7.2 Bertin’s many graphics of the French workforce data

There are just over \(100\) graphics in the chapter, all giving views of the same dataset. The exact number depends on how sets of multiple graphics and intermediate graphics illustrating how graphics were constructed are counted. About two-thirds of the graphics are maps of France, the rest are barcharts, scatterplots, histograms, and other graphic forms. Many of the graphics are drawn to show that they do not work well, as these remarks in the book show:

• In no case does the image yield useful information. [Referring to Figures 1, 2, and 3] (p. 103)
  
• It is difficult to draw a useful conclusion from this group of images. (p. 105)
  
• Note that the resemblances among the sectors are hardly visible and, in fact, are overwhelmed by the striking differences in total working population per department. (p. 121)
  
• In no case, however, can the notion of quantitative value be obtained from these images. (p. 124)
  
• This solution reduces considerably the information transmitted and opens the door to unjustifiable interpretations. (p. 126)
  
• Different value steps, applied to absolute quantities, result in completely erroneous images. (p. 130)

Comments in a similar vein can be found on pages 109, 119, 122, 131, and 135. The first edition of Bertin’s book was published in French in 1967. Perhaps he wanted to emphasise that not all of many different graphics are equally good. Judging by some of what is published today, this advice is still relevant.

Interestingly, the main (and pretty well only) conclusion that Bertin draws from all his graphics is that the data for two of the sectors are related. This conclusion is drawn by him from each of the groups of displays on pages 106, 107, 108, 109, 111, 112, 113, 115, 129, 135, and 137. He likes the final four maps on p. 137 best and writes: “The correlation between sectors II [Industry], III [Commerce] and the total population suggested by the earlier diagrams, is particularly striking here.” Bertin is referring to the total working population, the sum of the data for the three sectors. Since the two biggest sectors are highly positively correlated, it is obvious that the total will be positively correlated with them as well. Curiously this is not mentioned.