Join 3,572 readers in helping fund MetaFilter (Hide)


The pulsing popularity of political parties in America over the previous passage of years
November 22, 2010 9:26 AM   Subscribe

"Isarithmic maps are essentially topographic or contour maps, wherein a third variable is represented in two dimensions by color, or by contour lines, indicating gradations. I had never seen such a map depicting political data — certainly not election returns, and thus sought to create them".
posted by nomadicink (20 comments total) 33 users marked this as a favorite

 
Tie dye!
posted by MisterMo at 9:36 AM on November 22, 2010 [1 favorite]


Looks like battling and spreading viruses. This is pretty interesting, going to have to give some more time to look at how how he did this.
posted by shinyshiny at 9:59 AM on November 22, 2010 [1 favorite]


You can practically actually see the Civil Rights era unfold. Stunning.
posted by joe lisboa at 10:05 AM on November 22, 2010 [1 favorite]


Yeah, joe lisboa, all those white southern dems ran screaming to the republican party when they were compelled to desegregate.

This could be a good tool for a history class.

I love maps, thanks for posting.
posted by mareli at 10:50 AM on November 22, 2010


The earlier map set by David Sparks' (the choropleth presentation) is very interesting and appears to show the very strong influence of gerrymandering. The processing to generate this isarithmic set seems to have suppressed the gerrymandering effect and produced a very different and remarkably dramatic picture.
But interpreting the results? Still trying to get my head around the details of the data manipulation - which should be a fascinating and valuable exercise.
Thank you for an extremely interesting post!
posted by speug at 10:55 AM on November 22, 2010


These colors don't run...
posted by chavenet at 10:58 AM on November 22, 2010 [1 favorite]


That animation is possibly the most incredible piece of data visualization I've ever seen. You can literally see the geographic shadows of great political upheavals for the last century unfold before your eyes.
posted by Aquaman at 11:05 AM on November 22, 2010


What Aquaman said. I'm seriously impressed. Of course, the definition of what a Democrat and Republican are over the past 90 years hasn't been consistent.

It would be very interesting if the animator could derive data on whether an area was authoritatian, libertarian, conservative, or liberal based on the platforms of the popular candidates.
posted by hanoixan at 11:27 AM on November 22, 2010 [1 favorite]


I love the way it shifts along the eastern seaboard - also pure blue during the Great Depression.
posted by migurski at 11:31 AM on November 22, 2010


"the definition of what a Democrat and Republican are over the past 90 years hasn't been consistent"

This animation does a bang-up job of showing that! Throughout the animation, you can pretty clearly see communities of interest form and dissipate.
posted by migurski at 11:32 AM on November 22, 2010


Very nice. Would love to see his R source code, though.
posted by Jimbob at 11:47 AM on November 22, 2010


malecki says:
November 15, 2010 at 3:46 pm
This is really great – on the projection: you can do the transform by taking the same map, turning it into an sp object (a SpatialPolygonsDataFrame), projecting it using
spTransform
, then Hadley’s already made a ‘fortify’ method to extract the polygon coords and you’re good to go. But yeah, I’d definitely recommend using the ‘mapdata’ object instead of a detailed map – they do blow up filesizes! FWIW the USGS (AEA) contiguous states projection string is
spTransform(theMap, CRS("+proj=aea +lat_1=29.5 +lat_2=45.5 +lat_0=23 +lon_0=-96"))


Bastard stole my comment.
posted by Sebmojo at 11:54 AM on November 22, 2010 [1 favorite]


This has convinced me that I need to learn R.
posted by desjardins at 11:59 AM on November 22, 2010 [3 favorites]


This looked like nothing so much as mold growing on toast. And if that encapsulates my general malaise and cynicism about politics, well... so be it.

Cool visualization, though!
posted by hincandenza at 12:21 PM on November 22, 2010


I'm going to have to recommend this to all my friends who teach US history at the high school level. I wish I'd had something like this when I was a student.
posted by immlass at 1:34 PM on November 22, 2010


This is the coolest map I have seen in a while. Anyone taking a US history class should look at this.


I then used simple linear interpolation to create a smoothed transition from election-to-election, creating 99-interelectoral estimates of partisanship for each county. Using a custom function and the interp function from akima, I created a spatially smoothed image of interpolated partisanship at points other than the county centroids.


If I am reading this right, he is saying that he starts with the actual county-level vote then interpolates between elections. So that is adding some visual smoothing that is not really there. Even the static maps look a little too smooth to me. Anyone else.
posted by shothotbot at 2:49 PM on November 22, 2010


shothotbot, there are two dimensions to what he's done, a spatial dimension and a time dimension.

The spatial dimension was done to create a smooth, continuous surface across the US - he used data from each district, and did a linear interpolation to "fill in the gaps". You could do this for each election (every 4 years) separately and plot the maps.

However, it's trivial to include time as a model variable, and smooth over time, so that instead of getting one frame every 4 years, you can model as many intermediate frames as you like.

So what he did, for each county, he produced a trend line over time showing how that county moved from red to blue, then he took all those individual trends and did a spatial interpolation for each year.

It's not perfect - it is "adding some visual smoothing that is not really there", but that's what you get when you try to model highly dynamic, noisy systems with statistical models.

I'm impressed because I've done very similar visualization in R in the past (usually of meteorological data), but have never managed to get them to look that good. I've usually done them all in one step, though; using a Generalized Additive Model to combine a spline spatial model with a spline time model. My method would, however, probably produce something even more over-smoothed.

What the visualization does need, though, is a scale. In the early years, it appears the WHOLE UNITED STATES went completely Democrat...I worry that his colour scale isn't adjusted correctly. You remember how, after then 2008 election, someone came out with these maps showing the Red and Blue states...? Then someone else came out with different maps, where it wasn't a binary division, but rather a continuum, and the whole US was actually slightly different shades of purple? I think this analysis might need to go through that kind of treatment, because it might reveal less extremism than is apparent here.
posted by Jimbob at 4:23 PM on November 22, 2010 [2 favorites]


Jimbob, isn't that what we're seeing as white in this map? E.g., around 1948, when most of the country is a fairly pale color.
posted by hattifattener at 5:03 PM on November 22, 2010


You do see it a bit...but more recent years are still coloured fairly starkly red/blue when we know that in fact most places were quite evenly divided..I want to know exactly what sort of "swing" you need to one party or the other before it starts not being pale.
posted by Jimbob at 5:55 PM on November 22, 2010


30s and early 40s: Bright blue
60s: Bright blue
2008+: Pale blue with plenty of red

But Obama is the socialistest of all time.
posted by DU at 5:38 AM on November 23, 2010


« Older We need to talk about your cat because your cat is...   |   Yehuda Moon and the kickstand... Newer »


This thread has been archived and is closed to new comments