Annals of Bad Data Visualization

December 15, 2008

Slate had a good idea: provide a way to visualize the bailout. We’re dealing with very large numbers here, so in the average person’s mind it’s easier to put the various pieces of the bailout in proportion to each other if they are represented as shapes with a certain area rather than as a list of numbers. Also, there is the fact that this data is distributed over time; again, it’s easier to represent this data visually as timeline rather than a list of dates. So the task for any visualization of these disbursements is twofold: for any particular disbursement of bailout funds, show the proportion of the overall bailout fund that share represents and also show when it was disbursed.

But what a hash they made of it! This is what Tufte means when he talks about chartjunk: it’s not that the chart data is wrong, but rather that the visual representation of the data that was chosen actually obscures the data itself. So, where to begin...

Jawbreaker

Okay, so there are a couple of things that are not working here with this design. (They call it an "onion" but it looks like a jawbreaker to me.)

Slate's visualization of the bailout
Firstly, they try to represent the disbursement proportions as a series of concentric rings. If you are going to divide up a circle to represent proportions within a whole amount, there’s a time-honored way of doing that; it’s called a pie chart. But Slate wanted to overload the circle with data points: the jawbreaker or onion—whatever—is also a timeline: the inner-most ring represents the first disbursement the outer-most, the last. The problem with the rings it that it’s actually very hard for a person to calculate the area of one ring in proportion to another if the diameters of the circles the rings enclose are different. Let me show you what I mean. Which of these rings has a larger area?

Why the chart is hard to read
Trick question: they both have the same area. But that’s very hard to see and, if anything, my example is simpler to read: there are not 20 other layers of rings between the inner and outer ring as there would be in the Slate example.

Now, there is another possible way this diagram could be read: the width of the ring, not its area, represents the amount of the bailout disbursement. Applying that logic to my diagram above, then the blue amount would be larger (in fact, twice as large) than the orange amount. But if that is the case, what is the point of using a circle? Why not just do this?

A better way to show the data
That’s clear enough: the blue segment, which came first, is twice as big as the orange segment, which came last. But the fact that there is any ambiguity in the way the chart could be read is a bad thing in and of itself. While my visualization of data as segments makes more sense, there is another problem which both my example and the original Slate diagram share: they show the order in which the disbursements were made, but not when they were made. Looking back at my example, the orange segment could have come 6 days after the blue one, or it could have come six months later. The Slate example tries to remedy this by having a timeline below the onion, but the timeline obviates the need to represent the data chronologically in the onion. They may as well have just gone with a pie chart and a timeline. Each would have been a more comprehensible representation of the data.

Doing it right

That said, there is a way to represent this data which makes the proportion of each disbursement clear while also making it intuitive that these disbursements are part of a whole. It also shows the time at which each amount was paid out even more clearly than does the Slate chart and it does so in one single chart, rather than two:

The bst way to visualize the data.
Note that my data here is just made up, but this chart translates easily enough to the bailout data. To be honest, this stuff is really information deign 101. This chart basically represents only one part of a transaction. It shows what the government has payed out and when, but it does not show where the money goes. Now here’s what I think would be an interesting data visualization: show which banks and financial services firms received this largesse and how much they got. Of course, that may be a bit difficult, since the Fed is refusing to disclose this information. Here’s hoping the Obama administration will provide some desperately-needed transparency.

Tagged with: .

Leave a comment:

Comments are closed for this entry.