elections and bad graphs

 

Many people are deeply curious about the demographics of yesterday's rather startling voting results.

I think back to Edward Tufte and "The Visual Display of Quantitative Information", and the importance of making intuitive visual charts I'd say this CNN piece ,while providing some critical insight, fails in terms of clarity - for example:

It took me some pondering to figure out what I was looking at: each rectangular block is an *overlay* of the two voting patterns (D/R). So the height of the "roof" is where you see the winner-in-that-demographic sticking up above the votes for the less popular side.

I was pondering how to improve that. One obvious choice would be a stacked graph. Admittedly with so many totals hovering around 50 it might be harder to see who actually won (and the top of the stack, where it failed to hit 100%, would be "who voted for one of the two major parties"). Or possibly glue one candidate to the "ceiling" - though choosing which partyis over the other might look like bias.

Another helpful tweak might be to change the key floating above, so that the sample blue w/ top outline, red with top outline bars were both in middle, and *partially* overlapping. That would make it more clear that each graph leaned on overlap, and so the purple blend was the point where the candidates were neck and neck, so to speak.

Here's another misleading graph:

I saw this posted with the sardonic caption "Thanks Dads!". But Trump voters were just 58% of only 13% of the voters, or about 7.5% - Men without children were "only" 53%, but at 34% of the voting population they were 18%. A good chart would provide a feel both the size of each demographic as well as showing the breakdown.