It’s 2018. The 2016 U.S. presidential election is nearly two years in the past. But that didn’t stop the New York Times from unleashing a new map of the 2016 election results earlier this week. On the surface it’s a basic choropleth map: nothing new on that front. But this map drills down a bit further: showing the results by precinct, not just by county. The accompanying article sets out what the Times is trying to accomplish: “On the neighborhood level, many of us really do live in an electoral bubble, this map shows: More than one in five voters lived in a precinct where 80 percent of the two-party vote went to Mr. Trump or Mrs. Clinton. But the map also reveals surprising diversity.”
Kenneth Field has some objections to the map. “So you have smaller geographical areas. Detailed, yes. Accurate, certainly. Useful? Absolutely not because of the way the map was made.” It’s a choropleth map that doesn’t account for population: “An area that has 100 voters and 90 of them voted Republican is shown as dark red and a 90% share. Exactly the same symbol would be used for an area that has 100,000 voters, 90,000 of whom voted Republican.” It gets worse when that thinly populated precinct is geographically larger. (Not only that: the map uses Web Mercator—it is built with Mapbox—so Alaska is severely exaggerated at small scales.) There are, Ken says, other maps that account for population density (not least of which his own dot density map).
The Times map has a very specific purpose, and Ken is going after it for reasons that aren’t really relevant to that purpose. The map is aimed at people looking at their own and surrounding neighbourhoods: the differences in area and population between a precinct in Wyoming and a precinct in Manhattan wouldn’t normally come up. It works at large scales, whereas Ken’s point is more about small scales: zoom out and the map becomes misleading, or at the very least just as problematic as (or no more special than) any other, less granular choropleth map that doesn’t account for population. The map isn’t meant to be small-scale, doesn’t work at small scales, but then people regularly use maps for reasons not intended by the mapmaker. The mistake, I suspect, is making a map that does not work at every scale available at every scale.
Update: See this post for more reactions to the map.
As promised, Kenneth Field has uploaded the final web version of the quick-and-dirty dasymetric dot density map of the 2016 U.S. presidential election results, which he posted to Twitter last month. Unlike the quick-and-dirty version, the final version is in high resolution and can be zoomed in to quite a preposterous degree. One dot, one vote. [Kenneth Field]
Earlier this week, Kenneth Field posted a quick-and-dirty dasymetric dot density map of the 2016 U.S. presidential election results to Twitter. It quickly went viral. In a subsequent blog post, he goes into some detail about the process of making the map. “The screengrab was quick and dirty and while there have been many and varied comments on the ‘map’ it’s by no means the finished article. I want to create a hi-res version and also make a web map like the 2012 version. I don’t have time to do this in the next couple of weeks but it will happen. But I am aware of a number of issues and some have already spotted them as have many others.”
The maps that appear from time to time on xkcd are usually a lot more whimsical than the one Randall posted today: his somewhat belated “2016 Election Map” assigns one figure for every 250,000 votes for each of the 2016 presidential election candidates. As Randall says in the alt text,1 “I like the idea of cartograms (distorted population maps), but I feel like in practice they often end up being the worst of both worlds—not great for showing geography OR counting people. And on top of that, they have all the problems of a chloro… chorophl… chloropet… map with areas colored in.” This is an issue that election map cartographers regularly have to deal with, as many of my readers know well.
We’ve seen a lot of maps correlating election results with other demographic or geographic data, but SurveyMonkey’s exit polling on the correlation between politics and gun ownership seems particularly stark, particularly in the context of recent events. In the 2016 U.S. presidential election, nothing predicted who you’d vote for more than whether you had a gun in the house. If only gun-owners voted, Trump would have swept 49 states; if only non-gun-owners voted, Clinton would have won at least 48.
“Over all, gun-owning households (roughly a third in America) backed Mr. Trump by 63 percent to 31 percent, while households without guns backed Mrs. Clinton, 65 percent to 30 percent, according to SurveyMonkey data,” the New York Times reported. “No other demographic characteristic created such a consistent geographic split.”
Last month the New York Times mapped the U.S. cultural divide by looking at television viewing preferences. More precisely, the geographic distribution of viewership for the 50 most-liked TV shows. The correlation between Duck Dynasty fandom and voting for Trump was higher than for any other show. More surprisingly, the show most correlated with voting for Clinton? Family Guy.
Writing for the Portland Press-Herald, Colin Woodard compares the 2016 presidential election results to the eleven regional cultures he sets out in his 2011 book, American Nations. “The bottom line: the 2016 presidential election results exhibited the same regional patterning we’ve seen in virtually all competitive contests in our history, including those in 2008 and 2012. But by running on an unconventional platform, Donald Trump was able to erode his rival’s margins in certain nations.” He did better enough in rural Yankeedom and the Midlands to deny Clinton the victory in states she could not afford to lose. With plenty of maps to show the swing from the 2008 and 2012 votes. [Cartophilia]
Neil Freeman’s Random States of America creates election maps from an alternate reality. They apply real-world election results to randomly generated state boundaries, which can yield radically different results than what actually happened.
Mapping U.S. election results by county and state is a bit different than mapping results by electoral or congressional district, because counties and states don’t have (roughly) equal populations. Choropleth maps are often used to show the margin of victory, but to show the raw vote total, some election cartographers are going 3D.
The New York Times has a first-rate graphics department, and they’ve come up with some stunning ways to depict the 2016 U.S. presidential election results. They updated their maps of so-called “landslide counties” (see previous entry), which was straightforward enough. Their feature on how Trump reshaped the election map, with arrows showing the county-by-county swing (red and to the right for Trump, blue and to the left for Clinton), was unexpectedly good. But their maps of the Two Americas (above), imagining Trump’s America and Clinton’s America as separate countries, with bodies of water replacing the areas won by their opponents—Trump’s America is nibbled at the edges by coastlines and pockmarked by lakes; Clinton’s is an archipelago—is quite simply a work of art. Incredible, incredible work.
I’ve delayed posting maps of the 2016 U.S. presidential election results because—well, because like many of you I’m still recovering. But here we go. We’ll start with Benjamin Hennig’s cartogram of the results which, as cartograms tend to do, correct for the urban concentrations that made up Hillary Clinton’s vote, and demonstrate the rural nature of Donald Trump’s support. See it at Geographical magazine and Hennig’s website.