As predicted by data visualization expert Alberto Cairo, this election has brought out a “[maelstrom] of bad data and misleading graphics.” In particular, the red and blue (but mostly red) map of election results has been paraded around as a sign of Republican dominance, purporting to show “sanctuary city islands of blue in a sea of red" or why “Trump has better coverage than Verizon.”
Though we all know those blue areas actually have even more people than the red areas, it’s hard to visualize when they’re so much smaller. It’s natural that we subconsciously weight them lower. Below is my attempt at a better election map, one that gives a truer picture of the results.
County-level results from the 2016 U.S. presidential election
Color = winner and margin of victory, Height = total number of votes (all candidates).
Election maps are telling you big lies about small things
As with most maps that represent information using color, red/blue election maps are great for communicating categorical data (in this case, which candidate won county X?). But they don’t do a very good job conveying magnitude (how important is county X compared to other counties?).
For example, L.A. County alone has a population of over 10 million. That’s more than the combined population of 10 entire states. Maine, New Hampshire, Rhode Island, Montana, Delaware, South Dakota, Alaska, North Dakota, Vermont, and Wyoming together have a total population of just over 9 million.
One alternative that has become popular this year is to map the election results using a cartogram, something Professor Mark Newman at the University of Michigan has been advocating for a long time. His maps deform the shape of each state/county so that each area is sized proportional to its population. The one below also uses a spectrum of colors, rather than just red and blue, to show how close the vote was in each county.
I like cartograms and use them often myself, but they do have shortcomings. Namely, the shapes can become unfamiliar, making it difficult to recognize what the different areas are. Some people also find the deformations weird and uncomfortable to look at.
Another possibility is to use a tiled cartogram, popularized by FiveThirtyEight. It’s less weird-looking than a continuous-area cartogram (whether that is a good or bad thing) and the locations are more recognizable.
Tiled cartograms work great for quickly summarizing state-level results, as they are used on FiveThirtyEight. However, it gets increasingly difficult to build maps like this as you move to finer levels of granularity. At the county level, the hexagons would have to be extremely small to get the sizes and shapes right. For all intents and purposes, it would become a continuous-area cartogram like the map shown above.
A 3D map like the one at the top, sometimes called a prism map is another possibility. By extending each region into the 3rd dimension, it’s possible to show the relative importance of each region while retaining the map’s shape, keeping the areas recognizable. In this case, the height of each county corresponds to its total number of votes, though it could just as easily show population or share of the electoral vote.
For a closer look, see the full screen interactive version.
- Credit to Mark Kearney for the county-level election data, results as of 11/13/2016.
This post originally appeared on Metrocosm.