Kentucky tests new assessment tool to determine whether to keep defendants behind bars
"The detention decision is one of the most important decisions made," says Milgram, who was New Jersey's attorney general from 2007 to 2010. Nationwide, more than 12 million individuals are booked into local jails each year, and more than 60 percent of inmates currently in jail are awaiting trial, according to the Laura and John Arnold Foundation, a Houston-based philanthropic organization that focuses on criminal justice, education and public accountability. (See "Prison Break," ABA Journal, November 2013, page 10.) Municipalities spend more than $9 million a year to incarcerate defendants pretrial, the Arnold Foundation says.
Milgram now serves as vice president of criminal justice at the foundation's New York City office, which two years ago issued a pretrial release tool called the Public Safety Assessment-Court that assesses the likelihood a defendant will fail to appear if released while awaiting trial. The tool also diagnoses which defendants should stay behind bars because of the risks they pose to public safety.
DISCRETION AND DATA
Since July 2013, the commonwealth of Kentucky has served as the test case. Judges in all 120 counties have used the PSA-Court tool since then.
"Good, reliable data drive good bond decisions," says Kentucky Circuit Court Judge David Tapp. "It's definitely changed the way I decide. When I was a young judge, I relied on intuition and community expectations. Now I know with verified information what is statistically likely" if the defendant is released.
Judge Tapp says using the PSA-Court has not limited his discretion; instead, the tool has complemented his decision-making. The Arnold Foundation says its goal was never to replace individual discretion, but to enhance it.
"It enables us to make better arguments for our clients," says Scott West, general counsel of the Kentucky Department of Public Advocacy. "It allows a factual argument to rebut gut instinct."
The PSA-Court was derived from pretrial release cases across the country. "We started with 1.5 million cases from over 300 state and federal jurisdictions," says Milgram. "It's the first time a data set this large has ever come together." The data represent rural, urban and suburban districts.
The PSA-Court assesses the predictability of three different outcomes: whether the released defendant will appear in court, will commit a violent crime, or will commit any new crime. The data create a score of one to six, with one indicating low risk and six being high. According to research from the Arnold Foundation, a score of one means a defendant on release is 10 percent likely to fail to appear or commit a new nonviolent crime. If a defendant is a six, then the likelihood of failing to appear jumps to 40 percent, and committing a new nonviolent crime has a 55 percent probability.
The data show also that defendants identified as likely to commit violent crimes were 17 times more likely to be rearrested for violent activity than nonflagged suspects.
The test is based on nine factors, which are subject to change during the trial period and have not been disclosed. The Arnold Foundation will make those factors public when the tool is released nationally, which is expected to be this year. For the first six months that Kentucky used the PSA-Court, the average arrest rate for released defendants declined to 8.5 percent from 10 percent over the previous four years, the Arnold Foundation says. The percentage of inmates released increased to 70 percent from 68 percent.
The PSA-Court is not Kentucky's first attempt at a statistically verified assessment tool. The commonwealth used the Vera point scale, a pretrial assessment tool dating back to a New York City study from 1961 that used strength of family and community ties, prior record and employment to identify defendants who were good risks to appear in court. Kentucky adopted a modified risk-assessment tool in 2006.
In 2011, the Arnold Foundation came to Kentucky to find out what would be the least amount of information that could still determine a released defendant's failure to appear or likelihood of committing a new crime.
Through Kentucky's PSA-Court trial period, it was determined that questions about personal references and employment history were not predictive of pretrial release outcomes. Kentucky Pretrial Services, which assesses and monitors defendants for release and keeps tabs on those who are out, has since pared its statutorily required assessment interview to three minutes from 20.
"It's the most predictive assessment tool we've ever used," says Tara Boh Klute, general manager of Kentucky Pretrial Services.
According to Cherise Fanno Burdeen, executive director of the Pretrial Justice Institute in Gaithersburg, Maryland, only about 10 percent of pretrial service agencies use any kind of risk-assessment tool.
"We have had a risk-assessment tool since 1976, the Vera scale. Over time, modifications were being made to factors used to determine risk that were reactionary. The indicators for determining risk were not research-based," says Klute.
Not everyone is convinced, however, that big data is right for criminal justice reform. Perhaps the most prominent skeptic is Attorney General Eric Holder. On the PBS NewsHour with Gwen Ifill on July 31, 2014, Holder questioned the use of big data on the front end to determine detention. States are already "using that as a predictor, though, of how likely this person is going to be a recidivist. Using group data to make an individualized determination, I think, can result in fundamental unfairness."
The PSA-Court seems to be avoiding the fundamental unfairness that Holder worries about. According to the Arnold Foundation's initial six-month assessment, the tool has been race- and gender-neutral.
However, while the Arnold Foundation released positive results for the six-month period, Kentucky Pretrial Services reports a different outcome at 12 months. There was an increase in failures to appear and new crimes committed by those on release.
When asked about this discrepancy, Klute asserts that she remains confident in the assessment tool. The public safety rate in Kentucky has fluctuated between 90 percent and 92 percent since 2011, she says. During fiscal year 2014, the first year the PSA-Court was used, the public safety rate was 91 percent. Both Klute and Milgram reiterate that the tool is under constant revision and improvement, as more data become available.
From the experience in Kentucky and seven other jurisdictions nationwide, Milgram and the Arnold Foundation intend to expand the PSA-Court tool across the country. But, as Milgram notes, it is not as easy as cutting and pasting the assessment into a new jurisdiction. "[The PSA-Court] needs to be implemented in a way that has fidelity to the research," she says. "Getting jurisdictions to collect the nine factors is necessary so the tool can be used."
Judge Tapp is optimistic about the prospects of applying the PSA-Court elsewhere: "We need to continue to evolve. We need increased focus on better and better data."
This article originally appeared in the April 2015 issue of the ABA Journal with this headline: "Bond Ratings: Kentucky is testing a new assessment tool to determine whether to keep defendants behind bars."