The Distribution of Democracy
Soon after the 2004 presidential election, maps started appearing all over the internet. Red and blue maps showed the distribution of states voting for Bush vs. Kerry, and soon even more detailed maps appeared showing county-by-county how the votes fell into the Bush/Kerry camps. Because many of the large, empty spaces of the American West and large stretches of rural lands went Republican, the overal effect was of a largely red country with bits of blue here and there. Some people with a point to prove pointed at these maps and declared that the country was overwhelmingly supporting Bush, except for some fringe areas.
The problem, of course, is that much of Kerry's support came from very densely populated areas - the major cities of the Northeast, the Great Lakes region and the West. These votes were more compact, so they took up less space on the map. In response to these attempts to distort the truth, some folks started creating cartograms - maps that instead distort physical reality, changing the shape and size of parts of maps to equalize population density and give an even representation of the votes for both candidates. Cartograms of the red and blue states, as well as all the counties started to appear, showing how the red areas of the United States were not nearly so dominant as the standard maps would imply.
Another point well taken was that most states were very narrowly won by either candidate. States were not really Blue or Red, but somewhere in between, in shades of purple. And so, maps began to appear showing the votes for either candidate, both for the standard U.S. map and in cartogram form.
As for myself, I was curious to play around with the representation of votes in other ways, to see what other details could be revealed, simply by choosing different color ramps and different ways to apply them to the data. I didn't keep all the experiments, but here are some that you might find instructional:
First, the election results by percentage. Nothing fancy here, just showing the actual votes for Bush vs. Kerry. (For the case of simplicity and ease of interpretation, these maps ignore third party candidates, so percent is out of total votes for Bush and Kerry only.) There are actually a few counties that went 80 - 100 % for Kerry, but they're hard to see because they're so small.
There is a difference between percentages and percentiles, and this map is of the latter. Take all the counties, arrange them in order from the lowest to highest percentage votes for Bush (or Kerry), divide them into ten equal parts, and that's how you get the mapped groups here. It's a way of looking at the extremes in the data without making any assumptions about normal distributions and such. Note how the data are grouped in the graph in the large version of the map compared with the graph accompanying the previous map.
And the third in this series looks at the spread of the data along the normal distribution. Note the histogram in the graph in the large image. The county results do fall into the general bell-shaped pattern of a normal distribution. This graph divides up the data based on standard statistical concepts of standard deviations. This way you can visualize how far each county's vote was from the average (i.e., the mean value).
I do have a few more maps that would be fun to include here, but they're currently in rather rough form and need to be tweaked before I can include them here.