What shape is YOUR neighborhood in?

Most of the maps we use look familiar to us: for instance, most Americans can recognize the shape of the continental United States whether it's appearing on a weather map or in a newspaper article about the economy.  Sometimes, however, ordinary maps depicting geographic area aren't the most useful method of visualizing our information.

A good example of this was during the 2004 election.  Maps of the election results appeared on every channel and news website, updated in "near real time" as the election progressed.  Typically, these maps were presented as maps of the US with each state colored red or blue depending on whether the state had voted Republican or Democrat respectively.

One major problem with maps like these is that it's hard for our minds to easily combine two kinds of information -- the number representing each state's electoral votes (a function of population) and the color representing which candidate got those votes.  Instead, many people just see the giant swath of red through the south and central regions of the country, giving the impression of a Republican landslide.

Cartograms (wiki link) are one method of solving the problems posed by area-based maps.  Cartograms are special maps where the space on the map corresponds to a value other than land area, such as population or wealth.  For example, scientists at the University of Michigan generated a cartogram of the election results where each county in the US is depicted based on population instead of geographic size.  In addition, instead of using red/blue, they instituted a "spectrum" color scheme of various shades of purple to indicate the close nature of the results in many areas.  The resulting cartogram provides more  precise information than the preceding example - showing that the popular race was very close when compared to the electoral map as well as the unmistakable red/blue divide in urban and rural areas. 

One important thing to note: the county-level cartogram provides more precision, or granularity, than the simpler state-level map.  However, it's basically impossible to tell who won the election just from looking at the cartogram.  The state-level map gives viewers the ability to add up the numeric electoral votes.  This is a great example of how data visualization can be used to tell different stories!

So obviously as a neogeo-geek, I couldn't rest until I found a way to generate these sweet maps myself!  And we might as well make them local and relevant while we're experimenting.

Chicago, the city where MapTogether is based, has been divided into 77 "community areas" since the 1920s - these community areas were originally distinct neighborhoods, though changes over the years caused some to lose their unique character.  The community area borders have basically been unchanged for the past several decades, so they're frequently used for analyzing census data and other statistics.  The City of Chicago's GIS page has digital shapefiles corresponding to the community area boundaries, which is where I started with my neighborhood cartogram project!

For this project, I used uDig to open and edit shapefiles, ScapeToad to generate the cartogram images, Inkscape to polish them, and Kolourpaint for final image scaling and formatting.  ScapeToad is apparently capable of handling multiple variables (e.g., the example above uses both size and color to depict data) but for simplicity's sake, my experimenting was done in one dimension.

For each image, I've got the traditional land area-based community map on the left, and the cartogram on the right.  Cartogram legends show several sizes of squares, representing various values on the map, so you can see what the size of a particular region represents.  One of the primary uses of a cartogram, however, is to show geographic regions relative to each other in a nongeographic context.

First, I just wanted to see the size of population in various community areas.  This cartogram depicts 2000 census population data.  Any area that is small on the left map and large on the right is relatively crowded (lots of people in a little area).  An area that is large in physical area (as shown on the left) and relatively less populated (smaller on the right) is less crowded.  (You can click any of the cartogram images to see a larger version in a new tab/window.)

Cool!  OK, now that we've built one, let's look for something a little more relevant.  The census data has plenty of info on poverty level, aggregated by race and several other variables.  (Side note: the 2010 census can't come soon enough; lots of this data is showing its age.)  Let's examine a cartogram where area on the map (right image) represents the number of families in each area at or below the poverty level.  You can see the bulges in areas corresponding to economically disadvantaged community areas while some of the wealthier regions are barely visible.

Because all of the tools used to create these maps are Free/Open Source Software, anyone can make maps like these with their data.  We'll be on the lookout for good state- and county-level data so we can make more cartograms, but if you have a data set you'd like to see visualized, drop us a line!

Image and data sources:

1. image #1 - 2004 US Election state-based map; public domain from en.wikipedia.org
2. image #2 - 2004 US Election cartogram; CC-BY-2.0: Created by Michael Gastner, Cosma Shalizi, and Mark Newman of the University of Michigan. via en.wikipedia.org
3. Census data for poverty and population maps obtained from the Census 2000 report at the Northeastern Illinois Planning Commission website.