Hot Cartography Takes

from your friendly neighborhood cartographer

The New York Times published another neat 2016 election map. Yawn. Cartography Twitter has been embroiled in as close to a vicious battle as we can muster which is to say it’s been very polite. The principle criticism seems to be that this map floating around the internet is ugly, not particularly useful and perpetuates some dumb myths about the election. Cartographers are artists, data nerds and geographers which means we tend to run liberal and the 2016 election was a painful and irresistible data set to obsess over. This map is proof that we should stop because it’s cartographic masturbation and it’s making us dumber.

A county map from Brilliant Maps on the left compared to the New York Times map on the right.

While thematically identical to dozens of maps made in the past, what makes this map different is the granularity of the data. Before you could generally only get county results and this map shows results to voting districts which is vastly more precise and useful to political science nerds and Nate Silver.

We know urban areas are blue and midsize cities are blue but even very small towns in deep Trump country like Beckley WV are shown to be substantially more liberal than surrounding areas. You can really see that towns everywhere regardless of size tend to be Clinton supporting whereas rural areas are Trump supporting and that might be as interesting a finding as the notable exceptions which include the Cotton Belt, Indian reservations and Redding CA. Plenty of other maps have shown us this and I recommend looking up Kenneth Field to see some good examples of them. One of my favorites is 3D.

As this map loads, the page asks the audience a question that sets the theme: “Do You Live in a Political Bubble?” It’s an interesting question that the map does not do an effective job answering. One could conclude from this map that liberals never have to travel far to meet a conservative but conservatives are vastly more likely to live in a place where one would have to drive multiple hours to visit a district that voted differently from them. The most securely ideologically isolated are conservatives (at least geographically) despite the popular narrative to the contrary. But this isn’t altogether true.

As I see it, the problem with this map (in addition to the fact that the “3D” function is rubbish) is that for all its granularity it has no visual indication of population density at all and that’s a big deal to answering the Bubble Question. Elections are decided by people, not land. From my liberal bubble in Oakland, the map tells me that the nearest Trump supporting voting district is a 15 minute drive away without traffic.

Gif created by the peerless Dan Rademacher showing the area around Martinez, California.

A.) “without traffic” lol
B.) That Trump voting district I’m directed to reported just 33 votes, of which 18 went to Trump.
C.) Choropleth maps are good at showing vote margins, not actual populations of different voters. Since my district has thousands of voters, I can still probably find more Trump voters in my own neighborhood than in the closest Trump-voting district. Am I still in a bubble? Relevant XKCD.
D.) Nationally, Trump support is greatest where fewer people live which makes them more geographically isolated.
E.) Clinton won the popular vote by several million nationally but many districts were very, very close.

None of these points are adequately addressed by the map and all of them, I would argue, are important to a complete understanding of the political leanings of different regions. Any election map seeking to present a fair representation of how the United States electorate actually voted that does not clearly depict a popular vote win for Clinton is misleading and sloppy at best.

Then again, as I’ve written before, if you don’t like a map, maybe that says more about you than it says about the map. Maps have intention and maps have specific audiences. Andy isn’t wrong here and Jen gave me a real moment of pause; the map is useful for some people. The reason cartographers like this map is because the data is better than we’ve seen before and that’s something we really care about. The reason Andy is ambivalent is because the Bubble Question is less about results and more about political culture and the map does a decent job of generalizing political attitudes which is a useful thing to know even in isolation of the election results. The reason I still don’t like this map though is because I don’t think it accomplishes what it set out to do which is to show how politically divided the country is politically and geographically. Population density and total population are important in answering that question. While the map does show interesting things about small towns that weren’t evident in older maps, it does so at the cost of reintroducing the Myth of the Landslide Election. My take may have been a bit hot. It’s not a BAD map, but it’s not a particularly good one for general audiences. One of the things I love about cartography is that the data is only the backbone of the story. Style is the real muscle and flesh. That same data could have been used to ask more pointed questions that take better advantage of that data.

  • What if you compared voter turn out rates so you can see if an electorate is representative of the ideology of the general population? Since our voting turnout rates are very low, it would be interesting to see if motivated political minorities are electing politicians that are generally unfavorable to their district’s constituencies as seen from polling data. Do blue state Republicans vote as often as red state Democrats?
  • What if you had two maps side by side that show districts extruded in 3D by total population of Trump voters and of Clinton voters on separate maps instead of showing the vote margin between them so you see that the total number of Trump voters in urban areas is still larger than the total number of Trump voters in rural areas?
  • What if you could show the elective power of each individual vote weighted by senator, member of congress or electoral college vote?
  • The New York Times map shows that even very small towns tend to vote for democrats. Is there a threshold population density after which a district is more likely to vote for a Democrat or a Republican? Could you determine the population density of each district and create a distribution plotted against the electoral result margin and see which towns are disproportionately liberal or conservative for their densities?

Of course, none of these hypothetical maps effectively answer the Bubble Question either but they’re probably more interesting maps than simply republishing the same results in the same style with slightly better data. This deserved to be a republishing of the old maps but it probably doesn’t deserve much publicity beyond that without additional context. At conferences, all good research presentations end with a call for more research and all good maps call for more maps that better seek to describe a very, very complicated world. No one map can do the job. This map occupies an important but very crowded space in a much more interesting ecosystem of graphics, data and truth.