Tuesday, May 24, 2016

Stacked bar graphs and small multiples

Stacked bar graphs are tricky, particularly when you design more than one and you arrange them in a sequence: Only the bottom and upper portions are comparable to each other, as they sit on common baselines. However, there are cases when this graphic form is appropriate. See this elegant small multiple array just published by The New York Times.

What matters in this graphic is not to compare all parties, but to emphasize the hard-right ones, and then to compare them to all other parties as a whole. Therefore, I think that the decision of coloring all center-right and center-left parties identically makes sense: It's red versus white and gray.


  1. I'd have used a couple of gradients because the maker clearly needed to determine a right-ness threshold somewhere, and the current binary classification leaves too much in the dark. I'm sure the original methodology that led to the measure is not binary and thus various levels could have been extracted. It probably would not have impacted the trend or conclusion but would have increased credibility due to the higher, relevant information density that can be achieved by the gradients. For example the current chart lumps parties that have fascistoid origins with something like Hungary's Fidesz which was originally a liberal party and moved to moderate right but possesses none of the ultra right, antisemitic or fascist features Jobbik or Le Pen in France, both allegedly financed by Putin to destabilize the EU, possess.

  2. A part of me likes to think the NYT were inspired by a post I wrote about "ghost stacked bars" 12 months ago: http://www.tableau.com/about/blog/2015/5/vizzing-uk-election-rethinking-stacked-bars-39019