One of the more common forms of chart abuse is messing with the axis, starting from a non-zero point to make a trend look more substantial. But people often become dogmatic about this point, not realizing the point of the rule is to make meaningful differences meaningful (and only make meaningful differences meaningful).
But first, let’s show a somewhat egregious example of y-axis abuse:
The impression this chart gives is that welfare has quadrupled over the period of a couple years. With the axis set to zero, the trend looks a lot less ominous:
It’s easy to scoff and say this is the end of the story (it is, after all, a Fox news graphic we’re debunking here). But the truth is that the redone chart probably hides meaningful differences as well.
So what’s fair? First, the axis here probably shouldn’t be a zero, but if that’s the case we shouldn’t being using a bar graph (bar graphs should always have the vertical axis at zero, line graphs can be more flexible).
As far as what to set the Y-axis at — well is would depend on the historical trend of the thing being measured. We’re looking at a three year trend here — my preference would be to look at what the range of welfare has been as a percent of the population in the past and set the Y-axis to a bit more than that range.
Here’s a graph that gets it right — it takes a given year as an index value, shows the lead-up, and then the increases are off of the indexed year:
One way to think about this graph is this — this is the sort of information you’d look for if you were running a business. If you relied on the Fox News graphs, you’d be out of business in a very short time.
One final example of why the zero-based axis is not always a good idea. Here’s chart that the National Review posted to “prove” that climate change is not a problem (and no, I am not making this up):
That shows a chart, indexed at zero Fahrenheit, of the climate crisis, the growth in global temperature from about 57 degrees Fahrenheit to 60 degrees.
But think about this for a second. What is the meaning of the index here? Fahrenheit’s zero index has nothing to do with global temperature. It’s arbitrary. If you did this in Celsius it would be less pronounced. In Kelvin it would be unreadable. On the other hand, if we had a scale that started at 56 degrees F, the temperature would appear to have tripled.
A better strategy here is, well, basically anything else. But a simple way to do this is the way the los angeles times did this below. Take an average of the last 100 years or so and plot a line at the center, then show the deviations from that line:
Because of the statistical noise I’d probably put in a trend line, but even without it you can see the point — what we are seeing here is a trend with no parallel in recent history.