Ron Paul confuses the charts

Andrew Sullivan (link) re-printed this grouped column chart showing the result of a Washington Post-ABC poll on how voters say they would react to Ron Paul running as an independent candidate in next year's U.S. presidential election.

One aspect of this chart bothers me... depending on one's familiarity with the election politics, the need to read carefully both the titles at the bottom of the chart, and the legend, and possibly also the title of the chart (or the knowledge that the Republican wears red and Democrat blue) in order to orient onself. You can experiment by blocking out one or two of these three items.

Here's the same chart with a small number of fixes. Printing the legend onto the bars themselves makes the data more readable. This change necessitates flipping the columns over to horizontal bars. There are pros and cons to using a stacked chart versus a grouped chart.

Neither of these charts answer the burning question in the reader's mind, which is likely to be from whom would Paul take his votes. The key message from above is that the insertion of Paul is projected to make the identity of the Republican candidate irrelevant. The following flow chart emphasizes the shift in votes as opposed to the vote totals.

It appears that the Others/Undecided voters who can still swing the election do not consider Ron Paul as a desirable alternative. Most of Ron Paul's supporters would come from voters who would have cast their votes for the Republican or Democratic candidate (by a ratio of 3 Republican votes to 1 Democratic vote if Romney is running, or 3 to 2 if Gingrich is running).

Comments

[Brandt]

Isn't it easier to read, and just as informative, to simply put Paul between the others on the stacked chart? http://dl.dropbox.com/u/21362/two-or-three-party.jpg

[Jon Peltier]

I like the horizontal bar chart with labels on the bars. But I think it would be better keeping the bars grouped. Without the numeric labels, it's hard to tell that Obama's 51% beats Gingrich's 43%.

I like your use of a flow chart to track where Paul's votes come from.

[Jon]

It would be nice to see how an Obama vs Paul election would go too. That's probably a valuable piece of information.

[John Roth]

To me the truly interesting piece of information is that both results with Paul as a spoiler are almost exactly the same: two of them are identical and two differ by one point.

None of the versions make this obvious, although the flow chart at least makes it reasonably easy to find.

[Tom Hopper]

I am not a fan of the flow chart, as used here. The story of this graph is told almost entirely with numbers; take them away and the reader would learn nothing, except that some colored boxes point to others; the graphical elements serve only as highlighting for the numbers.

I think that a web plot, or bipartite graph, in which the width of ribbons connecting each side of the graph represent the number of votes transferring from candidate A to candidate B, would do a better job telling the story. Such graphs can be created in R with the "bipartite" package, and an example graph (much busier than your example) can be found here: http://www.oga-lab.net/RGM_results/bipartite/bipartite-package/bipartite-package_001_big.png

Another alternative might be to combine a cladogram-style decision tree with your bar graph of vote percentages. Something a bit like: http://www.scielo.br/img/revistas/babt/v52n6/a14fig02.gif. This might be the best option, as it would tell both stories in a single visual, and encourage thinking about the election results in detail.

[Account Deleted]

Adjusting the width of the arrows in the flow to show the relative sizes of the flows would make the chart clearer, and even better would be to adjust the sizes of the boxes too. (This would in effect give a bipartite graph as suggested by Tom Hopper.)

[Antonio]

Do horizontal bar charts violate self-sufficiency principle? IMO, it seems so. Indeed, I would not be so sad. I've never been sure that self-sufficiency should be always a virtue, and this case would be an example.

[how to convert flv to mp4]

Obama-Romney thing is really weird for me.
Hard to tell what i felt when Romney won here in Florida.
But i still think Obama is going to be re-elected.