Some time ago, my local newspaper published an article about operations research applied to an unusual domain - the Allentown police department. While the newspaper does not use the term 'operations research' (its mistake!), the new strategy is exactly that. For instance ("Allentown Police Will March to a New Beat", Morning Call, August 7, 2008): "The city will be divided into four 'police service areas,' which will be further divided into patrol areas or beats. There will be fewer beats, but increased staffing in some, to match demand" - that sounds like a resource allocation problem to me.
Also: "The department plans to classify calls based on their priority, responding to high-priority calls immediately while delaying responses to others, and in some cases, not responding at all." Oh, a queueing system with customer classes! The article also gives a new appreciation of what the Allentown police has to deal with: "Over the last 18 months, police have responded to 127 addresses at least 50 times each, mostly for noise complaints, parking problems, juvenile complaints or traffic complaints."
That is just mind-boggling to me. If you have to call the police every week and a half or so, or if the police is called about you at that frequency, doesn't that say something about how much your life is in a tailspin? (There is this article in today's paper about a man who has been "harassing the police with ''nonsense'' calls [101 calls in two years] about such things as pool water ruining his grass, children making noise and unfounded reports of damage to his property," Morning Call, October 28. Someone needs a hobby.) I guess that's the 80/20 rule in action, or even 90/10 (90% of the work is due to 10% of the people). In mathematical terms, that is called a power-law distribution. Populations of cities and net worth of individuals are considered to obey power-law distributions too.
The August article mentions "the department lacks a crime analyst who can map and track data," and explains that "the plan also calls for the city to implement COMSTAT, a widely used police program that among other things collects and analyzes data as a way to hold police brass responsible for increases in crime." Collecting data to identify patterns, which was popularized by Rudy Giuliani in New York City a few years back, has become an increasingly important part of police work. Wherever there are numbers, there is pressure to fudge them, in particular when it comes to public safety, and the police is no exception, as described in this article called "The Trouble With CompStat", by Robert Zink in PBA Magazine. He explains: "So how do you fake a crime decrease? It’s pretty simple. Don’t file reports, misclassify crimes from felonies to misdemeanors, under-value the property lost to crime so it’s not a felony, and report a series of crimes as a single event."
Sometimes, it seems that people have overly high expectations of what data can do for them. For instance, the article "Lehigh County goes forward on a crime data center but Northampton County isn't ready to follow suit", also from the Morning Call, dated March 30, 2008, states that "If established, the center would provide police with immediate information on crimes committed throughout the Lehigh Valley, and would help establish if a crime committed in Allentown is similar to one committed later in the day in Bethlehem or if a series of crimes may have been committed by the same person."
I think it would be great for police departments to pool their information to become aware of what is happening in other townships, but the farther you go in distance, the more difficult it is to see a relationship. If five cars on the same block are broken into, that might be the work of one person. If five cars throughout the Lehigh Valley are broken into, it's more likely to be the work of five different people. In the hope of establishing a correlation, one has to keep a lot more data, and the more data one has, the harder it is to make sense of it all. The article explains that the center would be staffed by five to eight people analyzing the data - for the center not to be a waste of money, these people had better be good at what they do. While it all does sound like a very good thing, it is probably more needed in areas of dense population.
There is no better example of a good integrated system for police work than what is described in a November 2006 article in the New York Times, "Connecting the Dots on 9 Holdups in 4 Boroughs." It describes the crime sprees of 3 criminals who robbed all-night pharmacies in New York City, two years ago. They were caught because New York had just established a citywide robbery squad, which allowed the police to "connect the dots early on in a string of crimes across several boroughs."
What is interesting about that squad is that it is staffed by detective-analysts, while the Morning Call article seems to imply the staffers at the Lehigh/Northampton Counties would just be 'data analysts' rather than police officers. (The NY police does have a Real Time Crime Center computer too.) Because of the great police work in 2006, which allowed the citywide robbery squad to identify cross-borough patterns, officers were staking out Duane Reades all over the city ("I was dreaming Duane Reades", says one), so after the last robbery occurred, they were able to chase and arrest the suspects quickly. In the end, good data is worthless if you don't have good people to act on it. Maybe, as the Lehigh Valley becomes more populated, we will see more multi-township squads making sense of the information collected by small-town police departments.
This also reminded me of Laura McLay's post "Make better figures and maybe a bar graph" in her blog Punk Rock Operations Research. (I had bookmarked it around the same time as the Morning Call article and never got around to writing about it. Until now.) A few months ago, Laura attended a conference organized by the National Institute of Justice; in her post, she recounts the talk by someone from the San Diego Sheriff’s Department, who described how his departments used to "produce a 28-page report every month", full of data metrics no one really could make sense of - when so much data is available, it is tempting to produce a lot of outputs, just because one can. To its credit, the sheriff's department realized this strategy was not working, and thanks to the book "Measuring What Matters", was able to turn the 28-page monster blob into an 8-page focused report that actually became widely read by the agency. When it comes to data, more is not necessarily better.
Food for thought for law enforcement in the Lehigh Valley.