Bad Neighborhood? Check the Cartogram

When I first moved to Washington, DC to join the Physics Central team, I had to quickly find a place to live. Although I was staying with some wonderful hosts who provided a comfy couch, I needed to find an apartment ideally in a safe, affordable and lively neighborhood – no easy task in DC.

Safety was one of my top concerns for some of the neighborhoods I surveyed, but reliable and useful information seemed out of reach. I heard plenty of anecdotal evidence about how terrible or tolerable certain neighborhoods were, but I had to take that advice with a grain of salt. The official crime data, which is overlaid on maps online, wasn't very helpful either.

Typical crime maps allow the user to see a selection of crimes (e.g. robbery, homicide, and theft) in a small radius for the past 30 days. But these maps don't account for population density, so they don't give the user a good idea of the likelihood of becoming a crime victim in a particular neighborhood.

Thankfully, a team of Argentinian physicists has combined population densities and geographic data to create crime cartograms – a much more efficient way to quickly assess a region's safety.

A population cartogram of the 2004 U.S. presidential election. Red states voted for George W. Bush; blue states voted for John Kerry; and states are sized relative to their population. Image Courtesy Michael Gastner, Cosma Shalizi, and Mark Newman from the University of Michigan.

Cartograms allow researchers to visualize a single variable, such as population or crime frequency, as geographic size on a map. For instance, the cartogram above looks similar to a traditional map of the United States, but the states have been re-sized to reflect their populations.

Herein lies the usefulness of cartograms: viewers can quickly attribute the size of a map's region to a chosen variable. Additionally, cartograms can color code regions with another variable, such as the voting preference of a particular state, allowing two layers of data visualization.

Although cartograms have been around for awhile, they started off looking pretty clunky. With newer software, however, researchers can now create cartograms that maintain the map's general shape while accurately representing the underlying data.

Argentina-based physicists Karina I. Mazzitello and Julian Candia decided to apply this method to homicide rates throughout Brazil. While the cartograms made problem areas more readily apparent, they also allowed the researchers a glimpse into what may contribute to higher crime rates.

The usefulness of the cartograms is probably best explained visually (All of the following images are courtesy Karina I. Mazzitello and Julian Candia via their arXiv article):

Original Map

An unchanged map of Brazil with color-coded homicide rates for each region. Darker regions have more murders while lighter regions have fewer.

Cartogram with Population Density

A cartogram with the regions re-sized to reflect the number of homicides per 100,000 residents. Meanwhile, higher populations are darker and lower populations are lighter. This cartogram reveals the higher crime rates in the northeast that may have gone unnoticed in the first map.

Cartogram with Socio-Economic Data

Finally, the team created the same cartogram as above but included socio-economic ratings in their grayscales. Darker regions have less socio-economic support at the municipal level, and these regions tend to have higher homicide rates. The authors noted that this correlation is much harder to detect in other graphs, such as scatter plots.

I think it would be fascinating to create crime cartograms for U.S. cities, allowing people to better evaluate crime rates at the neighborhood level. Although this requires quite a bit of data fetching and programming knowledge, it certainly can be done.

Crime data is available online for many large metropolitan areas, and researchers have provided the tools to create cartograms. Looks like I've got some homework to do for the next time I decide to move.

-------------------------------------------------------------------------------------

If you want to keep up with Hyperspace, AKA Brian, you can follow him on Twitter.