Although it sounds like something from a science fiction movie, the Santa Cruz Police Department is using a computer model to tell officers where a crime is going to occur before it happens.
The new practice, called the "Aftershock Model," was developed by Santa Clara University math professor George Mohler and tests have proven it to be 71 percent correct in its first three weeks of tracking home and car burglaries.
Other cities have tried other ways to predict crime. This is the first to use a formula that compares crimes such as burglary to an earthquake. It predicts that similar crimes will follow, like aftershocks.
Then it goes a step even past models tried in Minneapolis and Arlington. It tells officers the time of day a crime is likely to occur and locates it within a quarter of a mile, or three city blocks.
While Santa Cruz is the first testing ground for this project, the computer study was first funded by the Los Angeles Police Department, which plans to put it into effect in the fall.
"There were a couple of blank faces when we first explained it to the department," said police spokesman and crime analyst Zach Friend. "What else can we do? We have 30 miles to patrol and fewer people to do it with. This helps narrow it down."
The department started the program July 1 and it had one measurable success last Sunday, when officers caught a car burglar in a downtown garage where the computer predicted there would be a burglary. There may be plenty of other successes, but the model won't show it for some time because the presence of police should scare away burglars and stop the crimes.
Santa Cruz, with a population of 59,000 gets about 400-600 burglaries a year, including car break-ins, car thefts and items stolen from homes and garages.
Mohler, 29, who wrote the software in several weeks after years of studying crime at UCLA, agrees with other criminologists who say that burglars generally work in a small area and will return to the scene of a successful crime.
Each day during their briefings, Santa Cruz police get a list of the 10 most-likely places burglaries could occur that day based on the crimes that occurred in the weeks before. Some of them seem obvious, such as the downtown shopping district.
But others are contrary to expectations. For example, the model predicted the possibility of burglaries in several neighborhoods between Ocean Street and Soquel Avenue from 8 a.m. - 10 a.m. Those are hours when people have left for their jobs, but not ones usually thought of as times criminals work.
Not so, according to reports.
During a ride-along on patrol one day last week, Lt. Larry Richard could see why. There were car windows open all over. Car doors were unlocked. It almost seemed like an invitation for someone walking by to snatch something.
When we pulled into the parking lot by Gold's gym, a woman approached saying she wished he had been there a week before, when items were stolen from the back of her pickup. It was a tough lesson to learn, she said.
However, when we drove past, there were still many expensive items in the pickup bed, including a vacuum cleaner.
"It doesn't look like she really learned it," said Richard.
People who don't report burglaries make the system less effective. The model works best when more crimes are mapped out.
Some officers seemed skeptical about the program, thinking that years of patrol had honed their instincts to know where crimes are likely to occur.
However, Friend said that this program can help pinpoint things beyond instinct. For example, if there are burglaries in the area of Lighthouse Point only on Mondays and Tuesdays and an officer knows that, another officer working on Thursday or Friday may not know the days are so limited and waste time that could better be spent elsewhere.
This helps refine the focus of their patrol and gives them something to hang their gut feelings on, said Mohler.
Friend learned about the program from a Los Angeles Times article and found to his surprise that Mohler had moved from UCLA to Santa Clara, over the hill from Santa Cruz.
Police gave the mathematician eight years of crime records to build his model, including one that they left blank as a test. He made predictions about the missing year and they turned out to be correct.
That gave them the confidence to move ahead with the study.
Mohler said there are behavior theories behind why the math proves out. For one, criminals go back to where they have had success. They also tend to operate near their homes in neighborhoods they are familiar with.
His model puts dots on a map indicating where a future crime is expected.
"This allows police to rank different areas of the city and tells them where they should patrol every day and where they should send more people because of the higher risk of crime," he said.
Mohler said he thought the program could lessen crime by preventing it from happening. If a kid who was going to steal a CD player saw a patrol car, he might not do it, he said.
It's a different thing than a professional burglar who is stealing big items from homes.
The department will test it for six months and study the results. Friend says it can't lose either way because they will be no worse off from trying it.
"If it works, I'd expect the burglary rate to decline," he said, adding a corollary. "The prediction ability will also decline because there would be fewer burglaries to add to the data."