How (Not) to Visualize American Elections
Every two years since 1789, the American political media has faced the problem of how to visually convey the numbers that describe an election. Whereas covering elections in Israel and the Netherlands consists of dividing how many votes each party earned by the total number of votes cast, those attempting to report on the transfer of power within the world’s oldest extant democracy must boil down simple plurality districts, ranked-choice contests, party-caucusing independents, runoffs, and the ever-scrutinized Electoral College process into one visualization. Predictably, they do not always do a good job. While there are effective statewide election visualizations which make use of best practices to inform their audiences of the true number, distribution, and effect of the votes cast, today’s most common visualization technique employs size and color to confuse or even intentionally mislead its audience about the balance of power.
The most used and abused form of electoral visualization is the party colored states choropleth. Since the (in)famous 2000 presidential election, Americans have watched pundits partition a map of the 50 US states into “red” and “blue”. Most frequently utilized for Senate and Presidential races, the primary benefit of this visualization is that it builds off something all Americans know well: the locations and sizes of the fifty nifty United States. Voters and observers can quickly see how any one state voted and get a sense for the partisan geographic distribution of the country through color.
Those who turn to a choropleth in order to watch the expression of their self-determination unfold soon learn, though, that it falls short in several regards. In the Senate, each state’s two-member delegation is equally important to the balance of power. Looking at the above map, one’s natural instinct is to compare the relative amounts of red and blue to gauge which party controls more Senate seats. To do so, however, results in a skewed sense of the importance of large states’ delegations. Despite California being nearly thirty times larger than Connecticut, both have two Senators. Judging by the visual weight afforded to each in the choropleth, though, it is natural to assume that the former dwarfs the latter in significance. By conflating landmass with power, the choropleth fails to grasp the fundamental point of the Senate’s uniform representation.
There have been many attempts to rectify the choropleth’s inflation of sparsely populated land’s electoral significance. In Senate races, each seat is on the ballot every six years and staggered elections occur every two. Thus, only a third of the chamber is up for grabs in any given election year (unless a seat is vacated prematurely and a special election causes both of a state’s seats to be on the same ballot). To show that only a third of Senate seats are on the table and that each state gets two seats regardless of size, the Guardian uses something of a heat map of two hexagons at state centers of mass, each representing one seat. By representing each state’s two Senators separately, the plot uses shading to draw the distinction between seats up for election and not, easily conveying the circumstantial distinction of Oklahoma’s special election. By affording equal visual weight to each seat regardless of land area, this map better answers the question viewers truly care about — how many seats will each party have in the Senate? — while preserving the geospatial relationships of the states.
However poorly the party colored choropleth does at telling which party controls the Upper Chamber of America’s Legislative Branch, it does even worse at reporting on elections for the highest Executive office. Building off the bicameral Connecticut Compromise, the framers of the Constitution designed the Electoral College as a method to elect a President which balances the competing interests of uniform and population-adjusted state representation. Every four years, each state appoints as many electors to the College as it has delegates to Congress — a guaranteed two Senators plus a number of House seats roughly proportional to population. Thus in 2020, Alaska, Delaware, Vermont, Wyoming, and the Dakotas had the minimum three votes while California, Texas, Florida, and New York contributed 55, 34, 31, and 27 electors, respectively. Forty-eight states have laws and statutes which require the state’s electors to vote for the winner of its general election, effectively turning them into winner-take-all contests. Maine and Nebraska divide up electors to candidates proportionally to how they performed in the states’ general elections.
By filling each state with a single color, the choropleth does a poor job conveying the uniqueness 0f these two states’ processes. CNN attempts to do so by drawing a line between the states and a floating red-blue halved circle, perhaps giving the false impression that the parties split the states’ electors evenly. 270towin.com shades them with evenly spaced alternating red and blue stripes, promoting the same false perception. Many other maps simply leave this information out by coloring the whole state as if its delegation behaved the same as the other 48.
Even ignoring the poor job it does reporting these special cases, the choropleth fails to accurately answer the central question in a Presidential race: who’s winning? While it is still possible to count both big and small states on a map to understand Senate elections, tallying the Electoral College requires knowing the Congressional delegation size, and therefore population size, of each state — information that a choropleth simply does not contain. To accept the choropleth’s implicit visual assumption of the correlation between land area and electoral votes is to fall into one of the most common pitfalls of geographical data representation: the population density problem. The states of the Interior West are huge yet sparsely populated, while those of the Northeast are tiny and dense. Visually, 381-square-kilometer Montana appears to contribute more to Trump’s total than 27,000-square-kilometer Massachusetts to Biden’s. In reality though, the Bay State’s eleven Electoral College votes nearly quadruple Big Sky Country’s three. By implicitly correlating land area with votes, the party colored choropleth misleads its viewer about a Presidential election.
Just as there are visualizations which correct the association between land area and voting power in Senate races, so too are there techniques to accurately plot the Electoral College. Among the most effective methods is a “map” in which each state’s shape is discretely quantized into blocks or dots roughly resembling the state’s original borders. Each state consists of as many blocks as it has Electoral College votes and is positioned next to its neighbors in geographically realistic relative locations. This preserves the key strengths of the choropleth — quick lookup and revelation of geographic patterns — while allowing the viewer to assess the status of the overall contest for Executive power by visually comparing the amounts of red and blue. It even allows for an accurate breakdown of Maine and Nebraska’s proportional electors. By treating each state as a grouping of electors rather than a landmass, this visualization effectively communicates the true state of a Presidential race.
Requiring only a blank map of the 50 states and a copy of MS Paint, a choropleth may be the easiest electoral visualization to make. It is, however, among the most misleading ways to visualize American elections in regular use. Those who seek to report honestly on Senate and Presidential elections in the United States should always use a representation which assigns visual weight based on actual significance within the electoral system of the Constitution, not on area. Intentionally or not, those who fail to will obfuscate the truth and mislead their audience.