A Comparison of Repeat and First-Time Offenders
Note: For multivariate analysis of covariance, Wilks’s λ = 0.913, approximate F = 18.09, df = 6, 1136, p < .001. Univariate comparisons of first-time and repeat offenders were statistically significant only for all charges, F = 6.53, p < .01; all convictions, F = 82.46, p < .001; and for all court appearances, F = 38.15, p < .001.
Discussion
This study showed a consistent pattern of criminal behavior among sex offenders from the 1960s to 2009. Lifetime criminal activity remained highest overall for offenders against children (extrafamilial child sexual abusers, incest offenders, and adult/child mixed offenders). Offenders who crossed gender lines and who offended both against adults and children had the highest reoffense rates (extrafamilial offenders of both boys and girls, incest offenders of both boys and girls, and adult/child mixed offenders). Reoffense rates based on charges, convictions, or court appearances for sex offenses or total offenses were substantial and generally above 50%, consistent with other long-term follow-up studies (e.g., Prentky, Lee, Knight, & Cerce, 1997; Soothill & Gibbens, 1978) and with studies that examined juvenile records. It seems, however, that court appearances for sexual reoffending presents a different index of persistent crime from the other measures in this study. It is more time sensitive than the other measures and provides more clearly separated incidents of sex reoffending. Individual charges and convictions may not represent different incidents, whereas court appearances usually do. Nevertheless, each measure had its strengths and weaknesses, and it is recommended that all measures in this study be examined separately in research studies.
The study also supported Hood et al. (2002), who noted the difficulty in obtaining convictions for sex offenses. Only 41.4% of charges for a known sex offender sample were labeled as sex offenses and of these only 70% led to convictions. This is especially true of sexual aggressives, typically the most violent of sex offenders, who only had 21.8% of their charges as labeled sex offenses and an average conviction rate of 71.9%. These results argue for a more detailed examination of the incidents involved in charges, use of charges as well as convictions, and especially to examine nonsexual offenses.
The present study supports results of Langevin et al.’s (2004) that long-term follow up produces much higher reoffense rates than typically seen in short-term prospective studies. The mean span of criminal career for the present sample was more than 16 years and ranged up to 52 years, and cases were followed up on average for more than 21 years and up to 49 years. Noteworthy was the greater percentage of crimes labeled as nonsex crimes (only 41.4% of all charges were labeled as sex offenses), suggesting that these men have a more generally criminal lifestyle and/or sex offenses were plea bargained to other categories than sexual offenses. It also emphasizes the need to consider nonsex crimes when evaluating criminal history of sex offenders and in evaluating treatment effectiveness.
A significant change over the decades was in the candor of offenders concerning undetected/unreported crimes they committed. In the 1960s, inclusion of undetected crimes increased the reoffense rate for charges by 26.3%, but only by 1.5% in the 2000s, indicating that this measure is not as useful now as it was in the 1960s to1980s. This change may reflect to some degree the reporting laws concerning child abuse, which may have changed the candor of criminals with assessors.
There has been little change in incarceration practices for sex offenders over the last half of the 20th century and not surprising, time at large had little effect on reoffense rates for most groups; sexual aggressives, miscellaneous courtship disorders, and adult/child mixed groups excepted. In general, prison sentences averaged 2.7 years and probation 2.1 years. Furby et al. (1989) noted more than 20 years ago that sex offenders spent little time in prison and this continues to be the case for the majority of sex offenders in our study. Many (48.7%) spent no time in jail and 23.1% had been sentenced to a total of less than 2 years. This, however, is changing with current Canadian government policy demanding mandatory time in jail for offenders, for example, for possession of child pornography, so results may change significantly over the next decade.
The influx of immigrants over the decades may have influenced reoffense rates in complex and subtle ways that require more detailed study. The immigrants in our sample were older and better educated than Canadian-born offenders and had shorter follow-up times and shorter known criminal careers. The two groups did not differ in sex offender group membership or sentence length, but 30.4% of immigrants had sentences more than 2 years and it is unknown whether they were deported. Immigrants reoffended less on every measure in this study, but when cases with at least 20-year follow up are considered, only total reoffending was significant, again with foreign-born men scoring lower. Considering that sentence length was less than 2 years for 71.8% of the sample and 96% of the general population (Public Safety Report, 2006), it would appear that deportation only presents a problem in measuring repeat offending for a minority of cases.
Population mobility had increased over the five decades for the present sample, and it seems crimes, committed in more than one city and/or province, have affected on reoffense rates. Use of aliases also played a role in results, but it has declined in frequency over the past 40 to 50 years for the present sample. However, it is noteworthy that Langevin (2008) reported that designated dangerous offenders (n = 62) committed their crimes in significantly more cities and provinces than nondangerous offenders (93.5% vs. 42.9% for cities and 35.5% vs. 14.4% for provinces). Thus, mobility may be especially important for smaller but significant and dangerous subgroups of offenders.
Many changes have occurred over the past 40 years, and the present analysis is not meant to imply that only the child abuse reporting laws introduced in the 1980s had affected both charges and convictions of offenders victimizing children. As Furby et al. (1989) noted, social changes are problematic for long-term follow-up studies and may influence results in subtle and unknown ways. In the present report, number of charges and convictions for offenders victimizing children have diverged from offenders of adults and has increased in the 1990s and 2000s. However, the number of court appearances had declined and may reflect a reduction in long-term reoffending. Court appearances for sentencing/judge’s decision is a more time sensitive measure than charges and convictions, and it is possible that longer term follow up will also show changes in this measure as well. However, the greater number of offenders detected and facing criminal prosecution may also have reduced the longer term reoffense rates as well.
RCMP records continue to show more omissions if the offender served only provincial time/no time in jail (less than 2 years), but the 1960s showed the greatest deficits supporting Hanson’s (2006) claim about inactive files. Nevertheless, there is a consistent pattern of results indicating that use of only RCMP records in research may lead to missing a substantial number of sex and other offenses in every decade. Noteworthy, Trevethan, Crutcher, and Moore (2002) reported that 33% of dangerous offenders and 43% of long-term offenders never received a previous federal prison sentence prior to their dangerous offender applications, indicating the importance of combining provincial and federal criminal data sources. The present study suggests that using both federal RCMP and provincial records would capture 85.9% of cases but may exclude the most dangerous and mobile cases. Nevertheless, it provides a more complete picture of actual reoffense rates.
Sexual arousal and release are persistent features over the lifetime of most men, including sex offenders. The typically reported low reoffense rates in sex offender treatment studies may create unrealistic expectations among both offenders and therapists that once they are treated reoffending is unlikely or has a low probability of occurrence. The present report suggests that sex offenders need support and continued contact with therapists over an extended period of their lives, at least for more than 20 years. The Circle of Support and Accountability (Wilson & Picheca, 2005) provides a model of intensive and continued care for sex offenders, which the present study suggests should be the mode for sex offender treatment.
The full treatment history of the offenders studied was not documented and a treated offender group, followed up for a similar length of time as the present study, might be expected to show different results. Marshall (2006) has complained about the absence of long-term follow up of untreated sex offenders for comparison of treated cases. As noted previously, Langevin and Curnoe (2006) found that only 13.6% of their sex offender sample drawn from the present study completed therapy, indicating that therapy may have influenced only a minority of cases in the present study. The present results should not be construed as contradicting prospective recidivism studies, such as those on treatment effectiveness, but rather as suggesting that such studies may wish to consider separating out various repeat offending measures, which may contribute new information to their results. Court appearances, in particular, seem different from charges and conviction indices and to be more time sensitive. The present study also indicates that a largely untreated sex offender group may offend in significant numbers for more than 20 years after assessment, so follow-up times should be at least of this duration for more accurate reoffending information. It also indicates that self-report of criminal activity may be less valuable than it was in the 1970s and official statistics should be considered more reliable. It is unfortunate that the often used, valuable, and informative RCMP database misses so many sex offenses mainly due to lack of communication between provincial and federal record keepers.
The number of court appearances was more than five times for 4% sex offenses and 25% of all offenses. Some men reoffended up to 19 times for sexual offenses and 52 times for total offenses. Dangerous offenders are part of this group, but other persistent and habitual criminals are also included that merit closer further study, as they may offer useful information in predicting risk of reoffense.
Concern that inclusion of repeat offenders in the sample when first assessed would bias the results found some support, but when age, time at large, study follow-up time, and span of criminal career are taken into account in covariance analysis, differences in sex reoffense rates between repeat and first-time offenders were no longer significant. One may expect that three out of five offenders will reoffend at some time in their criminal career, regardless of being first-time or repeat offenders when initially assessed.
The present report represents one study in one setting and it should be considered in the context of other long-term (20 years or more) follow-up studies. It was limited by missing data, and it is possible that lifetime reoffense rates are higher than reported here. Nevertheless, the use of long-term follow up remains a significant recommendation in researching sex offenders for estimating the underlying reoffense rates. Sexual and general reoffending remain significant over a substantial part of sex offenders’ lives and it is recommended that both be examined.
Footnotes
Acknowledgements
The authors would like to thank barrister Douglas Lehrer for input to this study and the staff of Medical Records of the Clarke Institute of Psychiatry who were so helpful during the course of this study.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.