Going Around in Circles
A quick note: I very much meant to post this before I went away on vacation, but just didn’t click that “post” button in time, hence my very delayed response.
There’s a been a lot written about this little graphic by Vox over the past few weeks. The aim of the graphic, according to the accompanying article, is to highlight that we’re not necessarily donating to the right charities, in that those with the fewer deaths seem to be receiving the larger donations.
Note: The original graphic on Vox.com has since been edited. The circles are now sized accurately.
But the way the data has been encoded makes it very hard to ascertain this story from the visual alone. One of the main errors pointed out by Randy Krum on his article over on Huffington Post, is that the circles have been sized by diameter rather than area. This actually has a huge impact on the visual, making the differences between the different figures appear much more dramatic than they actually are, subsequently misleading the viewer.
David Mendoza over on his own blog has suggested that the most efficient and honest way to present this data would have been in a good-old scatter plot. As you can see, by presenting the data in this way the viewer doesn’t have to do much work in order to understand the story. As a viewer I can clearly see the heart disease has the biggest death toll, and yet remains significantly underfunded compared to something like breast cancer. This isn’t as easy to determine from the original chart.
Bubbles, as a rule, aren’t the best way to visualise any kind of dataset. We humans just plain aren’t good at determining the area of something, and then make the relevant comparisons.
That aside, the biggest issue for me is that circles have been used to encode two different kinds of data, these are then organised in vertical columns. At first glance, my eye wants to compare the left column to the right, but the two values (money raised and deaths) are not directly comparable.
The size of the largest circles also appears to be arbitrary. $257.85m in donations is shown smaller than 596,577 deaths, which to me, implies some sort of significance, and yet you can’t actually compare the two in that way. My suggestion would be to make the two largest circles the same size, if for no other reason than to make the size difference as you move down the graphic more comparable, but this is in no way a good fix.
One additional gripe, nowhere on the graphic does it state whether the donation figures are annual. It’s implied they are, but a bit of clarification would have been nice in this case, even if it is mentioned in the accompanying article.
Choose the Right Tool for the Job
It just goes to show how incredibly important it is to tailor the visualisation to the dataset and not the other way around. Different visualisation methods can help to emphasise / de-emphasise different points or stories. Unfortunately for the original graphic, the message is almost entirely lost. Sure, a simple scatterplot may not have the same “ooh that’s pretty” factor, but the message is much more clearly and effectively conveyed. After all, seeking clarity is why we visualise data in the first place and making aesthetic choices over functional ones defeats the point just a little.