This is turning up all over the place — at Brad DeLong’s, Crooked Timber, and this pair is from Cosmic Variance — it’s the most sublimely, awesomely, wickedly stupid example of fudging a curve ever. The two graphs below have exactly the same data points, and the only difference is the curve that was ‘fit’ to the distribution. Which one looks plausible to you?
The one on the left looks sensible and simple, and looks like it was actually drawn with some consideration of the data. The one on the right … not so much. I have no idea how anyone could think that particular curve belongs in there.
Now guess which one was actually published?
Hint: it was published in the Wall Street Journal editorial pages.
Someday, somebody’s going to write a book about the shenanigans at the WSJ that allows a clown college, the editorial staff, to exist and thrive within the bounds of an otherwise staid and sort of boring, but respectable, newspaper. That they could actually publish something like the ridiculous abomination on the right and no one said “Wait! What about our credibility?” is merely symptomatic of some really interesting pathology going on there.