The red-blue map is deceptive. The shades-of-purple map is actively counterproductive. The area-adjusted-for-population (“cartogram”) shades-of-purple map is simply absurd.

1. Human beings are good at comparing shades of gray.
2. Human beings are poor at comparing shades of color.
3. Common forms of color-blindness render "shades of purple" indistinguishable.
4. Human beings are good at comparing lengths.
5. Human beings are poor at comparing areas.
6. Human beings are even worse at comparing volumes, so don't even think of going there.

What to do? Produce two maps: one in which percentage of support for Bush is represented from 0% (white) to 100% (black), and another in which the same is done for Kerry. When the data becomes available, do this at the precinct level.

If you really feel the urge to adjust for population, then on both maps, project each county/precinct up by the number of residents who voted for that candidate, and publish the results as a true 3-D map (QuickTime VR, VRML, whatever) that can be rotated and zoomed. Resist the urge to project the opposition candidate’s areas down; comparing the length of lines going in different directions isn’t a good idea either.

Update: Source of and links to a bunch of deceptive El-04 maps here.

Update: this one is much closer to useful than the rest, although the perspective makes it difficult to fairly compare populations (give me a 3-D walkthrough!). Thanks, Bill.