Moneyballing Criminal Justice

TED Talk – October 2013

Anne Milgram, Former Attorney General of New Jersey

For a companion piece to Malcolm Sparrow’s book Handcuffed: What Holds Policing Back, and the Keys to Reform (see my review in the preceding post), take a few minutes to watch this fascinating TED talk from a criminal justice reformer who is renewing the drive to use data analytics to make better decisions in the U.S. criminal justice system. And in keeping with Sparrow’s message of collaboration, I would strongly encourage regulators to pay close attention to strategic advancements in the criminal justice system. There is much mutual learning to be had.

Anne Milgram is a former criminal prosecutor and Attorney General of New Jersey (2010-13), who has since moved on to found the Criminal Justice Initiative at the Laura and John Arnold Foundation.

The problem with policing in New Jersey

When Ms. Milgram assumed her post as AG of New Jersey, she soon grasped the lack of data and analysis about who was being arrested and jailed, and whether or not the work of the criminal justice system was making the public safer. In Camden, a city then rated as the most dangerous in the U.S., a group of detectives were assembled to manually trawl through case files and extract key data. When analyzed, this data showed that most arrests the department made were low-level drug arrests, near the police station. Ms. Milgram realized that the New Jersey criminal justice system was failing, in large part because their policing efforts were not data-driven.

Public safety is the most important function of government

In the U.S., 12 million arrests are made annually, 70-80% of which are for misdemeanours; and 75 billion dollars are spent each year on corrections. These numbers track with Sparrow’s concerns with U.S. policing, and with recent media coverage of excessive force being used in home raids by police to obtain small amounts of drugs. Without data-driven decision making, police and courts are focussing on the wrong criminal problems.

Ms. Milgram introduced the use of data analytics (“moneyballing”, a term borrowed from data-driven baseball management) into the justice system. Instead of low-level drug busts, they shifted focus to cases of state-wide importance relating to violence, guns, street gangs, and political corruption. The results were promising. In Camden, the murder rate decreased by 41% and there was an overall 26% decrease in all crime.

She also partnered with the federal Bureau of Alcohol, Tobacco and Firearms (BATF) to gain access to their firearms database, in order to trace illegal guns brought into New Jersey. From a regulatory perspective, I believe this is the kind of data sharing that needs to become the norm, not the exception.

An easy-to-use risk assessment tools for pre-trial judges

Ms. Milgram later joined the Arnold Foundation, where she recruited a team of data scientists, statisticians and researchers. Based on a huge data set – 1.5 million files culled from all levels of the criminal justice system, across America – the team was able to create a simple risk assessment tool for judges to use when determining whether to release individuals charged with crimes pending trial. Ms. Milgram observed that when making these decisions, judges did the best they could based on instinct and experience. But data pointed to the poor job the system traditionally did in this regard, incarcerating low-level offenders and releasing violent criminals and those likely to skip town before trial.

The tool asks only 9 questions, about the current charge, past convictions, and past failures to appear, to predict which persons are at risk to commit another crime or act of violence if released, and which are liable not to appear at their trials if released. The tool is not intended to completely replace judicial experience, but to complement it. As of the date of the TED talk, the tool was being piloted in Kentucky, with plans for a country-wide rollout within five years.

The bottom line: moneyballing works!