Ye Olde Pie Chart Debate
You may think that the debate over pie charts was a new one, but it has raged on for at least 100 years. Brinton started it in 1914, and great drama unfolded in the pages of the Journal of the American Statistical Association in the 1920s.
It all started with Willard C. Brinton – actually, it probably didn't, but that is the first reference I was able to find. Brinton, in his 1914 book Graphic Methods for Presenting Facts, shows several "circle diagrams" (the term pie chart seems to have come later):
Brinton was not a fan. In both figure captions, he criticizes the charts. He also makes some interesting statements about how they might be misread.
Eells' 1926 Study
Things got more interesting in June 1926, with The Relative Merits of Circles and Bars for Representing Component Parts by Walter Crosby Eells. Eells seems to have been bothered by his colleagues' criticism of pie charts without much research. He opens his paper by citing Opinions of Authorities, including Brinton:
There are more authorities listed, but Eells ran out of space or patience (or both) to quote them all. There was not shortage of criticism.
Eells then went on to run an experiment where he asked his students to judge the values on a number of pie charts. He found that his students actually did better with pies than with bars when judging fractions, and he also found some interesting gender differences. In Eells' words:
He didn't just ask them to read the values off the charts, but also to tell him how they were reading them. In particular, he responded to the claim by a man named Horace Secrist that pie charts were read by chord length (the straight line between the two endpoints of the slice on the circle). He found that half his subjects said that they were judging the charts by arc length, and a quarter each by area or central angle. In his charming 1920s way, he characterized the only person citing chord length:
Having established the superiority of pie charts, as well as the mechanism, he probably considered the case closed.
That, of course, could not be left unanswered. In the March 1927 issue, R. von Huhn's (great name!) Further Studies in the Graphic Use of Circles and Bars pointed out a long list of shortcomings. It was followed by some data from a study Frederick E. Croxton had run several years before, but apparently had not bothered publishing until he saw Eells' work.
Only a few months later, in December 1927, Croxton and Roy E. Stryker had a follow-up paper in the same journal, Bar Charts Versus Circle Diagrams. They were only concerned with "accuracy of judgment," and didn't have time for such trivial matters as "popularity and appeal":
Croxton apparently kept working in this area. In 1932, he published a paper with Harold Stein on Graphic Comparisons by Bars, Squares, Circles, and Cubes. The matter of graphics being used carelessly or without enough context clearly bothered him. It's nice to see that some things have not changed in eighty or so years. Modern infographics and their often sloppy use of graphical representations of data are nothing new.
The Pie Chart is Dead! Long Live the Pie Chart!
What also hasn't changed is the intense debate. Many studies have been run in the meantime, and they generally have found pie charts to be less accurate than bars. There are some valid use cases though, and they continue to be popular despite all the warnings.
When Cole Nussbaumer declares Death to Pie Charts, or Kaiser Fung criticizes a pie chart in his Pi Day posting, they are following an old tradition. Despite all that, though, pie charts do not appear to be going away anytime soon.
Happy Pi(e) Day!