Remember Moneyball? The book published in 2003, written by Michael Lewis? It argues that the Oakland A’s used data collected at the 161 games each season, year after year, to identify new indicators of success and completely revamp their scouting strategy.
By using readily-available time series data and statistical analysis it was determined that on-base percentages and slugging percentages were better indicators of offensive success than runs batted in and strikeouts. Using these indicators, the Oakland A’s changed their scouting strategy to secure undervalued players without breaking the budget. It changed baseball forever.
If sports teams have been using data to make predictive decisions and get more for their dollar, then why couldn’t local governments do the same thing? In fact, cities are already using predictive analytics to inform a number of decisions. Cities like Santa Cruz, Los Angeles, Seattle and others are successfully using existing data to fuel predictive policing analytic software.
Predicting criminal activity has been a major focus of the U.S. Criminal Justice System. One new approach is Intelligence-Led Policing (ILP). After 9/11, Intelligence-Led Policing kicked off as an approach for proactive policing, rather than reactive, by using risk assessment and risk management approaches.
Intelligence-Led Policing did not replace the community involvement and program-solving approaches of community policing models; it extended them to include research-based approaches, information and communications technology, and increased information intelligence in support of collaborative, multi-jurisdictional approaches to prevent crime, as reported by Charlie Beck and Colleen McCue in their article, “Predictive Policing: What Can We Learn from Wal-Mart and Amazon about Fighting Crime in a Recession?”
Designed at UCLA, the software is built on the same model used for predicting aftershocks from an earthquake. In Los Angeles, the software was piloted for one year and reduced crime by 13 percent, as reported on GeekWire. By generating prediction boxes that are as small as 500 square feet on a patrol map, officers are told to patrol the boxes that are flagged as potential crime areas in their spare time. As of July 2012, the program was implemented in five LAPD divisions that cover 130 miles and 1.3 million people.
The idea here is that police departments have a wealth of data that has been collected over a number of years for every neighborhood and block of a city. By using that pre-existing data that can tell a story about past experience, police cruisers can patrol areas that match the same characteristics to prevent crimes from occurring.
Predictive analytics also goes beyond pre-existing data, focusing on predicting future probabilities and trends based on observed events. It is a multi-perspective approach. Time Magazine reports that predictive policing uses a computer program developed by mathematicians, anthropologists, and criminologists, that integrates reasoning, pattern recognition and predictive modeling associated with domain knowledge.
Predictive analytics is about empowering cities to be proactive, instead of reactive, while still providing quality services and making progress towards goals and objectives.
In summer 2012, Seattle had an unexpected uptick in gun-related crimes. The city increased the number of officers patrolling the streets. As a result, the gun-related crimes decreased, but at high cost to the city. In response, the city began to consider predictive policing software.
In late February of this year, Mayor Mike McGinn announced that Seattle implemented predictive policing software in two precincts. The software uses data from 2008 to predict potential crime and it is estimated to be twice as effective as a human data analyst working from the same information.
By first piloting the software in two of five precincts and focusing on property crimes, such as theft and vehicle crimes, the plan is to revise and adjust the algorithm and methodology before broadening the software’s predictions to gun violence. Lt. Bryan Grenon, from Seattle’s West Precinct, comments, “As a community, we wanted to proactively target gun violence. We hope to implement that this summer.”
In Seattle the piloting of the program was important. Each precinct manages its officers differently, a process for collecting and distributing information was required, and measurements for progress and success needed to be established before the city could think about implementing the software in additional precincts or widening the scope beyond property crimes.
Based on the results seen in Los Angeles, Lt. Grenon is optimistic that the software will be effective in Seattle. Between calls, officers will spend time in identified areas that are flagged for possible crime. He understands that, “on busy days, [officers] can’t spend time there, but other days [they] can spend a considerable amount of time in those areas.”
Lt. Grenon was clear, “we think it’s critical to get our officers into those areas. The message we are sending to the officers is that the goal is not to make additional arrests, but to deter the crime.”
So far, officers in Seattle have been fairly positive about the new approach. With one month of piloting completed, Seattle expects to receive its first crime report since the software was implemented in the coming weeks. In the next month or two they will be expanding the software into five precincts.
For a cost of $73,000 for the software and an additional $45,000 per year for maintenance, the price of the predictive policing software in Seattle will likely limit the need for additional officers on patrol and reduce the number of arrests through place targeted patrolling and deterrence.
There are challenges with any new technology and new approach. Some fear that predictive policing could lead to unlawful stops and searches that violate constitutional rights. Others are concerned that police departments could use the approach to practice racial profiling or stereotype neighborhoods based on the results. As Andrew Guthrie Ferguson, Assistant Law Professor at the University of Columbia, questioned its validity in a courtroom, “You can’t cross-examine a computer – will it stand up in court?”
For all the fears, predictive policing as an Intelligence-Led Policing tool and not a community policing model is meant to be a preventative measure particularly applicable for petty crimes, such as theft and vandalism, and not violent crimes.
By taking living datasets that are strengthened each day, police departments can adopt at-scale approaches to managing crime levels, whether it is at a block, neighborhood or city level.
Using everyday data, in a creative way, applying a model used to measure natural disasters, and combining that in an innovative approach, some cities have propelled beyond others in policing efforts. Can’t wait to see what this model can apply to next!