You Learn Something Every Day: "Choropleth"

By James Fallows

John Gerth, of the computer science department at Stanford, sent a link to a powerful illustration, by Latoya Egwuekwe, of the spreading effects of unemployment across the country during this recession. The animated version is here, and a sample frame, showing conditions back in the palmy days, looks like this:

Thumbnail image for UnEmpRate.png

The map shows the rate county-by-county, with the hue darkening as layoffs spread.

I actually mentioned this same map several months ago, but at that time I didn't know the "choropleth" complication. It turns out that, in most familiar terms, this is the concept that "geography does not equal population." When we see a Red State/Blue State electoral college map in a close election, it looks like the country is voting overwhelmingly Republican, but that is because Montana, Wyoming, the Dakotas, etc occupy so much space and contain so few people. For anyone who has forgotten, the hair's-breadth 2000 election, in which the outnumbered-looking "blue" candidate received more popular votes than the dominant-looking "red" one:

Thumbnail image for 800px-ElectoralCollege2000-Large-BushRed-GoreBlue.png

The problem is greatly reduced, but remains, in county-by-county maps, like this one showing unemployment rates. My original homeland of San Bernardino County, Calif, which runs from just outside Los Angeles to the Nevada and Arizona borders, is bigger than several New England states combined, but most of it is mountain or desert.

Gerth's original note said the display "has the usual problems of choropleth maps (no weighting for population density), but when things are this bad, it hardly matters."

When I wrote back, "Choropleth????" (before looking here) he explained:

The conventional attempt to fix the weighting problem is to adjust the area by a variable like population resulting in a "cartogram", but this is a non-trivial calculation and is normally done at a coarser scale like states (e.g. electoral vote maps). Other ameliorations include creating callouts for densely populated metro regions like NY-NJ-CT.

I have long loved the Atlantic's use of these for statistics by county. It's amazing how much detail one can see given a little knowledge of our geography.

Would'st that this transferred so easily to other representations. I work in the field of information visualization and we really haven't had that much success in creating marvelous new displays for information which does *not* have a natural representation tied to dimensions of physical space. [And about the word itself,] IIRC, cartographers invented these long before computers. Since cartography (unlike computer science) is a real discipline, they get to use Greek without looking silly.

So there. And for anyone who hasn't seen it, a great comparison of different mapping ways to show presidential-election results here, including this county-by-county cartogram of the 2008 results:

US_Presidential_Election_County_Level_Cartogram.png

This article available online at:

http://www.theatlantic.com/politics/archive/2010/08/you-learn-something-every-day-choropleth/60720/