It's Not About Sentencing. Police Need to Make More Arrests.
In preparing for a White House conference this week, I’ve reviewed the recent data on the fight against crime. And it’s depressing. For example, the share of violent crimes in the U.S. for which arrests are made is shockingly low -- less than half in 2014, the FBI reports. For burglaries, the share was only 14 percent.
That, in turn, points to the great flaw in how we've gone about fighting crime: We've relied too much on longer prison sentences for those convicted. A wide variety of other evidence suggests that lengthy prison sentences do remarkably little to prevent crime and may well create more recidivism, a point highlighted by the Brennan Center for Justice (on whose advisory board I serve), by the White House Council of Economic Advisers and in a recent op-ed by Jason Furman and Douglas Holtz-Eakin.
A better approach involves not only expanding education and employment opportunities, to provide better alternatives to crime, but also increasing the odds that a criminal will be captured.
Raising that probability seems far more likely to deter crime than longer prison sentences do. After all, most people, let alone most criminals, seem more motivated by near-term consequences (the chance of arrest) than long-term ones. Boosting the perceived threat of capture may even reduce crime sufficiently to lower both the number of crimes and the number of people in jail.
Raising this capture rate, moreover, is eminently doable. Some people might recall how, in the mid-1960s television show, Batman informed Robin that their computer could predict where a criminal would strike next. (I know this because my four-year-old loves the show.) Today, many police departments have made this reality.
In the Italian city of Milan, for almost a decade, police have used KeyCrime, software that predicts robberies based on information about the date, time and location of past incidents, along with details about the criminals involved and their weapons.
Although many police departments across the world are developing similar algorithms, Milan provides an especially good test case because, for historical reasons, it has two separate police forces: the regular police and the gendarmerie. Each day, they alternate protecting different parts of the city. And while the regular police department has relied heavily on KeyCrime (which it invented), the gendarmerie has not. The result is as close as you can get to randomly assigning the software: Each day, it's used in some parts of the city and not others, and the locations constantly shift.
The data from this experiment indicate the software has been effective, Giovanni Mastrobuoni, an economist at the University of Essex, has found. For the police using KeyCrime, the share of repeat robberies leading to an arrest was more than 50 percent higher than it was for the gendarmerie. (That the differences occurred only among repeat offenders demonstrates that the software needs initial inputs about a criminal to predict future activity.)
Even with KeyCrime, the arrest rates were still quite low: From 2008 to 2011, the gendarmerie arrested less than 9 percent of repeat robbers, and the Milan police, a little more than 14 percent.
Beyond improved information technology, it is essential to reduce police response time, which substantially improves detection rates. This requires more resources for police -- which is consistent with the evidence that more police are associated with less crime -- along with more extensive use of predictive software. And it's crucial that police are integrated into the local community.
In addition to all these law enforcement measures, minimizing crime requires improving employment opportunities, education and health care. Simply locking up criminals who are caught and effectively throwing away the key has proven to be the wrong approach.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Peter R. Orszag at email@example.com
To contact the editor responsible for this story:
Mary Duenwald at firstname.lastname@example.org