Electoral maps: three cheers for the red, *white*, & blue

While I'm a longtime member of Daily Kos, this is my first diary using a new account.  The old one is pining for the fjords has ceased to be, has gone to meet its maker, is an ex-account.  (This was a voluntary move, not due to a banning.  My identity will be apparent over the next diary or so; for now, feel free to wonder and guess, because I know that we in the netroots love that sort of game.)

I've decided to kick off this account with a geeky, esoteric topic: map coloring.  This diary was almost entirely drafted before Kos proclaimed his love for color-coded maps yesterday, but that still provides a good reason to post it today.

I have a suggestion for political cartographers, the people who create all of those spiffy election maps.  I've made this suggestion before, I believe, in e-mail and the like, but I don't think I've ever done it publicly.  Ready?

Stop using red, purple, and blue in your political maps.  If you instead use red, white, and blue, they will be much more easily understood.

Proof -- definitive proof! -- or at least some sample maps that may prove persuasive, after the jump.

The 2000 election branded into our consciousness the colors red and blue, with fixed assignments to parties, and it didn't take long for people to realize that halfway down the shorter route along the color wheel between red and blue is purple.  (Or magenta, under additive color theory, but I'm going to call it purple here.)  And so we see political maps like the first one inthis well-done effort that indicate intermediate states between the candidates with shades from slightly purplish red through magenta through slightly purplish blue.  (The diary in question, in fact, tries other colors besides red and blue, such as a red and green combination in which brown is the intermediate color.)

The basic problem with the political maps of this sort is that people do a lousy job of distinguishing shades of purple.  Such fine distinctions are not a large part of our everyday experience; they are not meaningful to us.

By contast, people do a much better job of distinguishing shades of intensity for a given color -- especially whether a tint is present or not.  In other words, while people may do a poor job of distinguishing a slightly bluish purple from a slightly reddish purple, they do a great job of distinguishing a slightly bluish white from a slightly redding white.  The map color (or, if you prefer, "non-color") indicating equal support between candidates, therefore, should be white, not purple.  That means that, instead of mixing red and blue to various degrees to get shades of purple, political cartographers should simply include the appropriate portion of red or blue representing the person leading, and leave out the coloring for the minority.   Neutral states then become white; states tilting red or blue range from the palest pastel to the most intense color.  (Note that this would work with any pair of colors in a two person race.)

You'll want to see some examples.  OK, here's a map of election results in the imaginary state of Cartographia in an imaginary contest between Xena and Yanni:

I've labeled the counties in this slide just for the fun of it.  Now I'm going to ignore them almost entirely.  Too much retyping.  (Note that some of the county lines don't show up in some of the smaller maps that follow; I trust the reader to be able to follow the argument despite that.)

I've chosen what seems like a fair distribution of results in Cartographica, using map coloration where a 70% vote yields a "pure" red or blue and intermediate figures yields a mixture, with purple designating a lead of less than 5%:

The results where one candidate has 70% or 65% do jump out at you, but those five ranges in the middle are sort of hard to distinguish, aren't they?  Now see what happens when we attempt a fine-grained analysis in which every single percentage point between 54% and 46% for a given candidate gets its own color where the colors for 55% and 45% would stay as they were in the previous graph:

Pretty much useless, isn't it?  But, you may say, that's unfair; we could not possible expect a map that depicted the middle 40% of the possible election outcomes with colors between pure red and pure blue to do a decent job of distinguishing between values at the middle 10% of possible election outcomes.  If that's what you think, keep reading.

Now, here is the version of the first colored map above where white, rather than purple, designated a tie between the candidates and ranges vary between that white at 50% support to pure red or blue at 70% of the vote for each of the candidates:

This, I suggest, gives a much better sense than its purple-centric counterpart of what candidate was doing better, and by how much, across counties.  (It could be even better, but I haven't spread the intensity of the red and blue hues as much as I could because I reserved all of the more pastel shades for the map I'm about to show you.)  We'll now take the range in between 55% for Xena and 55% for Yanni -- that is, colors less red than that of Lipstick County and less blue than that of Otherwise County in the above map -- and assign each single percentage point its own color using the above scale:

This is the exact same information as that present in the map above that looks like an almost undifferentiated sea of purple.  All we've done is circumvented our inability to distinguish close shades of purple and relied instead on our ability to distinguish among shades of pastels in two different hues.

Here are the two "70% red to 70% blue" maps side-by-side to allow for ready comparison:

And here's the same thing with the "54% red to 54% blue" maps:

I'm not a political cartographer, I don't have access to the raw data that yields these maps or the programs that create them, and even if I did I don't know how to program them to transform them from red/purple/blue to red/white/blue.  But some of you may have that knowledge and power, and I hope that you'll consider running a red/white/blue version of your maps and seeing whether your readers find them easier to comprehend.

(Oh, and before someone asks: what do you do with states that haven't yet voted, which current custom is to leave them as white?  My answer is: color them gray.)

This diary may seem trivial, but increasing public understanding of what is actually going on in the world is anything but.  This is a relatively minor change that I think will substantially boost public understanding.  I'll hope to see some maps recolored using this scheme so that we can see how much of a difference it makes.

•  We should still call them "purple states"

  "Purple states" is a good term, and we certainly don't want to go around calling swing states "white states."  But at some point you have to accept the human visual perceptual system for what it is, and we are just not designed to parse purples.

  If you're one of the people who posts maps here, this may have been e-mailed to you, and I hope you've enjoyed it and that you'll give the red, white, and blue a chance.

  by Seneca Doane on Wed Mar 12, 2008 at 03:23:41 AM PDT

•  Deuterium County, is that where

  the Fusion Power Lab is?

  by Kimball Cross on Wed Mar 12, 2008 at 03:31:08 AM PDT

  •  I'd answer you, but

    I'm afraid that then you'd ask me something about Foo County.

    by Seneca Doane on Wed Mar 12, 2008 at 03:35:50 AM PDT

•  what color is your map?

  That's up to you, my friend.  :)

  by cabaretic on Wed Mar 12, 2008 at 04:00:17 AM PDT

•  Many, many people are red/blue color blind

  and can't tell the difference between the two, which further makes map reading problematical.

  by JFinNe on Wed Mar 12, 2008 at 04:10:01 AM PDT

  •  No, red-green is the most common color blindness,

    followed by yellow-blue. My hunch is this scheme would work for >99% of people.

    I like it a lot! Subtle differences are important.

    by Wee Mama on Wed Mar 12, 2008 at 05:24:14 AM PDT

    •  The reason I know you're right

      is precisely because I am red/green color blind.  Green looks blue to me.  Or what I Think blue looks like. :)

      by JFinNe on Wed Mar 12, 2008 at 06:49:49 AM PDT

      •  What kind of color-blindness do you have?

        Wikipedia offers a decent primer on this.  Your yellow should look like the generally perceived yellow, and your blue should look pretty much like most people's blue.  Depending on your condition, your red is what most would call olive, either dark or medium, your orange and green are both lighter olive, and your purple is a different shade of blue.

        So, given that last, this approach would be a lot easier for you.  Rest assured, though, that "midrange purple" chart is hard for even people with all three sets of cones pigments to perceive.

        by Seneca Doane on Wed Mar 12, 2008 at 11:37:11 AM PDT

•  It doesn't have to be white

  Any third color will do. The point is that you're blending each extreme with a neutral color rather than blending the two extremes with each other.

  On the other hand, blending the extremes does paint a truer picture. The fact is, reality is purple. We fixate so much on tiny differences -- imagining, somehow, that a 58-42 split is vastly more significant than a 56-44 split -- that we overlook just how substantial the minority is, and how much is lost when its point of view is disregarded.

  by Geenius at Wrok on Wed Mar 12, 2008 at 05:30:44 AM PDT

  •  I don't think any third color would do as well

    Let's say that we used green as the third color, with Yanni's support trending from vivid blue through shades of cyan to green, and Xena's from bright red through shades of yellow and brown to green.  That's not as meaningful, to me at least, as when my eyes are looking for the presence or absence of a given shade (light pink, light blue).  I don't think it would stand out the same way and lend itself to ready interpretation.  But it would be nice for one of these cartographers to give it a test so we could see.

    As for the truth of the picture -- well, they're both true.  I agree that a red/white/blue scheme overemphasizes the difference more than a "shades of purple" scheme, but that's pretty much what we want when we're viewing political maps.  While a 58-42 split may not be vastly more significant than a 56-44, it a a difference to note and perhaps get exercised about, as we've seen this campaign season.  A few ticks towards a deeper blue (or red) seems like an appropriate change to designate it.

    by Seneca Doane on Wed Mar 12, 2008 at 11:23:32 AM PDT

    •  I've done thematic maps

      with red and green as the extremes and yellow as the neutral color. It does work.

      by Geenius at Wrok on Wed Mar 12, 2008 at 04:18:31 PM PDT

      •  Yellow would be the best neutral to use

        because it's perceived as closest to white.  But I was not arguing that white is the only color that could work -- although I think it's the best one -- I was arguing against the notion that "any third color will do."

        by Seneca Doane on Thu Mar 13, 2008 at 12:00:53 AM PDT