Hi,
I am looking for feedback on a graphical method I’ve devised for comparing groups. I think it can be used for many different kinds of quantities, some of which I show on my website http://mason.gmu.edu/~
The plot shown below (and at http://mason.gmu.edu/~
This representation shows several interesting features of an election at once. The intensity of support for each candidate can be easily compared among states by looking at the widths which are aligned along a common vertical axis. The box height represents the total number of voters living in a state in which a given candidate won. This is similar (but with important differences) to the electoral college votes of that state. The box areas give a measure of the importance of that state’s margin in building the national popular vote margin for a candidate.
One can see at a glance that about 40 million people voted in states that went strongly (> 10% margin) for Clinton, 25 million in states that voted strongly for Trump, and around 20 million in states that went narrowly (<2%) for Trump. Clinton’s national total relied on a few states that were both populous and highly partisan, while Trump had a large number of low-population partisan states. Collectively the states Trump won had substantially more voters than the Clinton states (75 million to 55 million) even though Clinton got more total votes.
I included the cumulative graph on the right of the stack in order to facilitate adding together the margins from different states in each stack. This shows that just the 8 states which were the most pro-Clinton accumulated as much of a margin (number of votes) for her as all the pro-Trump states did for Trump.
The Washington Post and Wikipedia have been showing cartograms of the US to represent the vote; the stack of boxes shown here loses the geographical information but makes it easier to see the relationship between numbers. I think this conveys a lot of information quickly and efficiently and is a good tool for understanding and comparing elections… but I’m curious to see what other people think.
Thanks.
Hi – it took me a while to grasp how to read the chart, but once I had that, I found it insightful.
One suggestion is for the colour scheme. Firstly, I’m not sure what the population density adds to the overall story. For a non-US person, I find the number of electoral college votes per state more interesting. Could the colours be used to represent that? Regardless of what the colours are used for, I think the current sequential palette could be improved – the saturation seem off, eg the lilac looking much softer than the others.
But overall, great to see new spins on this type of dataset
Tom
Thanks Tom. The colors aren’t central to the idea here. I discuss it more on the web page linked to. It was motivated by noticing that the D states included some geographically small ones which still had pretty big populations – ie, states with high population densities. So do crowded states vote D and sparsely-populated ones vote R? I think the answer is mixed – almost all the purple states (high density) are D, and almost all the yellow and orange (sparse) are R (some high density ones are also swing states with pretty low margins), but the fact that WA and OR are pretty firmly D suggests that it may be more a question of geography (ie west coast) rather than population density.
Which states had inconsistent colors? I don’t notice it on my computer. Graph was generated with Matlab, just used their colors specified with a few different RGB sets.
Yeah, similar to first comment – found it a bit confusing at first then insightful.
Yes the colors could be better for example just taking a smoother gradient https://www.flickr.com/photos/kylehailey/2659893170/in/album-72157617786524697/ or a single color gradient https://www.flickr.com/photos/kylehailey/2659066021
The thing that sticks out to me is the small margins at the top but tbig impact. The impact is hard to determine thought as it’s the electoral college votes and not total voting that makes the difference.