August 15, 2006

[Statistics] How To Lie With A Chart

The New York Times printed an article today showing the ridiculously low amount of Americans that believe in evolution compared to the rest of the world.

They pointed to the article summary in Science, but since they knew that most of their readers didn't subscribe to Science, they ran a chart as well. (This links to a local copy in case they change the chart.)

Notice anything odd about the chart? There's a hash mark at the top at the 50% mark, and an unfortunately believable curve of "Yes" answers going down. At the bottom, there's another hash mark. The position of the hash mark at the bottom would lead you to believe that just over 50% of the population of the U.S. believes in evolution. You would be wrong.

So you can compare apples to apples, here is the same chart with a straight line down from the 50% mark...notice anything different?

Now, the obvious defense here is "the bottom hash isn't a hash, it's indicating the label underneath. You know, the one that says 'White areas represent an answer of not sure.'" So why not put that label at the top with the rest of the legend? If "white" is too hard to indicate, why not use a different color?

It's a poorly designed chart, and given that most people are only going to glance at the chart rather than dissect it, it's a potentially misleading chart as well.


Morgan said...

Did you e-mail the editor?

Michael Russell said...

Once I was able to find the E-mail address, yes, but I still have not received a response and the chart on their site is still the incorrect version as of this writing.

Sarkie said...

I love none skewed charts.

Worst Graph Ever Chart