That Map Does Not Mean What You Think It Means

Last week I showed a wonderful heat map of crime from Trulia, that made it more clear where crimes were actually occurring. Unfortunately, the map does not really have much information; anyone who lives in the area already knows where the crime is. To illustrate, compare the crime map:

Heat map of Pacific Beach, from Trulia.com
Heat map of Pacific Beach, from Trulia.com

To this map:

Alcohol Permits in Pacific Beach
Alcohol Permits in Pacific Beach

Suspiciously similar? The second map shows the locations of alcohol permits in the same area, from the California Alcohol Beverage Control department. In Pacific Beach, most of the crime is associated with (a) a high density of people, and (b) alcohol consumption. The high density of people is due, of course, to the existence of the bars, and the people are in the area in the evenings, during the hours when crime is most likely to be committed.

This a classic case where there are many interrelated factors, and this means that the “crime” map isn’t just a crime map. It’s a also a lots-of-people-drinking-at-night map.

For a community planner, this effect poses a significant problem. With no other information, it would be easy to pursue an ineffective solution to the crime problem.  Additional policing might sound good, but probably won’t work  because most of the police ( and firefighters and ambulances, ) are already in the area. Reducing the number of alcohol permits probably would work but if that were possible, the resident would already have done it. There might be something else going on, but the overwhelming influence of density of people and alcohol obscure any other effects.

If we want to tell the story about how why we should limit the number of alcohol permits, these two maps provide powerful evidence, but if we want to find things we can do to reduce crime in the short term, we’ll have to remove the effect of alcohol and people. Such an analysis might require us to know how many people are in each census block at different times. We might also want to know how much alcohol is being sold, or the demographics of the people in the area. These additional data can be very difficult to acquire, so in a future post we will look at some options for making better guesses and extracting more information from our crime maps.