What Political Maps Get Wrong

On election night this month, most news outlets chose to show Republicans taking over the senate with maps. Maps that updated, maps with shading signaling a race lead, win, or loss, and maps with drop down menus that had drop down menus. For those of us who love maps, and love how the internet has led to an explosion in cartography, this might have been a welcome addition to the news coverage. On the contrary, the ways that maps are being employed most often in these midterms and other national elections muddies the very phenomena they are trying to display.

Let’s start with the most misleading map in recent political memory. It’s the county-by-county election returns from the 2012 presidential election, with counties won by President Obama in blue, and counties that Mitt Romney carried in red:

Following Obama’s re-election victory, this map was touted by right-wing pundits across the board, including Rush Limbaugh, because it shows that the majority of the counties in the U.S. voted for Romney. That’s a lot of red. Of course, there is no prize for winning more counties than your opponent, just electoral college votes. Because the population of the U.S. is tremendously clustered in and around cities (approximately 80% the country’s population now live in cities), maps that display counties or states without indicating the discrepancy in population in those areas provide strongly misleading visual information.

The psychology of it is quite simple. The human brain perceives information in a relative way. The bathroom of a concert venue may seem quiet, only because the main stage is shaking with noise. When presented with a map displaying election results, we read area as votes because that’s how we read almost any other chart, from bar graphs to pie charts. Seeing that Montana, Idaho, and Wyoming are all red seems much more powerful than tiny, blue Massachusetts, even though there are more people in Massachusetts than those Northwest states combined. In fact, there are more people in New York City alone than in New Hampshire, Rhode Island, Montana, Delaware, South Dakota, North Dakota, Alaska, Vermont, and Wyoming, combined. Can you even make out New York City on the county-by-county map? Our smallest state, Delaware, has more people than Alaska, which is as big as 21% of the entire contiguous U.S. Maps that plot area of states alone hurt more than they help.

In fact, there are more people in New York City alone than in New Hampshire, Rhode Island, Montana, Delaware, South Dakota, North Dakota, Alaska, Vermont, and Wyoming, combined. Can you even make out New York City on the county-by-county map?

What is most misleading about basic state or county maps being used in elections is the misconception that a given area on a map represents any people at all, let alone a uniform amount of people. The smallest geographical unit the U.S. Census Bureau uses is called a block. In 2010 there were about 11 million census blocks, and 4.8 million of those blocks had a population of zero. Zip. When this fact is mapped, (carefully plotted by Nik Freeman), you realize how much of the country is still decidedly empty:

This point is driven home quantitatively by simply plotting all 50 States in terms of total area versus their total population. In the two graphs below, the second of which has Texas, California, and Alaska withheld because they are such outliers, you can see that there is no direct correlation between state size and population. That characteristic of the country devalues non-dynamic maps which implicitly display area as a proxy for people and votes.

This is not to say that standard maps depicting political results don’t provide any useful information. At different granularities they can adeptly display regional patterns of political support, but that’s so often outweighed by the false messages they signal. The more our country’s population urbanizes, the more these maps will distort the picture of political opinion.

So what do we do? We don’t have to throw away the political map entirely, we just have to find creative ways to reconcile the population disparities. One way to do this is to superimpose the population figures on top of political results. The New York Times has a version of this buried in their 2012 election hub called “size of lead,” and it nicely brings the viewer’s attention to the population centers (such as coastal cities) rather than to vast swaths of the country with very small populations. This map as a whole is less colorful, which allows for visual discrimination between densely and lowly settled regions. Each part of the country is still visible in its normal format (as opposed to “cartograms” that too harshly distort geography), and you can understand what regions are having the most influence on national races.

A simple visualization that I’ve put together is to separate states from their contiguous placements, and put them in order of electoral college votes, as of the 2012 presidential election. This does several things, first, it makes states more individually noticeable (particularly small ones) than if they are colored the same as their neighbors. Second, you can see quickly that of the top 12 states in terms of votes in the electoral college, Obama won 10 of them. There’s no way to see that in a traditional map without requiring a lot of user-effort.

Additionally, you can also see how the size of each state relates to its electoral weight. All of the states here are on the same scale (except Alaska), which illustrates population differences across area. Clearly New Jersey, the smallest state in the top 12 in electoral votes is densely populated, especially compared to Virginia and North Carolina. This map alone does not provide a complete picture of national election results (it inherently misses regional trends), but it doesn’t misrepresent results the way traditional maps so commonly do, because population plays a role in its structure. This basic adjustment to the well-worn U.S. map yields a surprising amount of insights, and further modifications can do the same, without compromising legibility.

To quote George E.P. Box, a famous statistician, “essentially, all models are wrong, some models are useful.” It is the same with maps, whether we’re trying to show the best route to take home, or national election returns. The demography of our country provides a unique challenge to political cartographers, but not something that creative thinking can’t overcome. Let’s see what we can put together before the Iowa Caucus, or Super Tuesday, and make these models more useful.