Advertise here.

iceland.jpg
A Junk Charts reader tipped us about this chart that appeared on the front page of one of their newspapers. It announced: "Iceland banks are in a universe of hurt".

As we all know, Iceland is bankrupt, and all its banks are spinning out of orbit. But is this outer-space concept successful in portraying the story? We think not.

Read the detailed critique below.

MORE

The chart is in essence a bubble chart with the individual bubbles arbitrarily given colors and set into planetary orbits. While the colors don't mean anything and the distance from the center also seems meaningless, the biggest problem is the distortion of scale.

The information in a bubble chart is conveyed through the relative areas of the bubbles. But here, the bubble sizes appear arbitrary too. We measured them, and found that neither the radii nor the areas were proportional to the data shown on the chart. The graph below compares the relative areas of the circles depicted on the chart, and the actual sizes of the banks as printed.

For instance, the ratio between the largest and the smallest bubble was one to 960 but the ratio of the largest and the smallest bank was one to 500.

redo_iceland.jpg

Look for our previous posts on bubble charts and on finance.

Thanks Andrew for letting me post here. And thanks to Arnar F. for the tip.

7 COMMENTS

But it's even worse than that. Planets are spherical, and the volume of a sphere varies with the cube of its radius. So representing values with 3D objects like this is incredibly misleading. The 'sun' in this solar system, for instance, with a diameter roughly 3x that of the largest planet illustrated, ought by rights to have a value some 27 times larger: that is, nearer 50,000m kr than 9,553m kr.

Truly terrible.

Fri 21 Nov 2008 at 9:23 PM
jblondon

They should have used a bar chart, but as Jorge points out in The Inner Beauty of Business Charts, people think they have to sex up their charts to attract attention.

Speaking of bad charts, your line chart is improperly constructed. You should plot bank size as X and bubble size as Y. Deviation from the line Y=X is an indication of inaccuracy. You could place the amount of distortion in data labels.

Fri 21 Nov 2008 at 10:40 PM

Three major problems with the redesigned chart, two of which have been mentioned above:

Spheres show volume, not area. Why does the scatter plot compare area?

Why is the x-axis, which shows the amount of distortion as a percentage, evenly divided into non-sequential and non-proportional values?

What does the most dramatic feature of the scatter plot, the connecting lines between points, in particular the huge spike between the last two data points, actually mean? In this case, it means nothing. The order of data points on the x-axis is arbitrary, there are no intermediate values in the data, and there is no actual or implied connection between the data points. So why does the chart connect them?

This scatter plot has more distortion in it than the original planet chart, though it purports to be a rational analysis. If you want to recycle Tufte badly you could save time and effort by just cutting up one of his books and putting it on the curb in clear plastic. Please, no more Junk Charts!

Tue 25 Nov 2008 at 1:45 AM
Jason

You're criticizing a chart, but *THIS* on the x-axis of your own chart?

0% 365% 46% 13% 24% 24% 43% 29% 91%

If that order isn't bad enough there's a "24%" twice! Come on!

Tue 25 Nov 2008 at 8:25 AM
sebseb

I am the author of Junk Charts and am glad to see strong reactions to the guest post. Andrew is definitely right when he noted to me that the statistical graphics perspective would enrich the discussion in this forum. For those with an open mind, I encourage you to browse through my blog. We feature both good and bad charts, using concepts due to Tufte but also Cleveland, Wainer, etc. In many cases, we provide alternative charts that may work better. The idea is that charting is serious business, and it may take multiple drafts to find the right one. The "right chart" is one that clearly reveals the key message coming from the data, clear of clutter.

In terms of specific points:
- a key concern of Cleveland is how readers perceive sizes and proportions; given that we are looking at a 2-D surface, it is more than likely that readers respond to the area, rather than the volume, of the circles
- the new chart was not intended to replace the original; it serves to point out the degree of distortion. Since the top line shows the perceived sizes and the bottom line the actual, the gap between the lines has this information. The unusual axis labels are chosen to highlight this point.
- putting lines on the plot is a controversial topic. See the posts on profile plots on my blog to see other reactions. My point of view is that the lines merely track the motions of our eyeballs when we read such a chart so why not make it explicit?
- as with the spirit of Junk Charts, we are always open to constructive ideas for better charts

Tue 25 Nov 2008 at 4:56 PM

I absolutely appreciate the discussion on your post Kaiser, it is something that should occur more frequently on infosthetics. I also admire your courage to critique other people's work. Therefore, constructive comments or further ideas for future posts are indeed always welcome.

Wed 26 Nov 2008 at 3:21 AM

http://en.wikipedia.org/wiki/Levels_of_measurement

Wed 26 Nov 2008 at 6:32 AM
ADD A COMMENT
Commenting has been temporarily disabled.