Design & Geography

There is no ideal way of giving a snapshot of an election, particularly for a country the size of the United States, so lets look at three ways of visualizing it (including 2 methods created with the help of my sister at visualizingeconomics.com). Typically the election has been visualized geographically by the media focusing on who will, or has won, each state and county in what are technically called choropleth maps. John King of CNN does a particularly good job on the CNN interactive election map zooming in looking at vote percentages and turnouts vs. previous elections. These methods serve their purpose in clearly showing who has won the Electoral College vote and highlighting key areas within each state that have contributed to the victory. In this case use of political geographies is quite useful. However, this has led to the map “area bias” at both the county and state level.

This occurs when the variables of interest, such as votes, are not mapped directly but a land area is filled in as a proxy. For election maps these tend to be states and counties. Combined with this issue the use of only two categories Obama vs. Romney winning or losing a state has contributed to the false dichotomy of red state/blue state America. So the mountain states like Montana or plains states like South Dakota seem to hold a larger visual weight than Pennsylvanian even though Pennsylvania is much larger, over 12 times the population of these states, and has over 6 times the numbers of electors. On election night I was at a DC digital week event and the person next to me remarked that it looks like Romney has one Florida because of all the red counties.

Looking at the top county map we used three categories for each candidate: close results less than 10 percent, a moderate win 10 to 40 percent, a very large win greater than 40 percent. We have moved from red state to red county but it looks like a Romney win visually. This map shows a seeming giant stream of support for Romney not in the south but the prairies and mountain west into Nevada. Visually that is what dominates the story. The color schema helps limit the negation the voters on the losing side but the complexity of voter patterns is still simplified.

The first way to give the viewer another perspective is the cartogram map (bottom map). The counties have been changed geographically to account for their size in total population using this software. Using the same scale we see a different story of large urban areas of Obama support on the coasts and great lakes surrounded by light blue or red suburbs. To give a quick feel, LA county which has about 10 million people is now the same size as Michigan just to the right of Cook County. Montana (just under 1 million people) which looks like it has been pressed down on like the rest of the plain states is closer to the size of DC (both about 650,000 people). Brooklyn and Miami-Dade are about the same size on the cartogram map each with about 2.55 million people and are larger than all but a few counties. The intense Romney support is in small rural counties that have turned into “webs” of support in the interior of the country. The suburbs tend to be more light blue and light red around medium to dark blue urban cores. This map shows the regional, urban, rural, suburban character of the election better than the other top map although certainly not without problems.  Visually as the large dark blue urban areas predominate, it looks too much like a big Obama win more than 3 percent at least. In some counties particularly large western counties the area of analysis does not remotely correspond to the variable of interest, votes, so LA County (and neighbours) might appear filled in but much of it is desert even though 10 million people live in this county. Cook county (Chicago) 5.5 million people is almost completely populated although not evenly. The area bias problem still exists with the cartogram map but the rural bias has been reduced. You can compare this map with a different cartogram color schema here showing the 1932 Presidential election.

The second way to show the votes are with a dot density map. In our map each 2000 votes for either candidate are represented by one dot. This shows a more textured view in urban areas especially in the eastern portion of the United States. It also shows red in heavily blue counties except for places like DC or San Francisco which are all blue on the map and close to 90 percent Obama in real life. One drawback with county level data is in the large western urban counties like Las Vegas in southern Nevada or LA County the dots are too dispersed as the counties are larger than the urban areas. In addition the rural areas seem to disappear into the background, less so east of the Mississippi. Despite these problems it weights the geographic regions well showing that Romney like Obama got most of his votes from medium and large urban areas. It also shows the closeness of the election visually unlike the first choropleth maps and does not distort area like the cartogram map. These three maps together and their relative strengths provide an excellent snapshots of the election for national voting patterns at the county level.