<< Cunningham Catch-22 | Main | News Scan >>

Promise and Pitfalls of Sex Offender Research


As I discussed in a previous post, there’s much talk about sex offenders but a lack of good science. One of the most discussed areas in terms of sex offenders is risk of recidivism. While some say recidivism risk is relatively low among sex offenders, others disagree and praise the severe civil restrictions mandated for many sex offenders. Where does the truth lie? Like so many things in life, it’s a mixed bag.

Some food for thought about sex offender recidivism research:

1. Defining the population. Each study defines “sex offenders” differently. When the media speaks of sex offenders what they’re really taking about are child molesters (although, civil commitment statutes theoretically can be applied to all sex offenders). Yet, some sex offender studies include flashers, rapists, frotterists, inter alia, while others only focus on child molesters. Thus, there’s a real problem when people say "research has shown sex offenders..."

2. Defining recidivism. How do we know a sex offender has committed another sex offense? Depending on the study, arrests, convictions, or incarceration is the measure. This partially explains the variation in reported recidivism rates. But none of these are perfect measures of recidivism. Moreover, it is generally accepted that sex offenses are underreported for a variety of reasons. Considering that so much sexual offending occurs within families, this fact seems particularly salient.

3. Length of follow-up after release. Some studies look at 5 years, some 6 months, and a few have examined 25 years. Generally, the studies that examined longer time periods have shown higher recidivism rates. While this finding is intuitive, the numbers are striking. A 1997 study showed a 19% recidivism rate among child molesters during a 5 year period that ballooned to 52% during a 25 year study period. Likewise, rapists also had a 19% recidivism rate at 5 years that increased to 39%. Other studies have shown up to a 40% recidivism rate for any crime of violence over 4 years for rapists.

4. Base rates. Since sexual offending is relatively rare, it suffers from a low base rate problem. This makes statistical analysis difficult, with the potential that generalized statements of recidivism are prone to inflated estimates which overrate recidivism risk.

5. Meta-analysis. This is a popular research methodology these days whereby all studies meeting certain criteria are combined into a single new study to arrive at what is known as an effect size. Meta-analysis adds power to statistical computations but it has one major weakness: It is only as good as the data in the original studies. If the original studies were conducted poorly (and many sex offender studies are), but published nonetheless, they can be used in a subsequent meta-analysis which will report erroneous findings.

6. Static v. dynamic risk factors. In risk assessment, we talk of static factors (i.e., those that cannot change, e.g., gender) and dynamic factors (i.e., those that theoretically are amenable to change, i.e., substance abuse). While there is much discussion about sex offender treatment modifying dynamic factors, the bare truth is that static factors account for most of the risk variance. Simply put, dynamic factors don’t seem to mean much in terms of reducing risk (but are used to justify treatment programs). Thus, there are serious questions as to whether treatment can ameliorate risk, especially in high-risk offenders. Moreover, the dynamic factor associated most with risk, substance abuse, is notoriously difficult to eliminate long-term.

7. Measures of sexual deviance. There are many and many are poor. The Abel screen is frequently used, despite serious concern among many researchers (and the courts in Massachusetts) that it lacks validity. Same goes for the penile plethysmograph.

8. Not all sex offenders are the same. Incestuous offenders have the lowest risk, followed by opposite-sex child molesters, followed by same sex child molesters. Those who offend against strangers, juvenile sex offenders, and offenders with diverse victims (male and females) are more likely to re-offend. Given the heterogeneity of sex offenders, we should be cautious about any blanket statements regarding recidivism risk among sex offenders.

What on earth are we to make of all of this? Two points: First, there’s much heterogeneity within the sex offender population. What is true for offender A may not be true for offender B. It's easy to condemn all sex offenders -- their crimes are terrible -- but malice means being unrighteously spiteful. A just society is just in its judgments and punishments; not all sex offenders deserve the worst punishments. Second, we desperately need better studies that examine recidivism risk over the long haul and can give us truly some understanding of whether sex offender treatment is effective. As I mentioned before, I don’t think there’s much evidence that sexual offending is a mental illness, nonetheless, you don’t need to be mentally ill to benefit from psychological help – and these folks badly need it.

Edit: A helpful comment has shown that Prentky cautioned against applying the recidivism rate he published due to the heterogeneity of the sex offender population. This is likely true -- the sex offender population is heterogeneous -- but his results still stand empirically.


The one thing Mr. Erickson fails to take into account is that some 'sex offenders' aren't really sex offender at all and never should have been deemed as such in the first place. Teens having consensual sex are just one example of some of the types of people that specious sex offender laws defile with the forced label of "sex offender."

Amen, realitcheck! No discussion of so-called sex offenders should neglect the issues of consensual teen sex and other ridiculous offenses such as public urination and "mooning".

The statement, "It's easy to condemn all sex offenders -- their crimes are terrible" contradicts the point the author appeared to be making initially. There's nothing really all that "terrible" about teenage sex, not being able to find a bathroom, or a young person's lighthearted prank.

Some states even include adultery, fornication, and "lewd and lascivious cohabitation" among registerable offenses. Who among you should/could be a sex offender? Hmmm. While possibly morally reprehensible for some, really not all that "terrible".

The great tragedy is that even articles that address the heterogeneity of offenders leave out these types of offenders which STILL leaves the public with the idea that even if not all sex offenders are alike (and some are more prone to recidivism than others),thay have all, in fact, done something that is incredibly offensive and vile.

While I do appreciate articles that point out that not all offenders are alike, we need to educate the public about these lesser offenses. Seeing on the registry that your neighbor committed a crime of "public indecency" or "carnal knowledge" may not really be what it seems! Maybe not so "terrible" at all.

Under "3. Length of follow-up after release." above you are misquoting recidivism rates for child molesters. You describe the rates as applicable to all child molesters but the study you quote from has a AUTHOR WARNING that the rates SHOULD NOT be misapplied. i.e., to all
child molesters.

For a complete explanation of the problem please se the following, as it is quite a explanation, due to an error by Prentky:


"Since sexual offending is relatively rare"

Rare relative to what? Studies routinely find that significant fractions of women report having been raped at least once. Victim ages tend to be young.

Some empirical data on this compiled here:


It's rare in terms of reported crimes. In determining base rates for crimes, we can only make estimates based on reported crimes, not reported incidents.
For what it's worth, it may be worth noting as well there have been numerous criticisms of the types of statistics you point to (which is a USDOJ study). These criticisms range from overly broad definitions of sexual assault to non-representative samples. Whether these criticisms are valid is debatable and would require a lengthy post of its own (which I will try to get to in the future). Nonetheless, even the Center for Sex Offender Management (which I have been critical of for other reasons) acknowledges this fact.

Thanks for commenting.


Reported crimes versus committed crimes is a big problem in research, and it is particularly problematic for sex crimes as reporting can vary based on cultural and social factors which may change over time and vary between jurisdictions.

In the research for this post, we found that the number of rapes per capita reported was five times as high in England and Wales as it was in Italy. Is the number committed that much higher? I very much doubt it.

Leave a comment

Monthly Archives