How to Lie with Statistics

Chapter 6 covers problems with other visual interpretations of statistics: the bar chart and the pictorial graph. Huff notes the appeal of representing a number with a picture but warns of the ease with which these images can misrepresent data.

He begins with a section on bar charts. Because they represent comparisons between quantities, any alteration to the chart or shape of the bars can exaggerate differences. Bar charts, like line graphs, can be truncated. Huff also warns against charts that make the bars broader and taller or use three-dimensional shapes, as these can make the differences appear more significant to the eye.

His next section focuses on the pictorial graph. Like the bar chart, it represents the differences between numbers but uses representative pictures instead of simple bars. Huff again warns against images that alter the width and height. A picture might be twice as tall, for example, but its width and volume increase. Huff notes a variety of publications and advertisements that use this method, including differences in wealth within a population or the increase in the steel industry’s production capacity. Huff describes the problem as follows: “To say ‘almost one and one-half’ and to be heard as ‘three’—that’s what the one-dimensional picture can accomplish” (72). For example, the steel industry image reports a 50% increase in production in one decade but adjusts visual proportions to “give a visual impression of […] over 1500 per cent” (71). The artist’s interpretation of the data inflates the perceived difference.

Huff concludes with another caution on the interpretation of pictures in statistics. While the difference in size between items, such as bags of money, may lead the viewer to wrong ideas about the amounts represented, taking this information literally doesn’t necessarily cause issues. However, this difference can become a problem when the size of the figures is not meant to represent reality. Huff uses pictorial graphs of changes in the population of cows and rhinoceroses as examples. The number of milk cows in the US tripled in the period being studied, and an image presents a cow labeled 1936 that is “three times the height” of a cow labeled 1860 to demonstrate this increase (72). However, Huff notes that a reader taking a quick glance at this image might interpret it as indicating cows’ size increase over that period of time, represented by a larger and smaller cow. The reverse is true for the images of rhinoceroses that he presents, which show an enormous animal labeled “1515” above a very small one labeled “1936.” Not only do these two graphs have the issues he previously discussed, but they can also lead to the inaccurate impression that the physical size of the animals is increasing or decreasing over time.