I found that part really interesting since the requirement for a zero baseline today is limited to bar charts (see Few) whereas this insists you use it for line charts as well (which most modern data viz textbooks will say is okay because you judge the length of a bar chart but the position of a point in a time series).
This senseless insistence on beginning the vertical axis at zero is repeated in Huff's popular book, How to Lie with Statistics, where he also claims that failing to adopt this rule results in "misleading" graphs.
It depends on whether you want to show relative or absolute change. Huff's warning about scales not starting at zero occurs when the chart has a caption such as "A massive five percent growth!"
In my experience, not starting a scale at zero often DOES produce a misleading graph.
