California Crime Predicting Company Now Forecasts Where Guns Will Be Used

PredPol, the first company in the nation to sell crime predicting computer programs to police departments, is adding crimes with guns to its list of those it can predict.

The company, which had its first contracts with the Santa Cruz and Los Angeles police departments and has spread across the country and into England, has focused so far on crimes against property, such as car theft, car burglary and home burglary.

It says the new model can better predict where a gun crime will occur 66 better than current practices, which include a variety of computer programs and the instincts of law officers.

"One way to think about it is that you are Moneyballing crime," says company founder, anthropologist Jeff Brantingham, referring to the way the Oakland A's used obscure and unstudied statistics to put together a better baseball team for less money.

"Look at how players were chosen over the decades, with gut instinct. Then along comes Billy Beane and he says 'We're going to do it a different way. We're going to use statistics.'"

So, for example, rather than looking at how many times a player gets a hit, the new statistic would be how many times he gets on base, including such things as walks or getting hit by a pitch. If he leads the league in getting hit by pitches, he may perform better for the team than other statistics would infer.

The program won't say for certain a crime will occur on a certain block the next day, but it can build a model that suggests it may be likely to occur.

For PredPol's gun analysis, the company looks at crime statistics, including possession of weapons, assaults and batteries to predict "hot spots" for the next crime. While some have compared the techniques to the movie Minority Report in which police could predict crime by reading minds, Brantingham compares it more to using computer programs to pick stocks, which use a number of factors to predict a likelihood of whether a stock will go up or down.

The numbers of crimes and their locations are all the data they need, says the company president. They don't look at people's crime records, or when burglars get out of jail, which some agencies might do.

"As a civilian what would you assume if you had a large number of assault calls on a corner at 2:30 a.m.?" asks Brantingham. "Hypothetically, you could assume there is a bar there that doesn't discourage people from drinking too much and doesn't look out for them when they leave. You don't have to have a data base that maps the bar. You just have to know the numbers of crimes on a corner."

The company tested its predictive powers by taking data from Chicago's huge number of shootings and predicting where the next ones would occur and came up with the 66 percent figure. It claims to be 100 percent more accurate in predicting property crimes than past practices.

Kent, England, with a population of 1.3 million, has used PredPol in sections of the county and claims property crimes have gone down 6 percent in its first six months. Los Angeles and Santa Cruz claim burglary and car theft reductions of 12-25 percent.

Here is a TED talk by two of the PredPol founders.

Seattle is the first city to add the gun prediction model.