One of the newer ideas in corr sent is to provide judges before sentencing the findings of assessments predicting likely recidivism and how similar offenders have done with various sentences. I know enough about how the data are put together not to put absolute faith in this, but overall I think it would be a good move to inform decision-makers and to put evidence into practice. However, what happens if a judge does all that and then those well-known statisticians and methodological wizards on the appellate level decide that, well, they’re not convinced? Don’t have to wonder in the 7th Circuit. And it’s not good news, as EvidenceProf blog notes:
The STATIC-99 is an actuarial risk prediction instrument designed to estimate the probability of sexual and violent reconviction for adult males who have already been charged with or convicted of at least one sexual offense against a child or a non-consenting adult. An expert uses STATIC-99 to match a sex offender's characteristics to characteristics found in studies of convicted sex offenders to determine their likely recidivism rate.
The Seventh Circuit was recently presented with STATIC-99 evidence in the case, United States v. McIlrath, 2008 WL 90084 (7th Cir. 2008). 31-year old Christopher McIlrath was on an internet chat room and thought he was chatting with a 15 year-old girl. He wrote her that he would travel to her state to have sex with her, but the "girl" was actually a detective conducting a sting, and he arrested McIrath upon his arrival. McIlrath thereafter pleaded guilty to traveling across state lines to have sex with a minor and was sentenced to 46 months imprisonment. This sentence came after the trial judge discounted forensic psychologist Eric Ostrov's use of STATIC-99 to determine that McIlrath's characteristics matched the characteristics of offenders 9 to 13 percent of whom were found to have repeated their offense.
. . . The Seventh Circuit, however, found that there are a plethora of problems with STATIC-99:
-even its advocates only claim that it has "moderate predictive accuracy;"
-estimates of recidivism are bound to be too low when one is dealing with underreported crimes such as sex offenses; and
-STATIC-99 has too limited a number of potentially relevant characteristics.
The Seventh Circuit then found that McIlrath had not addressed these criticisms and held that "without any effort by the defendant's lawyer to establish the reliability of Dr. Ostrov's methodology -- or even to explain it -- the judge was entitled to discount his prediction." Now, as noted, this was not a decision that addressed the admissibility of STATIC-99 results, and maybe McIrath's lawyer simply dropped the ball. But it certainly seems like the Seventh Circuit was holding that STATIC-99 results are too unreliable to be admissible as expert evidence, and it should be interesting to see how courts in future cases deal with this evidence.