It is also important to recognize that, once reported to law enforcement, only a subset of sex offenses result in the arrest of the perpetrator. Grotpeter and Elliot (2002) found that only 2.5 percent of sexual assaults and 10 percent of serious sexual assaults resulted in an arrest and Snyder (2000) found that an arrest was made in only 29 percent of reported juvenile sexual assaults. In addition, a number of studies have found that sex offenders disclose in treatment or in surveys that they had committed a large number of sex crimes before they were first caught or arrested. Abel and his colleagues interviewed paraphiliacs (i.e., those with a diagnosed psychosexual disorder) under conditions of guaranteed confidentiality and found that only 3.3 percent of their self-admitted hands-on sex offenses, such as rape and child molestation, resulted in an arrest (Abel et al., 1988). Simons, Heil and English (2004) found that only 5 percent of rapes and child sexual assaults self-reported during prison treatment were identified in official records. Likewise, another study found that only 1 percent of contact and noncontact sexual offenses self-reported during treatment were identified in official records (Ahlmeyer et al., 2000).
Studies also have demonstrated a \"disproportionate and patterned attrition of sexual offenses and sexual offenders from the criminal justice process\" (Larcombe, 2012, p. 482). While case attrition (the dropping of a legal case by authorities, for various reasons) occurs for all types of offenses, it appears to be particularly pronounced for sexual crime and offenders (Gelb, 2007). Moreover, certain types of sexual crimes and offenders are more likely to be subject to criminal justice system processing and ultimately conviction, and these cases are not representative of sexual offenses or sexual offenders overall (Lievore, 2004; Kelly, Lovett & Regan, 2005). As Larcombe (2012, p. 482) points out, police, prosecutors, jurors and the community tend to take more seriously those assaults that are \"clearly interpretable as violence\" and \"least similar to potentially appropriate sex.\" Further, among all sexual offenders, those who have had \"prior contact with the police\" and those who have assaulted \"children, male victims and female victims who are strangers\" are most likely to be arrested, charged and prosecuted (Larcombe, 2012, p. 493; Statewide Steering Committee to Reduce Sexual Assault [SSCRSA], 2006; Kelly, Lovett & Regan, 2005; Temkin & Krahé, 2008). Research indicates that victim characteristics can also play a role in attrition. For example, females who are young, who have disabilities or who are members of other vulnerable populations have been found to be \"proportionally overrepresented as victims of rape\" yet underrepresented among rape cases processed in the criminal justice system (Larcombe, 2012, p. 489; SSCRSA, 2006). This systematic and patterned attrition of sexual offenses within the criminal justice system ensures that the relatively small number of sex crimes that are reported, prosecuted and ultimately result in conviction do not reflect \"the most common or injurious forms of sexual violence experienced by women and children\" (Larcombe, 2012, p. 483). Hence, findings from recidivism studies need to be interpreted within the context of sexual assault incidence, prevalence and attrition research.
Findings from these studies, like those from the Harris and Hanson (2004) analysis, demonstrate how the recidivism rates of sex offenders increase as follow-up periods become longer. In the study conducted by Harris and Hanson (2004), sexual recidivism rates increased from 14 percent after five years of follow-up to 24 percent after 15 years of follow-up. In the study conducted by Olver, Wong and Nicholaichuk (2008), sexual recidivism rates for treated offenders increased from 11.1 percent after three years of follow-up to 21.8 percent after 10 years of follow-up. In a somewhat older study, Hanson, Scott and Steffy (1995) found that first-time recidivism for a sexual/violent crime occurred between 10 and 31 years into follow-up for 10 percent of a sample of 191 child molesters released from a Canadian prison16
Using a sample of sex offenders similar to the one employed in the Prentky et al. (1997) study, Knight and Thornton (2007) examined the recidivism rates of rapists as part of a larger study aimed at evaluating and improving risk assessment schemes for sex offenders. As in the Prentky et al. study, everyone in the sample had been referred to the Massachusetts Treatment Center for civil commitment evaluation between 1959 and 1984. But unlike in the Prentky et al. study, not everyone in the sample had been civilly committed. Still, generalizing findings from the analysis to rapists overall may be problematic given the high-risk nature of the sex offenders in the study and the length of time that has passed since these individuals committed their referral offense. Two broad findings are worth noting. First, based on a new charge for a serious sexual crime (those involving physical contact with a victim), Knight and Thornton found that rapists in their study recidivated at a rate of 12 percent after three years of follow-up and 20 percent after 15 years of follow-up).32 Second, when the researchers examined the pace of recidivism in a comparative analysis involving rapists and child molesters, \"consistent with the hypothesis that rapists are criminologically more generalists than child molesters,\" rapists were found to have faster rates of recidivism than child molesters for non-sexual victim-involved and victimless crimes\" (Knight and Thornton, 2007, p.77).
The accuracy of Doren's (1998) estimate regarding the long-term propensity of rapists to reoffend and the contention that any nontrivial proportion of sex offenders may show first-time recidivism 20 years or more following release from incarceration or discharge from probation, both remain subject to debate. Harris and Hanson (2004, p. 11), for example, in discussing their findings concerning the long-term sexual recidivism rates of rapists and child molesters, stated, \"The decreasing rate of offending with age suggests that the rates observed after 15 to 20 years are likely to approximate the rates that would be observed if offenders were followed for the rest of their lives.\"
A relatively large body of research exists on the recidivism rates of child molesters. While unreported crime affects all recidivism research, it is particularly problematic in recidivism studies of child-molesting offenders as several studies have demonstrated that the likelihood that a sexual assault will be reported to law enforcement decreases with the victim's age (Kilpatrick, Saunders & Smith, 2003; Smith et al., 2000; Sorenson & Snow, 1991).
The study of sex offenders released from state prisons in 1994 by Langan, Schmitt and Durose (2003) included a large sample (4,295) of child molesters. The researchers reported that 5.1 percent of the 4,295 child molesters released from prison in 1994 were rearrested for a new sex crime within three years of their release, 14.1 percent were rearrested for a violent crime and 39.4 percent were rearrested for a crime of any kind. Similar to the pattern for rapists in the study, child molesters with more than one prior arrest had an overall recidivism rate nearly double (44.3 percent compared to 23.3 percent) that of child molesters with only one prior arrest.
Three other studies mentioned in the prior discussion about the recidivism of rapists also make contributions to the knowledge base about the recidivism patterns of child molesters. As part of their larger study designed to evaluate risk assessment schemes for sexual offenders, Knight and Thornton (2007) examined the recidivism rates of child molesters. Their analysis examined the recidivism of child molesters who had been referred to the Massachusetts Treatment Center for evaluation between 1959 and 1984. Again, given the high-risk nature of these offenders and the length of time that has passed since these individuals committed their referral offense, findings from the analysis may have limited application to child molesters today. Still, several findings from the analysis are worth noting. First, Knight and Thornton (2007, p. 7) found a recidivism rate for child molesters of 12 percent after three years of follow-up (24 of 97 offenders recidivated) and 20 percent after 15 years of follow-up (18 of 91 offenders recidivated) based on a new serious sexual charge.35 Second, when the researchers examined the pace of recidivism in a comparative analysis involving child molesters and rapists, they found no difference in the speed of sexual recidivism between these two types of sex offenders. However, they did find that child molesters recidivated at a slower pace than rapists for both non-sexual victim-involved and victimless crimes.\"
Findings from Harris and Hanson's (2004) analysis are particularly compelling because they document differential rates of recidivism for different types of child molesters based on follow-up periods of five, 10 and 15 years. For all child molesters in the analysis, the researchers found five-, 10- and 15-year sexual recidivism rates based on new charges or convictions of 13 percent, 18 percent and 23 percent, respectively. Table 2 presents the study's recidivism estimates (based on new charges or convictions) for five-, 10- and 15-year follow-up periods for molesters of boys, molesters of girls and incest offenders.
Table 2 shows that molesters of boys had the highest rates of sexual recidivism. Different patterns of reoffending within child molester populations have been found in other studies as well, with molesters of boys having higher recidivism rates than other types of child molesters (see, e.g., Seto, 2008). It is important to keep in mind that the recidivism rates observed for child molesters, and for incest offenders particularly, are impacted by underreporting even more so than recidivism rates for other types of sex offenders, as research has shown that child victims who knew their perpetrator were the least likely to report their victimization (Smith et al., 2000). 59ce067264