Erickson on STATIC-99

Steve Erickson at Crime & Consequences has a nice post in part responding to my post about the courtroom use of Static-99.  Here is the key part of his post regarding the predictive tool:

People may scoff at the simplicity of risk assessment measures such as the Static-99, but the truth is these instruments are reliable insofar as they measure known recidivism risk variables that time and time again show (or should have shown as with Duncan) that the best predictor of future behavior is past behavior. This inconvenient truth will bedevil our policy debates about sex offenders and helps explain why it is so difficult to generate public opinion to moderate our criminal justice system's approach to these offenders. We should avoid a "one size fits all" approach to the ever increasing defined population of sex offenders, but until those who advocate reform admit that in cases like Duncan and numerous others the system's best response would have been lifelong incapcitation early in his sadistic criminal career, the public opinion will only continue towards more punishment and retribution.

Despite the fact that Erickson is responding to my post, I don't disagree with his assessment at all.  He is much more versed than I regarding prediction tools.  And I believe generally in deferring to his expert opinion just as I think judges and juries should make their decisions regarding recidivism based upon expert opinion.  My problem concerns the use of STATIC-99 as evidence.  Putting a really rough diagnostic tool before a judge and jury virtually ensures they will mishandle it.  STATIC-99 is nothing like the MMPI or other more complete diagnostic tools.  It simply measures 10 basic characteristics that can be correlated to sex offender recidivism.  Some of those characteristics are seemingly benign history factors (length of prior relationships, having had a live-in relationship, etc.).  They are especially poor indicators with younger offenders (and for that reason STATIC-99 is not supposed to be used with offenders under age 18).  Psychiatrists and psychologists know the limitations of the tool or at least they should.  Judges and juries are not versed in the ins and outs of a tool like STATIC-99.  And for a judge to allow it to enter the courtroom as evidence strikes me as an idea likely to yield a very bad result.

Instead, I support an expert using STATIC-99, other diagnostic tools, review of an offenders history, and clinical evaluation.  Based upon that information, an expert testifies as to recidivism risk and explains how he or she reached that opinion.  However, some prosecutors have attempted to introduce STATIC-99 records directly into the record.  Similarly, STATIC-99 results have appeared in probation officer reports.  In such cases, the fact-finder gets very limited information that is almost always going to result in overprediction of recidivism.  The reliance on experts won't yield completely reliable results because our ability to predict future dangerousness is still very limited.  However, I think allowing competing experts to battle it out is a lot better than allowing diagnostic tools into the record.

Comments

[Ilah]

"People may scoff at the simplicity of risk assessment measures such as the Static-99, but the truth is these instruments are reliable insofar as they measure known recidivism risk variables that time and time again show (or should have shown as with Duncan) that the best predictor of future behavior is past behavior."

STATIC-99 is precisely that: static. It will give the same recidivism risk to a person released yesterday as it will to a person released--and offense-free--fifty years ago. It can measure generalities, but so broadly that resources may be squandered based upon its results. It's a snapshot--nothing more.

Duncan's history isn't a tale of the need for risk assessments, nor for the need for harsher post-release restrictions for sex offenders. It is a tale of the system's failure to use the tools it already had available. The implied logic of relating the story seems to be that it's perfectly acceptable to continue post-release retribution upon sex offenders because Duncan was permitted to work with children even after admitting his crimes.

A public looking to end such a practice might want to look at the professionals who made the decisions in the Duncan case.

[George]

The precursor to the Static-99 was the RRASOR:

"What can be said about the use of the RRASOR as a sex recidivism prediction instrument for this population? The low correlation coefficient as well as the low area under the curve ROC statistic lead to the conclusion that the predictive accuracy of this instrument on this particular group of sex offenders is slightly better than chance." ["Ten-Year Recidivism Follow-Up Of 1989 Sex Offender Releases," State of Ohio Department of Rehabilitation and Correction, April 2001.]

The Static-99 may be more accurate, but is it accurate enough?

The same study also found: "A study of sex offender recidivism done by the New York Department of Corrections followed a group of sex offender releases for nine years. This study found that the rate of return to prison for committing a new sex crime was 6%, compared to the 8% new sex crime rate of this ten-year follow-up study."

So, in the vast majority of cases, there is no comparison to Duncan. It would probably be unfair to compare Duncan the majority of the 6% or 8% who did reoffend as well. In short, child murder for sexual reasons is so rare there is not enough in the sample to do a valid study. However, the FBI did a profile of child murderers in the context of sexual crimes and found the most common reason for the murder was to avoid detection.

If only it were as simple as something like a Static-99.

[Guy]

How often is it that only results from a STATIC-99 assessment are presented before a judge as opposed to being part of a comprehensive sex offender evaluation?