Assessing Sex Offender Recidivism Using Multiple Measures

Abstract

While the recidivistic activity of sex offenders has received considerable attention from researchers, most studies have been limited by using a single measurement of recidivism. Using arrest/conviction episodes as the unit of analysis, the present study tracked 389 convicted sex offenders for up to 10 arrest/conviction episodes using 11 different measurements of recidivism for an average of 15.7 years. Logistic regression was used to create a model that successfully predicted recidivism with approximately 70% accuracy. The rate of recidivism as defined by new convictions for sex offenses was approximately 10% overall.

Keywords

sex offender recidivism longitudinal logistic regression classification

Spurred on by high-profile cases such as that of Polly Klaas and Megan Kanka who were brutally murdered by convicted sex offenders (Jenkins, 1998), the incarceration of sex offenders has sky rocketed over the last two decades. Utah leads the nation in having the highest percentage of sex offenders incarcerated (33%) as part of its overall prison population. Paradoxically, Utah, with an incarceration rate of 234 per 100,000 residents, has one of the lowest overall rates of incarceration in the country. The assumption that fuels Utah’s high sex offender incarceration rate is that sex offenders are basically incorrigible with high rates of recidivism that persist over extended periods of time. The purpose of this study is to examine the accuracy of this assumption.

This study tracked 389 offenders convicted of felony sex crimes for an average of 15.7 years. Offenders, incarcerated at the Utah State Prison for their initial sex offense, were supervised by the Utah Department of Corrections following their release. Using 11 definitions of recidivism, offenders were tracked for 10 arrest-conviction episodes following their initial conviction for a sex offense. The objectives of the study were to (a) determine the frequency of recidivism for each of the 11 definitions of recidivism for Episodes 1 to 10, (b) construct a predictive model of sex offender recidivism, and (c) determine the extent of recidivistic activity based on arrest-conviction episodes over an extended period of time.

Prior Research on Sex Offender Recidivism

Longitudinal studies of sex offender recidivism have produced a wide array of results. Some research validates beliefs that sex offenses are high-frequency crimes, offenders usually repeat the offense, and sex crimes are undercounted because many acts are never officially reported (Doren, 1998; Firestone et al., 1999). Other research concludes that the probability for sexual reoffending following the initial offense is relatively low (Barbaree, 1997; Bench & Allen, 2011; Wollert, 2001). One explanation for these disparate views is the methodological differences between studies for such factors as sample size, operationalization of recidivism, regionality, length of time offenders were tracked, and the population studied (Losel & Schmucker, 2005).

A U.K. study by Thornton and Travers (1991) examined the recidivistic activity of sex offenders that were incarcerated for at least 4 years and were released from prison in 1980. A review of records 10 years later revealed that 15% of rapists and approximately one third of child molesters had been convicted of new sexual offenses during the follow-up period. Another U.K. study, based on 900 randomly selected sex offenders released from prison in 1987, tracked recidivism over a 4-year period of time. Reconviction data indicate that only 7% of offenders were convicted of new sexual offenses (P. Marshall, 1994).

Research by Hanson, Steffy Richard, and Gauthier (1993) examined the recidivistic activity of 197 child molesters released from a provincial correctional institution in southern Ontario between 1958 and 1974. Offenders were followed for varying lengths of time from 15 to 20 years. Recidivism was defined as a sexual offense or a violent offense that led to conviction. Results of the study suggest that child molesters are at risk of reoffending for many years, with a reconviction rate of 5.2% for the first 6 years and 1.8% for the following years.

Hanson and Bussiere (1998) conducted a meta-analysis of 61 studies of sex offender recidivism consisting of a total of 28,972 offenders. Using a follow-up period of 4 to 5 years, they found a recidivism rate of 13.4% for sexual offenses, 12.2% for nonsexual violent offenses, and 36.3% for any type of offense. Characteristics that predicted sexual recidivism were oriented toward measures of sexual deviance and, to a lesser degree, conventional criminological variables.

More recently, Hanson and Morton-Bourgon (2005) analyzed 115 studies of sex offender recidivistic patterns as well as new offenses committed by sex offenders for nonsexual crimes. Using a variety of recidivistic measures, the results show a sexual recidivism rate of 13.7% and a violent nonsexual recidivism rate of 14.3%. The study concludes that the strongest predictors of sexual recidivism are variables associated with sexual deviancy and antisocial orientation.

Langan, Schmitt, and Durose (2003) conducted a study of the recidivistic patterns of 262,420 offenders released from 15 state prisons in 1994. The sample included 9,691 offenders that had been incarcerated for sex crimes. Offenders, tracked for 3 years following their release, were rearrested at a rate of 5.3% (517 of 9,691). The conviction rate for sex offenders convicted of new sex offenses was 3.5%.

Sample and Bray (2003) examined the recidivistic patterns of 146,918 sex offenders and nonsex offenders arrested in Illinois between 1990 and 1997. Recidivistic activity was measured at intervals of 1, 3, and 5 years. At the 5-year interval, 6.5% of offenders previously convicted of a sex crime had been arrested for new sex offenses. This is considerably lower than 38.8% of the offenders who were initially convicted of robbery charges and arrested for new robbery offenses. Offenders convicted of homicide (5.7%), kidnapping (2.8%), and stalking (5%) are the only offense-specific groups that had lower rates of recidivism than sex offenders.

Hanson and Morton-Bourgon (2007) conducted a meta-analysis of sex offender recidivism based on 577 findings from 79 distinct samples. They found that the recidivism rate for new sexual offenses was 12.4% (72 studies), the recidivism rate for violence (sexual and nonsexual) was 17.5% (36 studies), and the general recidivism rate (all offenses) was 30.1%.

Other studies show a much higher rate of recidivism. For example, Prentky, Lee, Knight, and Cerce (1997) followed 115 extra-familial child molesters over a 25-year period of time. Using “new sex offense charge” and “conviction for a new sex offense” as outcome measures, they documented a rate of 52% for reoffending.

Generalizing from the results of studies on sex offender recidivism is difficult because of the variety of outcome measures used to define recidivism. An additional consideration is that many studies rely on a single outcome measure. For example, several studies operationalize recidivism as rearrest for a sexual offense (Hout, 1997; Marques, Day, Nelson, & West, 1994; W. Marshall & Barbaree, 1988; Nagayama-Hall, 1995; Sample & Bray, 2003; Song & Lieb, 1995). Other studies restrict the definition to new convictions (Friendship, Mann, & Beech, 2003; Gordon & Nicholaichuk, 1996; W. Marshall, Eccles, & Barbaree, 1991; Nicholaichuk, Gordon, Andre, & Gu, 1995; Rice, Quinsey, & Harris, 1991). A limited number of studies take a more “bottom-line” approach by defining sex offender recidivism as reincarceration (Nicholaichuk et al., 1995). Realistically, the measurement of recidivism is more complex than one specific measure (Polizzi, MacKenzie, & Hickman, 1999). Consequently some studies, such as one conducted by Prentky et al. (1997), use multiple measures, including new charges, convictions, and imprisonment as measures of recidivism.

Studies that rely on the use of a single recidivistic measurement run the risk of either underestimating or overestimating the actual rate (Bench, Kramer, & Erickson, 1997). For example, the use of arrest data as an outcome measure for recidivistic activity tends to overestimate recidivism. Numerous studies show that a high percentage of offenders arrested by the police are never tried in court. A study by Reaves and Smith (1995) for the Bureau of Justice Statistics shows that approximately 46% of felony arrests are declined for prosecution or dismissed by the court.

The choice of data used for sex offender research also varies from study to study. Some analyses, such as McConaghy, Blaszczynski, and Kidson (1988), Maletzky (1991), W. Marshall et al. (1991), and Meyer, Cole, and Emory (1992), rely on official records of recidivism and such unofficial records as self-reports and informal reports (i.e., significant-other interviews, Children’s Aid Society records, reports from patients’ legal representatives, reports from social agencies). Research by W. Marshall and Barbaree (1988) shows that unofficial records report over 2.7 times the number of offenses as official records.

Part of what is known about sex offender recidivism comes from studies that have compared the recidivism rates of treated vs. untreated offenders (Friendship et al., 2003; Hanson, Broom, & Stephenson, 2004; Looman, Abracen, & Nicholaichuk Terry, 2000; McGrath, Cumming, Livingston, & Hoke, 2003; Worling & Curwen, 2000). Losel and Schmucker (2005) conducted a meta-analysis of 69 studies containing 80 independent comparisons between treated and untreated offenders. Their findings report an overall recidivism rate of 11.1% for offenders who were treated and 17.5% for offenders who were untreated. A meta-analytic review conducted by Hanson (2002) of 43 studies (n = 9,454) reveals a sexual recidivism rate of 12% for sex offenders who received treatment and a rate of 16.8% for comparison groups. Nagayama-Hall (1995) reviewed 12 studies and found an average sexual offender recidivism rate of 19% for offenders who received treatment and 27% for untreated offenders.

The present study attempts to address some of the deficiencies in previous studies by conducting a longitudinal analysis of sex offender recidivism using arrest-conviction episodes as the unit of analysis. In addition, the study operationalizes recidivism using 11 specific measures to clarify the full breadth of recidivistic activity over an extended period of time.

Method

A longitudinal analysis of sex offender recidivism using arrest-conviction episodes rather than arrest-conviction incidents was conducted for this study. Recidivistic activity was tracked using 11 measures containing 51 inmate and crime-related variables. Stepwise logistic regression was used to determine significances of the variables used in the final model. The resulting model was tested for predictive accuracy.

Participants

The sample consisted of 389 male sex offenders who were under the supervision of the Utah Department of Corrections at some time between 1979 and 2005. The sample was restricted to offenders for whom complete recidivistic information was available. Offenders in the study received sex-offender-specific treatment in a community correctional center. The treatment program was based on cognitive-behavioral and relapse prevention models. Treatment modalities included cognitive restructuring and relapse prevention group therapy, deviant sexual preference-modification, conditioning group therapy with penile plethysmography, and psychoeducational classes. Instruction included training in assertiveness, anger management, sex education, stress management, parenting, victim empathy, thinking errors, and relapse prevention. All sex offenders who are released from the Utah State Prison prior to serving full sentences are required to participate in treatment. Only a very small percent choose to serve their full sentences.

The offenders were predominantly Caucasian (92.4%) with a mean offender age of 33 and a median age of 32. The range was 17 to 73 years of age. The offender IQ distribution was roughly normal, with a mean of 102.8 and a standard deviation of 20.48. The average number of years of education was 11.76. The distribution of marital status was as follows: 33% married, 31% divorced, 9% separated, and 27% never married. The mean number of arrests prior to the initial conviction was 3.6 with a range of 1 to 63. A controlled substance was used at the time of the offense in 13.1% of the cases, and alcohol was used in 20.4%.

Offenders were tracked for 10 arrest-conviction episodes following their initial conviction for a sex offense. The 11 definitions of recidivistic behavior used in this research are listed in Table 1. An episode is defined as an event in which the offender is the subject of a technical violation, an arrest, arrest warrant, single or multiple charges, and single or multiple convictions. Offenders were tracked for a median of 15.2 years, with minimum and maximum values of 11 and 24.2 years, respectively. The mean length of time tracked was 15.7 years with a standard deviation of 2.79 years. Although offenders were tracked for up to 10 episodes, the actual time at risk was calculated from the time the offender was released from prison until his next involvement with the criminal justice system. The time at risk ranged from 3.4 years to 23 years with a mean of 13.7 years. Offenses tracked include felony and misdemeanor violations as well as sexual and nonsexual crimes. Fifty-one independent variables were assessed in relation to the various measures of recidivism. Table 2 contains the list of variables.

Table 1.

Definitions of Recidivistic Activity Used in This Analysis.

1. Technical violations

2. Warrant issued nonsex offence

3. Warrant issued sex offence

4. Arrested misdemeanor nonsex offence

5. Arrested misdemeanor sex offence

6. Convicted misdemeanor nonsex offence

7. Convicted misdemeanor sex offence

8. Arrested felony nonsex offence

9. Arrested felony sex offence

10. Convicted felony nonsex offence

11. Convicted felony sex offence

Table 2.

Independent Variables Included in the Original Logistic Regression Model.

Adult touching of child’s clothed body	Recidivism type
Age at first arrest	Severity of present offence
Age of offender	Sexually abused as an adult
Age of victim	Sexually abused as child
Anal intercourse with victim	Simulated intercourse with victim
Child instructed to expose genitals	Substance abuse type
Child masturbated adult	Total arrests
Completed treatment	Total arrests for sex offenses
Digital penetration of victim	Total charges per episode
Drug use at time of present offence	Total convictions for nonsex offenses
Education Level	Total convictions for sex offenses
Escapes from a secure facility	Total felony arrests prior to key offence
Exposure of adult genitalia to child	Total felony nonsex arrests prior to key offence
Fondling of child’s unclothed genitals	Total female victims for present offence
IQ of offender	Total juvenile arrests
Length of custody	Total juvenile convictions
Marital status of offender	Total male victims
Most serious nonsex offence	Total nonsex misdemeanor arrests prior to key offence
Most serious offence per episode	Total number of arrests for nonsex offenses
Most serious sex offence	Total parole violations prior to primary sex offence
Offender exhibitionism	Total probation violations prior to key sex offence
Offender failed treatment	Total sex misdemeanor arrests prior to key sex offence
Offender intoxicated at time of offence	Total number victims for the offence
Offender views pornography	Type of force used on victim
Open mouth kissing	Vaginal intercourse
Reason for program failure

Recidivistic activity was confirmed using information obtained from the Utah Department of Corrections, the Utah Bureau of Criminal Identification, and the National Crime Information Center (NCIC). Use of multiple data sources allows recidivistic activity to be detected at state and national levels. In cases where multiple recidivistic charges occurred within a single episode, the most severe charge or disposition, as defined by the Utah State Code, was used. For example, following his release from prison, one of the offenders in the study was arrested for the felony charge of “Aggravated Sexual Abuse of a Child” but was ultimately convicted of “Lewdness With a Child,” which is a misdemeanor under the Utah State Code. The conviction for lewdness with a child was coded as the most severe disposition for the episode. In another instance, an offender was initially arrested for larceny. However, a subsequent investigation revealed that he was the suspect in a rape. The offender was ultimately convicted for this felony charge which in turn was coded as the most severe disposition for the episode.

Results

Table 3 displays the frequency of types of recidivism by offenders in the sample for Episodes 1 through 10. The results show the latest dispositional information available within the time limits of the study.

Table 3.

Type of Recidivistic Activity for Episodes 1 to 10.

Disposition	Frequency	Percentage
Revocations—technical violations	301	54.74
Warrant issued—nonsex	11	2.00
Warrant issued—sex	20	3.63
Arrested misdemeanor—nonsex	0	0.00
Arrested misdemeanor—sex	0	0.00
Convicted misdemeanor—nonsex	111	20.18
Convicted misdemeanor—sex	19	3.45
Arrested felony—nonsex	0	0.00
Arrested felony—sex	0	0.00
Convicted felony—nonsex	48	8.73
Convicted felony—sex	40	7.27

Offenders who violated conditions of their release, the largest category of recidivism, account for approximately 55% of all recidivistic activity. Conditions required of sex offenders are usually comprehensive and include restrictions such as refraining from the use of drugs and alcohol, prohibitions against frequenting parks, playgrounds, and schools, and requirements for attending therapy, paying restitution to victims, and securing employment.

Convictions for nonsex misdemeanor offenses comprised the category with the next highest frequency (20.86%). Offenses committed in this category include possession of a controlled substance, driving under the influence, and minor theft.

Logistic Regression Analysis

Stepwise logistic regression was used to create a predictive model of sex offender recidivism where the outcome or dependent variable is defined as 1 for recidivism and 0 for nonrecidivism. Recidivism was operationalized to include any conviction or guilty plea to a misdemeanor (sex or nonsex) or felony (sex or nonsex) offense. Episodes, rather than individuals, were used as the unit of analysis. The number of episodes for offenders ranged from 0 to 10 depending on recidivistic activity. This approach, suggested by Allison (1984), effectively increases the sample size from 389 to 581 and avoids the necessity of conducting a separate analysis for each successive event. Recidivistic activity decreases substantially at about the third episode, after which there are too few repeat offenders for meaningful analysis. Although some biasing occurs in using episodes as the unit of analysis, there are two major advantages to this approach. First, it allows for the creation of a single model that takes into consideration the entire time period tracked in the study, and second, the weighting of individuals is in the direction of offenders who recidivate most often and are of most interest to the criminal justice system.

For this study, recidivism is defined as a conviction for a felony or misdemeanor offense (sex or nonsex). The definition was chosen because of the stringent requirements associated with “beyond a reasonable doubt” that must be demonstrated by the prosecution to obtain a conviction.

Although 51 variables were initially included in the analysis, the stepwise procedure eliminated all variables that were associated with the crime itself and retained only four variables, all of which were associated with the offender. Table 4 presents the results of the logistic regression analysis. Only four variables were found to be significant: age at first arrest, technical violations (yes = 1, no = 0), whether the inmate failed treatment (yes = 1, no = 0), and whether the offender was intoxicated at the time of the offense (yes = 1, no = 0).

Table 4.

Results of the Logistic Regression Predicting Recidivism.

Variable	Coefficient	Standardized coefficient
Age at first arrest	−0.039*	−0.231
Technical violations	1.032*	0.232
Failed treatment	0.829*	0.217
Intoxicated	0.508**	0.123
Constant	1.027
−2 Log Likelihood	684.178

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

When using a variable-reduction technique on many variables, a potential problem is multicollinearity. For this study, the highest correlation between all pairs of the original 51 independent variables was a single pair at 0.85. This correlation is not high enough to create dependencies between the pair, and neither variable was significant enough to be retained in the final model. The correlations for the independent variables in the final model ranged from −0.27 to 0.13. Table 4 includes the significant variables, their coefficients, and the standardized coefficients to show their relative importance in the model. The likelihood of recidivism for each of the significant variables is shown in Table 5 which reports the odds ratios along with 95% confidence intervals. The odds ratio for “age at first arrest” is less than one which means the regression coefficient is negative. This result indicates that the older the offender was at first arrest, the less likely he was to reoffend for any crime (sex or nonsex). For the other three variables the regression coefficients are positive: Sex offenders with a history of violating the conditions of their release were 2.8 times more likely to recidivate than those with no history of technical violations, offenders who failed treatment programs were 2.3 times more likely to recidivate than offenders who did not fail treatment, and offenders who were intoxicated at the time of the offense were approximately 1.7 times more likely to recidivate than offenders who were not intoxicated.

Table 5.

Odds Ratios With 95% Confidence Interval for Significant Variables in the Logistic Regression Model.

		95% Confidence Interval
Variable	Odds Ratio (exp(B))	Lower Limit	Upper Limit
Age at first arrest	0.962	0.945	0.979
Technical violations	2.805	1.684	4.673
Failed treatment	2.290	1.524	3.441
Intoxicated	1.662	1.065	2.593

The overall percentage of recidivism predicted correctly by the model is 70.2. As noted in Table 6, recidivists were predicted correctly in 71.5% of cases, and nonrecidivists were predicted correctly in 65.3% of cases. The welcome finding here is that the model is more successful at predicting recidivism than nonrecidivism, and predicting recidivism is more important for public safety.

Table 6.

Classification Results.

	Actual
Predicted	Recidivist	Non-recidivist	Total
Recidivist	331 (71.5%)	132 (28.5%)	463 (100%)
Non-Recidivist	41 (34.8%)	77 (65.3%)	118 (100%)
Total			581

Note. The model correctly classifies 71.5% of recidivist and 65.3% of nonrecidivist. Total correct classifications are 70.2%.

The extent of recidivistic activity was determined by tracking offenders for up to 10 reconviction episodes. The results, fashioned in the form of a recidivistic funnel, are detailed in Figure 1. The number of offenders was standardized to 1,000. This technique is used by the Bureau of Justice Statistics and many criminology textbooks for ease of interpretation. Reported results refer to misdemeanor and felony convictions.

Figure 1.

Sex offender recidivism funnel for the first three episodes standardized to 1,000 offenders (convictions only).

Episode 1 contains all offenders who were convicted of new crimes following their initial release (17% of the original group). Eighty-three percent had no further convictions beyond their initial felony sex offense. Approximately 7.7% of the recidivists committed new sex offenses, and 9.3% committed nonsex crimes.

The study population for Episode 2 includes only the 17.0% of original sex offenders who recidivated in Episode 1. Twenty-seven (27.1%) percent of these offenders (4.6% overall) were convicted of new crimes in Episode 2. Ten (10.4%) percent of offenders who were convicted of sex crimes in Episode 1 were also convicted of sex crimes in Episode 2 (0.8% overall). Of the offenders who were convicted of nonsex crimes in Episode 1, 10.8% were convicted of sex crimes in Episode 2 (1.0% overall), while 30.1% were convicted of nonsex crimes (2.8% overall).

By the third Episode, only 1.5% of the original sex offender population was convicted of another crime, one third of which were sex crimes (0.5% overall). There were virtually no offenses committed beyond Episode 3. Offenders may have discontinued their criminal behavior or may have been incarcerated for such extended periods of time that they were unable to commit further crimes during the study period.

Conclusion

The pattern of sex offender recidivism that emerges from this study echoes the results of a number of prior investigations which suggest that most sex offenders are not convicted of new sex offenses after their release from prison. Initially, 51 independent variables were included in the prediction analysis. Stepwise regression reduced the number to four significant variables: age at first arrest, technical violations, treatment failure, and intoxication at the time of the offense. The resulting model was 70% successful in predicting previous offenders who would reoffend.

The first predictive variable, age at first arrest, shows a negative correlation between the age of the sex offender and recidivism. This relationship between age and crime in general has been well documented by prior research (Cohen & Land, 1987; Kercher, 1987;) Hanson and Bussiere (1998). In a meta-analysis (Hanson, 2002), Hanson finds that recidivism declines with age and that this desistence may be attributed to a decrease in sexual interest and/or a decline in the opportunity for sexual encounters.

Researchers continue to explore the exact relationship between age and sex crimes. Langan et al. (2003) tracked the recidivistic patterns of approximately 9,700 offenders for a 3-year period following their release from U.S. prisons. Results indicate that the downward trend in recidivism begins after the age of 44. Several studies have focused on the importance of age at release versus age at the time of first sexual offense (Barbaree, Langton, & Peadcock, 2005); others focus on the age/recidivism connection as a function of offender typology (Prentky & Lee, 2007). Further research will hopefully enhance our understanding of the age/sex offender relationship.

A history of technical violations is the second significant variable noted in this research to be predictive of recidivism by sex offenders. This result is somewhat surprising in view of the large number of variables that were examined. Underlying assumptions for this analysis were that released sex offenders would develop extensive criminal histories prior to reconviction, and that sex offense would be one of many types of crime manifest in these histories. Several studies suggest the need to examine this assumption. Fisher and Thornton (1993) point out that many sex offenses are “one-time incidents.” Viewed from this perspective, it is not untenable that technical violations are found to be significant while many other variables are not strong predictors of sex offender recidivism.

Our third finding, inmates who “failed treatment” are at higher risk for reoffending than those who completed treatment, reinforces the results of studies by Hout (1997) and Hanson and Bussiere (1998); This finding also may be related to the importance of technical violations as predictors of recidivism. The failure to comply with treatment rules or with conditions of release may be the manifestation of a general reluctance to follow society’s expectations.

The fourth predictor of recidivism, the relationship between intoxication and offending, is also well-established in criminological literature. A study by Maguire and Pastore (1999) found that 33% of prison inmates and 22% of federal prison inmates were under the influence of illegal drugs during the commission of their crimes. Alcohol impacts criminal activity by inhibiting higher-order cognitive functions and increasing awareness of sexual stimuli (Abbey, Zawacki, Buck, Clinton-Sherrod, & McAuslan, 2001; Peterson, Rothfleisch, Zelazo, & Phil, 1990). In addition, alcohol is thought to distort the ability to consider victim suffering and to alter a perpetrator’s sense of morality (Abbey et al., 2001). Other debilitating consequences of alcohol consumption include reduced inhibitions, decreased impulse control, irrationality and excitability, all of which are linked with the commission of sex offenses and other violent crimes.

Discussion

Several recent high-profile cases have focused attention on sex offender recidivism. Professionals and concerned citizens understandably want to limit the risk from incorrigible sex offenders. Consequently, all states require sex offenders to register with law enforcement, some states use civil commitment following incarceration, and a number of politicians have called for mandatory life sentences regardless of the sex offender’s criminal history.

The current interest in sex offenses has spawned assumptions about the behavior of sex offenders that are contrary to the findings of numerous empirical studies showing that sex offender recidivism is surprisingly low. The tendency to overlook these findings may be due in part to methodological differences that have led to conflicting results. Furby, Weinrott, and Blackshaw (1989) note that “the methodological weaknesses and lack of uniformity are an almost inevitable result of the conditions under which most recidivism studies have been conducted.” They add that many sex offender studies have been undertaken with a mandate to answer policy makers’ questions in an unrealistic amount of time and have been hampered by data limitations, inadequate funding, and a lack of research expertise. Fisher and Thornton (1993) argue that there is no consensus about how to operationalize sex offender recidivism, and that study findings encompass various definitions of recidivism that may include rearrest, return to prison, reconviction for a felony offense, and reconviction for another felony sex offense. Short follow-up periods have also been cited as a weakness. Quinsey (1984) observed after reviewing recidivism studies on rapists that “the differences in recidivism across these studies is truly remarkable: clearly by selectively contemplating the various studies, one can conclude anything one wants” (p.101).

The present study overcomes some of these limitations by including many variables that have not been examined by previous research. In addition, recidivism was tracked over a long period of time and episodes were used as the unit of analysis rather than criminal justice incidents. A noteworthy and unexpected finding of our study is the high number of variables that are not predictive of sex offender recidivism. The initial analysis included variables that have had a long-standing association with criminal recidivism in general, such as prior involvement with pornography, sexual abuse as a child, and number of prior arrests. None of these factors were significant predictors in our study.

Several considerations should be noted regarding the results of the present study: (a) Utah has a long history of austere penalties for sex offenders. Many offenders who receive prison sentences in Utah might receive less-severe sanctions in other states. The extent to which the population of this study is similar to the larger community of sex offenders is unclear. (b) As this was a panel study, it may not be representative of offenders incarcerated at a later point in time. The panel was constructed using all felony sex offenders under the supervision of the Utah Department of Corrections during a specific period of time. Practices affecting incarceration are contingent on ongoing changes in laws, sentencing practices, policies, and bed space. Replications using our unique methodology to measure multiple forms of recidivism over an extended period of time are needed to compare the findings of the present study with other sex offender populations. (c) The results of our study are consistent with a number of other longitudinal studies of sex offender recidivism. Studies by West (1987), Hanson et al. (1993), Hanson and Bussiere (1998), Langan et al. (2003) also found comparatively low rates for sex offender recidivism.

Although more than 35 years of research indicates that the incidence of sex offender recidivism is not only low, but much lower than many other criminal offenses, the hope that society will achieve certainty in predicting recidivism is unrealistic. Nonetheless, it is important for public safety to continue efforts to identify the small group of sex offenders who will become chronic recidivists. With this objective in mind, it is hoped that this study offers encouraging directions for future research.

Footnotes

Authors’ Note

This paper is dedicated to the memory of our friend and colleague, Steven P. Kramer, PhD, who assisted in this study before his untimely death. The views expressed in this paper are those of the authors and do not necessarily reflect the official policy or position of any agency or employee of the state of Utah.

Declaration of Conflicting Interests

The author(s) declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Lawrence L. Bench is a faculty member at the University of Utah as well as a consultant with the National Institute of Justice. A guest lecturer at the National Academy of Sciences and Zayed University, Abu Dhabi, and a recent retiree from the Utah Department of Corrections, his research interests include sex offender recidivism, experimental design, validity assessment, and criminal behavior alteration through cognitive intervention.

Terry D. Allen is an adjunct professor of sociology at Weber State University, teaching courses in statistics, research methods, criminology, and deviant behavior. A former research statistician with the Utah Attorney General’s Office, he began his career as a Utah State Prison correctional officer.

References

Abbey

Zawacki

Buck

P. O.

Clinton-Sherrod

A. M.

McAuslan

(2001). Alcohol and sexual assault. Alcohol, Research, & Health, 25, 43-51.

Allison

P. D.

(Ed.). (1984). Event history analysis: Regression for longitudinal event data. Newberry Park, CA: SAGE.

Barbaree

(1997). Treatment outcome research. Sexual Abuse: A Journal of Research and Treatment, 9, 89-160.

Barbaree

H. E.

Langton

Peadcock

(2005). The effect of age-at-release on recidivism of high-risk-sex offenders. Unpublished manuscript.

Bench

L. L.

Allen

T.D.

(2011). A Logistic Regression Analysis of Factors That Predict Sex Offender Recidivism. In Schwartz

B. J.

(ed.), Handbook of Sex Offender Treatment (pp. 26-1 – 26-16). Kingston, NJ: Civic Research Institute.

Bench

L. L.

Kramer

S. P.

Erickson

(1997). A discriminant analysis of predictive factors in sex offender recidivism. In Schwartz

B. K.

Cellini

H. R.

(Eds.), The sex offender: New insights, treatment innovations, and legal developments (Vol. ll, pp. 1-15). Kingston, NJ: Civic Research Institute.

Cohen

Land

(1987). Age structure and crime: Symmetry versus asymmetry and the projection of crime rates through the 1990s. American Sociological Review, 52, 170-183.

Doren

D. M.

(1998). Recidivism base rates, predictions of sex offender recidivism, and the “sexual predator” commitment laws. Behavioral Sciences & the Law, 16, 97-114.

Firestone

Bradford

J. M.

Marcia

Greenberg

D. M.

Larose

M. R.

Curry

(1999). Prediction of recidivism in incest offenders. Journal of Interpersonal Violence, 5, 511-531.

10.

Fisher

Thornton

(1993). Assessing the risk of re-offending in sexual offenders. Journal of Mental Health, 2, 105-116.

11.

Friendship

Mann

R. E.

Beech

A. R.

(2003). Evaluation of a national prison-based treatment program for sexual offenders in England and Wales. Journal of Interpersonal Violence, 18, 744-759.

12.

Furby

Weinrott

M. B.

Blackshaw

(1989). Predicting relapse: A meta-analysis of sexual offender recidivism studies. Psychological Bulletin, 105, 3-30.

13.

Gordon

Nicholaichuk

(1996). Applying the risk principle to the treatment of sexual offenders. Forum on Corrections Research, 8(2), 36-38.

14.

Hanson

R. K.

(2002). Recidivism and age: Follow-up data from 4,673 sexual offenders. Journal of Interpersonal Violence, 17, 1046-1062.

15.

Hanson

R. K.

Broom

Stephenson

(2004). Evaluating community sex offender treatment programs: A 12-year follow-up of 724 offenders. Canadian Journal of Behavioural Science, 36, 87-96.

16.

Hanson

R. K.

Bussiere

M. T.

(1998). Predicting relapse: A meta-analysis of sexual offender recidivism studies. Journal of Consulting and Clinical Psychology, 66, 348-362.

17.

Hanson

R. K.

Morton-Bourgon

K. E.

(2007). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis. Public Safety and Emergency Preparedness Canada, Department of Justice Canada.

18.

Hanson

R. K.

Morton-Bourgon

K. E.

(2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73, 1154-1163.

19.

Hanson

R. K.

Steffy Richard

Gauthier

(1993). Long-term recidivism of child molesters. Journal of Consulting and Clinical Psychology, 61, 646-652.

20.

Hout

(1997). Community-based sex offender program evaluation project (1997 Report to the Legislature). St. Paul: Minnesota Department of Corrections.

21.

Jenkins

(1998). Moral panic: Changing concepts of the child molester in modern America. London, England: Yale University Press.

22.

Kercher

(1987). Causes and correlates of crime committed by the elderly. In Borgatta

Montgomery

(Eds.), Aging policy. (Vol. 9, 256-280). Beverly Hills, CA: SAGE.

23.

Langan

P. A.

Schmitt

Durose

(2003). Recidivism of sex offenders released from prison in 1994. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

24.

Looman

Abracen

Nicholaichuk Terry

(2000). Recidivism among treated sexual offenders and matched controls: Data from the Regional Treatment Centre (Ontario). Journal of Interpersonal Violence, 15, 279-290.

25.

Losel

Schmucker

(2005). The effectiveness of treatment for sexual offenders: A comprehensive meta-analysis. Journal of Experimental Criminology, 1, 117-146.

26.

Maguire

Pastore

(Eds.). (1999). Sourcebook of criminal justice statistics 1998. Washington, DC: Bureau of Justice Statistics.

27.

Maletzky

B. M.

(1991). Treating the sex offender. London, England: SAGE.

28.

Marques

J. K.

Day

D. M.

Nelson

West

M. A.

(1994). Effects of cognitive-behavioral treatment on sex offender recidivism: Preliminary results of a longitudinal study. Criminal Justice and Behavior, 21, 28-54.

29.

Marshall

(1994). Reconviction of imprisoned sexual offenders (Research Bulletin No. 36). London, England: Home Office Research and Statistics Department.

30.

Marshall

Barbaree

(1988). The long-term evaluation of a behavioral treatment program for child molesters. Behavior Research and Therapy, 26, 499-511.

31.

Marshall

Eccles

Barbaree

(1991). The treatment of exhibitionists: A focus on sexual deviance versus cognitive and relationship features. Behavior Research and Therapy, 29, 129-135.

32.

McConaghy

Blaszczynski

Kidson

(1988). Treatment of sex offenders with imaginable desensitization and/or medroxyprogesterone. Acta Psychiatrica Scandinavica, 77, 199-206.

33.

McGrath

R. J.

Cumming

Livingston

J. A.

Hoke

S. E.

(2003). Outcome of a treatment program for adult sex offenders - From prison to community. Journal of Interpersonal Violence, 18, 3-17.

34.

Meyer

W. J.

Cole

Emory

(1992). Depo provera treatment for sex offending behavior: An evaluation of outcome. Bulletin of the American Academy of Psychiatry and the Law, 20, 249-259.

35.

Nagayama-Hall

(1995). Sexual offender recidivism revisited: A meta-analysis of recent treatment studies. Journal of Consulting and Clinical Psychology, 63, 802-809.

36.

Nicholaichuk

Gordon

Andre

(1997). Long-term outcome of the Clearwater Sex Offender Treatment Program. Paper presented at the 14th Annual Conference of the Association for the Treatment of Sexual Abusers. New Orleans, LA. October 1997.

37.

Peterson

J. B.

Rothfleisch

Zelazo

P. D.

Phil

R. O.

(1990). Acute alcohol intoxication and cognitive functioning. Journal of Studies on Alcohol, 51, 114-122.

38.

Polizzi

D. M.

MacKenzie

D. L.

Hickman

L. J.

(1999). What works in adult sex offender treatment? A review of prison- and non-prison-based treatment programs. International Journal of Offender Therapy and Comparative Criminology, 43, 357-374.

39.

Prentky

R. A.

Lee

A. F. S.

(2007). Effect of age-at-release on long term sexual re-offence rates in civilly committed sexual offenders. Sex Abuse, 19, 43-59.

40.

Prentky

R. A.

Lee

A. F. S.

Knight

R. A.

Cerce

(1997). Recidivism rates among child molesters and rapists: A methodological analysis. Law and Human Behavior, 21, 635-659.

41.

Quinsey

V. L.

(1984). Sexual aggression: Studies of offenders against women. In Weisstub

(Ed.), Law and mental health: International perspectives (Vol. 1). New York, NY: Pergamon.

42.

Reaves

Smith

(1995). Felony defendants in large urban counties. Washington, DC: Bureau of Justice Statistics.

43.

Rice

M. E.

Quinsey

V. L.

Harris

G. T.

(1991). Sexual Recidivism among child molesters released from a maximum security psychiatric institution. Journal of Consulting and Clinical Psychology, 59, 381-386.

44.

Sample

L. A.

Bray

T. M.

(2003). Are sex offenders dangerous? Criminology & Public Policy, 3, 59-82.

45.

Song

Lieb

(1995). The twin rivers sex offender treatment program: Overview of recidivism studies. Olympia: Washington State Institute of Public Policy.

46.

Thornton

Travers

(1991). A longitudinal study of the criminal behavior of convicted sexual offenders. Paper presented at the Prison Psychologist’s Conference, Her Majesty’s Prison Service.

47.

West

(1987). Sexual crimes and confrontations (Cambridge studies in criminology). Aldershot, UK: Gower.

48.

Wollert

(2001). An analysis of the argument that clinicians under-predict sexual violence in civil commitment cases. Behavioral Sciences & the Law, 19, 171-184.

49.

Worling

J. R.

Curwen

(2000). Adolescent sexual offender recidivism: Success of specialized treatment and implications for risk prediction. Child Abuse & Neglect, 24, 965-982.