For Data Visualization, Circles Don't Cut It
During President Obama's State of the Union speech on Tuesday night, the live feed at the White House website featured PowerPoint-style infographics and bullet points to accompany the President's statements. One of those charts was a comparison of the GDP of the United States and its nearest rivals:
Notice anything odd? The 2010 GDP of the US isn't quite three times that of China, but you could easily fit four of those China circles in the US one, if not more. In fact, the area of those circles are proportional to the square of GDP, as Dan Meyer illustrates with amusing snark. Scaling a variable to one dimension (here, diameter) of a two-dimensional shape is a rookie mistake in data visualization, an issue Edward Tufte popularized in his book The Visual Display of Quantitative Information. And given that Tufte was recruited by the White House to improve communication about data that chart is, well, a bit embarrassing.
But even when the area of is appropriately scaled, circles are still a poor tool for communicating information about numbers. Pie charts are circles, and it's well known that pie charts suck when it comes to comparing proportions. And even simple circles are hard to compare, as amply demonstrated in this demonstration from the blog Contrast:
The answer is that the area of A is greater than that of B, C, D and E together, but that's hardly apparent to my eye. The whole post is well worth a read -- it's chock filled with examples of good and ba data visualizations -- and a fine testament to the principles of Tufte.
Other Posts by David Smith
The moderated business community for business intelligence, predictive analytics, and data professionals.