Chartmanship

Stephen Potter, in his famous Gamesmanship books, defined gamesmanship as the art of winning without actually cheating. Likewise, Chartmanship is the art of using graphs to mislead without actually cheating. Here is one of the more innocuous examples from The Times (February 15th, 2001) used to illustrate a story headed Doctors in crisis as complaints soar.

The graphic designer wishes to emphasise the effect required by the story, so he uses three standard techniques:

(a) suppress the zero, so that visual appreciation of the proportional increase is removed.

(b) make the vertical range exactly equal to the range of the data.

(c) use an aspect ratio that increases the apparent slope of the curve.

The really annoying thing about this sort of distortion is that it is so unnecessary. An honest version of the graph illustrates the point being made quite adequately. Here is a version with the zero restored and a square aspect ratio.

If we want to mislead in the other direction, we can further change the aspect ratio and scale

To a trained mathematician, all three graphs are the same, but to the average lay reader they give quite different instant impressions. As we have said, this is a fairly innocuous example, the sort that can be found in newspapers on a daily basis. Graphic designers can really go to town and create a woefully misleading impression, still without cheating. Here is an example from the excellent little book by A K Dewdney 200% of Nothing (which everyone resistant to being conned ought to read) that turns disaster into triumph:

Advanced Chartmanship - A case study

Let us take some raw data that you wish to present to the lay public, say a plot of temperature against time:

Now, it is fairly clear that the graph shows two flattish periods and two rising periods. Let us also suppose, for the sake of argument, that in order to satisfy your paymasters you need to emphasise the importance of the second rise, while diminishing that of the first. There are three techniques that help. The first is to put in a base line at the most helpful level, which is at the beginning of the period you wish to emphasise. The second is to chop off the first plateau, which reinforces the illusion that the second is a natural level. The third is to make use of colour. By perception and tradition the hot colour is red and the cold colour is blue. The result is a form of the same chart, but one which gives a completely different impression to the non-mathematician:

The curves are taken from the web site of the Climate Research Unit at the University of East Anglia.

John Brignell

February 2001

Index