Chart distortion
Drawing a chart so that the picture says something the numbers do not: an axis that starts near the top, a pictogram that grows in area when the quantity grows in length, two series on scales chosen to cross where the argument needs them to.
A chart is read as a shape before it is read as data, and the shape is the designer's. Cut the vertical axis at ninety and a two-point rise becomes a cliff; double a pictogram's height to show a doubling and its area has quadrupled; leave the axis unlabeled and the reader supplies one. The numbers printed on the chart may all be correct, which is the device's defense; the lie is in the geometry, and geometry is not checked.
Darrell Huff named the gee-whiz graph in 1954 and Edward Tufte put a number on it, the lie factor, the ratio of the effect shown to the effect in the data. Exploits Anchoring and Framing Effect.
Historical examples
Huff's How to Lie with Statistics showed the same government payroll figures twice, once on an axis from zero, where the line was nearly flat, and once on an axis cut at twenty, where the same rise filled the chart. He called the second a "gee-whiz graph," and the book, still in print, sold more copies than any statistics text written since.
Huff, Darrell. How to Lie with Statistics. New York: W. W. Norton, 1954, ch. 5.
At a House hearing on Planned Parenthood, Representative Jason Chaffetz displayed a chart with two lines, cancer screenings falling and abortions rising, drawn to cross dramatically. The chart had no vertical axis: the screening figure was about 935,000 and the abortion figure about 327,000, so on a shared scale the lines would not have met. The source was an advocacy group, and fact-checkers rated the chart misleading.
PolitiFact, "Chart shown at Planned Parenthood hearing is misleading and 'ethically wrong'," 1 October 2015; Vox, "This chart is a crime against data," 29 September 2015.
A Times graphic on new fuel-economy standards drew the required mileage, rising from eighteen to twenty-seven and a half, as a road widening across the page, and the drawn increase was 783 percent against a real increase of 53 percent. Tufte used it to define the lie factor: the ratio of the effect shown to the effect in the data, here 14.8.
Tufte, Edward R. The Visual Display of Quantitative Information. Cheshire, CT: Graphics Press, 1983, 57–58.
Find a chart in a news story or advertisement. Does the vertical axis start at zero? Redraw it so that it does, and describe what changes.
In the wild
Short examples of the technique at work, each with what to notice. The same examples are questions in the quiz, where the answer is not given away.
A bar chart shows this year's sales as a bar twice the height of last year's; the numbers beneath are 4.1 and 4.3 million.
What to notice. The axis has been cut so that a five percent rise doubles the bar. The numbers are right and the picture is not.
A pictogram shows the national debt as a money bag, drawn three times taller and three times wider than the bag for ten years ago, to show it has tripled.
What to notice. Height and width both tripled, so the area is nine times larger. The eye reads the area.
A line chart of crime has no numbers on the vertical axis and ends at the year before the rate fell.
What to notice. The missing axis and the chosen endpoint are both in the drawing; the data would say something else.
Sources
- Huff, Darrell. How to Lie with Statistics. New York: W. W. Norton, 1954. ISBN 978-0393310726
- Tufte, Edward R. The Visual Display of Quantitative Information. 2nd ed. Cheshire, CT: Graphics Press, 2001. ISBN 978-1930824133
- Cairo, Alberto. How Charts Lie: Getting Smarter about Visual Information. New York: W. W. Norton, 2019. ISBN 978-1324001560