Tipping The Scale
©2021, George J. Irwin. All rights reserved.
Want to know how to mislead an audience in just one simple step? Read on...
Back when I was not quite yet in Black Belt Land, my then supervisor showed me two examples of a line graph. This is one of your more simple devices for plotting a series of points over time. The two samples were both illustrating the number of orders shipped through our warehouse. The base data was the same for both charts. However, one showed a nearly straight line that moved up modestly from right to left, and the other one had that line rising at such a sharp rate that it nearly rocketed off the top of the chart.
How is that possible?
If you’re having a bad flashback from Algebra Class, you probably know. It’s not just the points that need to be plotted to create a graph, it’s the determination of the x and y axes, also known as the horizontal line at the bottom of the chart, which usually helps depict the rationale for the sequence of the data points (often but not always a time series) and the vertical line along the left side of the chart, which shows the value of each data point. How many of you still remember which is which by saying “vertical column,” by the way? Or perhaps you’re like me and think of the smash hit “Physical” by Olivia Newton-John and the creative use of the word “horizontally” (look it up, kids).
Anyway, the way in which that y-axis, or the vertical line, is created has a lot to do with how your line of data points looks. Let’s say the lowest value to be graphed is 11 and the highest is 29. If you set the bottom of the y-axis to 10 and the top of the y-axis to be 30, it will look like there is quite a lot of variability in the data, because the points taken together consume almost all of the space between the bottom and top values. The eye fools the brain into thinking that whatever it is you’re measuring is all over the place. But if you choose 0 and 100 as the lowest and highest points of the y-axis, the points seem to “cluster” inside a small segment of the available space, and wow, that process sure looks stable to me. Of course, we have no idea whether it is or not. The “scale” of the y-axis has nothing to do with whether those data points are telling a good story or a bad one, but it sure can create a false impression, no?
This misleading rises to a high art when you don’t hold scales constant for common data sets. The examples my supervisor showed me illustrated that point. The first y-axis scale gave the viewer the idea that the number of orders wasn’t going up that much, because the scale was, well, tipped. That would be the version to show to the Union. The other version had the scale too small, and therefore the line looked like Fabulous News... for Management. Wow, wer’e doing great! Genius, that’s what it is!
Not exactly. I can’t tell you exactly what the optimal scale should have been for this chart, but I can tell you that it should have been consistent over time (i.e. the same y-axis each time that chart was created) and based on something a little more reasonable, if not scientific.