October 27, 2013

Black swan crash lands on Florida SVP program

Audit finds low recidivism, critiques reliance on inflated Static-99 risk estimates

Dan Montaldi’s words were prophetic.

Speaking to Salon magazine last year, the former director of Florida's civil commitment program for sex offenders called innovative rehabilitation programs "fragile flowers." The backlash from one bad deed that makes the news can bring an otherwise successful enterprise crashing down.

Montaldi was referring to a community reintegration program in Arizona that was derailed by the escape of a single prisoner in 2010.

But he could have been talking about Florida where, just a year after his Salon interview, the highly publicized rape and murder of an 8-year-old girl is sending shock waves through the treatment community. Cherish Perrywinkle was abducted from a Walmart, raped and murdered, allegedly by a registered sex offender who had twice been evaluated and found not to meet criteria for commitment as a sexually violent predator (SVP).

Montaldi resigned amidst a witch hunt climate generated by the killing and a simultaneous investigative series in the Sun Sentinel headlined "Sex Predators Unleashed." His sin was daring to mention the moral dilemma of locking up people because they might commit a crime in the future, when recidivism rates are very low. Republican lawmakers called his statements supportive of "monsters" and said it made their "skin crawl."

Montaldi's comments were contained in an email to colleagues in the Association for the Treatment of Sexual Abusers, in response to the alarmist newspaper series. He observed that, as a group, sex offenders were "statistically unlikely to reoffend." In other words, Cherish Perrywinkle’s murder was a statistical anomaly (also known as a black swan, or something that is so rare that it is impossible to predict or prevent). He went on to say that in a free society, the civil rights of even "society's most feared and despised members" are an important moral concern. A subscriber to the private listserv apparently leaked the email to the news media.

The Sun Sentinel series had also criticized the decline in the proportion of paroled offenders who were recommended for civil commitment under Montaldi's directorship. "Florida's referral rate is the lowest of 17 states with comparable sex-offender programs and at least three times lower than that of such large states as California, New York and Illinois," the newspaper reported.

Audit finds very low recidivism rates 

In the wake of the Sun Sentinel investigation, the Florida agency that oversees the Sexually Violent Predator Program has released a comprehensive review of the accuracy of the civil commitment selection process. Since Florida enacted its Sexually Violent Predator (SVP) law in 1999, more than 40,000 paroling sex offenders have been reviewed for possible commitment. A private corporation, GEO Care, LLC, runs the state’s 720-bed civil detention facility in Arcadia for the state's Department of Children and Families.

Three independent auditors -- well known psychologists Chris Carr, Anita Schlank and Karen C. Parker -- reviewed data from both a 2011 state analysis and an internal recidivism study conducted by the SVP program. They also reviewed data on 31,626 referrals obtained by the Sun Sentinel newspaper for its Aug. 18 expose.

All of the data converged upon an inescapable conclusion: Current assessment procedures are systematically overestimating the risk that a paroling offender will commit another sex offense.

In other words, Montaldi’s controversial email about recidivism rates was dead-on accurate.

First, the auditors examined recidivism data for a set of sex offenders who were determined to be extremely dangerous predators, but who were nonetheless released into a community diversion program instead of being detained.

"This study provided an opportunity to see if offenders who were recommended for commitment as sexually violent predators, actually behaved as expected when they were placed back into the community," they explained.

Of the 140 released offenders, only five were convicted of a new felony sex offense during a follow-up period of up to 10 years. Or, to put it another way, more than 96 percent did not reoffend. "This finding indicates that many individuals who were thought to be at high risk, were not," the report concluded.

Next, they analyzed internal data from the program itself. As of March 2013, 710 of the roughly 1,500 men referred for civil commitment were later released for one reason or another. Of those, only 5.7 percent went on to be convicted of a new sexually motivated crime.

Interestingly, this reconviction rate is not much different than that of a larger group of 1,200 sex offenders who were considered but rejected for civil commitment after a face-to-face evaluation. About 3 percent of those offenders incurred a new felony sex offense conviction after five to 10 years, with about 4 percent being reconvicted over a longer follow-up period of up to 14 years.

Logo on wall of sex offender hearing room in Salem, MA
"The recommended and the non-recommended groups differed by less than 2 percent in the percentage of offenders obtaining a new felony sex offense conviction after release," the investigators found. "Such a minor difference is surprising and indicates that the traditional approach to determining SVP status needs to be improved. There are too many false positives (someone determined to fit the SVP definition when he does not, or someone determined to be likely to re-offend but he is not)."

Overestimation of risk was especially prevalent for older offenders. Only one out of 94 offenders over the age of 60 was arrested on a new sex offense charge, and that charge was ultimately dismissed.

Finally, the auditors reanalyzed the data obtained by the Sun Sentinel newspaper via a public records request. Of this larger group of about 30,000 paroling offenders who were NOT recommended for civil commitment, less than 2 percent were convicted of a new sex offense.

What the public is most concerned about, naturally, is sex-related murders, such as that of young Cherish Perrywinkle. Fourteen of the tens of thousands of men not recommended for civil commitment had new convictions for sexual murders. This is a rate of 0.047, or less than five one-hundredths of 1 percent – the very definition of a black swan.

Static-99R producing epidemic of false positives

Determining which offender will reoffend is extremely difficult when base rates of sex offender recidivism are so low. However, the auditors identified an actuarial risk assessment tool, the widely used Static-99R, as a key factor in Florida’s epidemic of over-prediction. Florida mandates use of this tool in the risk assessment process.

Florida Civil Commitment Center
In 2009, government evaluators in Florida and elsewhere in the United States began a controversial practice of comparing some offenders to a select set of norms called "high risk." This practice dramatically inflates risk estimates, thereby alarming jurors in adversarial legal proceedings. The decision rules for using this comparison group are unclear and have not been empirically tested.

The recidivism rate of the Static-99R "high risk" comparison sample is several times higher than the actual recidivism rate of even the highest-risk offenders, the auditors noted. Thus, consistent with research findings from other states, they found that use of these high-risk norms is a major factor in the exaggeration of sex offender risk in Florida.

(It is certainly gratifying to see mainstream leadership in the civil commitment industry coming around to what people like me have been pointing out for years now.)

"The precision once thought to be present in using the Static-99 has diminished," the report states. "It seems apparent that less weight needs to be given to the Static-99R in sexually violent predator evaluations."

What goes around comes around

Due to the identified problems with actuarial tools, and the Static-99R in particular, the independent auditors are recommending that more weight be placed on clinical judgment. 

"It now appears that clinical judgment, guided by the broad and ever-expanding base of empirical data, may be superior to simply quoting 'rates,' which may lack sufficient application to the offenders being evaluated."

Ironically, the subjectivity of clinical judgment was the very practice that the actuarial tools were designed to alleviate. I have my doubts that clinical judgment will end up being all that reliable in adversarial proceedings, either. Perhaps the safest practice would be to "bet the base rate," or estimate risk based on local base rates of reoffending for similar offenders. This, however, would result in far fewer civil commitments.

Consistent with recent research, the auditors also recommended re-examining the practice of mandating lengthy treatment that can lead to demoralization and, in some cases, iatrogenic (or harmful) effects.

Although the detailed report may be helpful to forensic evaluators and the courts, it looks like Florida legislators aiming to appease a rattled public will ignore the findings and move in the opposite direction. Several are now advocating for new black swan legislation to be known as "Cherish’s Law."

As sex offender researcher and professor Jill Levenson noted in a commentary on the website of WLRN in Florida, such an approach is penny-wise but pound-foolish: 

“Every dollar spent on hastily passed sex offender policies is a dollar not spent on sexual assault victim services, child protection, and social programs designed to aid at-risk families…. We need to start thinking about early prevention and fund, not cut, social service programs for children and families. Today's perpetrators are often yesterday's victims."

* * * * *

Photo credit: Mike Stocker, Sun Sentinel
BREAKING NEWS: Montaldi has just been replaced as director of the civil commitment facility by Kristin Kanner, a longtime prosecutor from Broward County, Florida who headed that county's Sexually Violent Predator Unit for almost a decade. Not only does she have a JD in law from the Florida College of Law, but she holds undergraduate degrees in psychology and public policy from Duke. Word on the street is that she is an extremely competent and ethical person. It will be interesting to see how she will be treated by the media and politicians in the event that any black swan crash lands on the facility during her watch.

 * * * * *

The full report on the Florida SVP program is available HERE.  

Related post: 

Systems failure or black swan? New frame needed to stop "Memorial Crime Control" frenzy (Oct. 19, 2010)


James Manley, Ph.D. said...

Karen, once again a great job of reporting. The ever-increasing and consistent data refuting the validity of the High Risk group is good news. Now there are at least 15 studies, covering state and federal research, that report low base rates. Interesting will be the response of State evaluators. Ongoing use of the High Risk sample group is clearly not indicated and the perpetuation of which can only reflect a biased evaluator or system.

Anonymous said...

Great article as always! Keep up the excellent work Karen! :)


Jeffrey C. Singer, PhD. said...

Sorry I missed this 6.5 years ago but nothing has happened to undermine the accuracy of this blog piece in the ensuing years. Thank you Karen!