Over at a website called “GOOD”, they have a really bad chart on the effects of bike commuting on obesity. Here’s a static image of the chart:

What makes this chart bad? Well, in the first place, it took me a few minutes to even figure out what the various bits mean. The columns are countries; the green part is the proportion of trips made on a bike, the orange part proportion of trips made on foot. Now, already we have a problem: why the hell did they curve some of the bars? It makes it impossible to see exactly how long the bar really is—and I’m not sure how the chart designer decided how far the bars go after they turn. Did the designer take into account the length of the quarter-circle for the turn when drawing the long bars? Or do you just add the vertical and horizontal parts and ignore the curving part? We don’t know, and because of that, you can’t accurately compare the lengths of the bars. You have to look at the percentages—and if you’re making us look at the percentages to get information, why are you making a chart?
The second part involves the obesity percentages. It took me a while to find them, as it took me a bit to figure out that the decoration on the ends of the bars are incredibly rotund cyclists. Using the cyclists to represent obesity is really stupid, because it’s their width that represents the obesity rate, but because they are circular, a reader’s eye perceives area—and area increases with the square of width. Hence, when you look at the UK and Norway, you would think that the obesity rate in the UK is about four times that of Norway, but it’s only twice that of Norway.
The length/area thing is a key tool to use when trying to lie with statistics— the author of this chart, deliberately or not, is effectively lying to us.
Another big problem with the chart is that, since you need to compare lengths of the bars with their widths, it’s almost impossible to see the trend. You can easily compare countries whose bars are next to each other, but that doesn’t seem to say much—for example, Finland has lots of biking and walking, but their obesity rate is basically the same as that of Canada and Australia, where people walk and bike far, far less than in Finland. If the chart is trying to tell me that biking and walking to work helps you be thin and healthy, the actual data are not bearing that out.
In the end, each country has two numbers associated to it: walking/biking rate, and obesity rate. Why not just make a scatter plot? Then, if there’s a clear correlation, it will be visibly obvious, and outliers can be quickly identified.
In the end, I’m saving this chart, so if I ever teach a quantiative literacy class, I can use this as an example of how not to communicate quantitative data.