The use of charts and data visualisation is becoming an increasingly important part of how we report data. An eye-catching chart will always attract much more attention than a lengthy report on a webpage, no matter how interesting the findings. With people following hundreds of profiles online, and only having the time to quickly flick through a report, we are spending just seconds viewing content before moving on. The need to grab people’s attention quickly is all the more pressing, and charts are a great way to do this.
But there is never going to be a completely neutral way to visualise your data. What data is included, what axis scales are used, and even the type of chart are all decisions that can affect what story the chart tells. The stats may be correct but through a series of design choices a graph can, intentionally or otherwise, distort the data to the extent that it is misleading.
We’ve all heard about Fake News, but should we be equally wary of Fake Charts? And what aspects should we be cautious of, so we can protect ourselves from being misled?
One way that charts can be misleading is by using dual axes. This is the use of two Y-axes with different scales, often used to show correlations in the data when the range is too large for just the one.
I recently came across a great example of dual axes being used to give the completely wrong impression. Created by anti-abortion organisation Americans United for Life and presented during a congressional hearing investigating Planned Parenthood, it appears to show a huge increase in abortion procedures at the same time as a reduction in cancer screenings and prevention services delivered by the organisation.
Unsurprisingly, many people online pointed out that the use of dual Y axes without showing the scale for either, as well as a time series that displayed a completely straight line instead of a realistic jagged change between the years, made for a very misleading chart. This is what it should have looked like:
(Data from Politifact)
By using a single axis, the chart shows that while abortion services have increased, it is by a tiny proportion compared to the decrease in cancer screenings, and the two trends do not overlap.
Largely, dual axes are used when the data for the two variables are on vastly different scales. If the data is similar enough that the chart could have been plotted using just one axis, it is worth checking the chart more carefully.
Truncated axes, when used wisely, can make a chart clearer by decreasing the amount of unused white space, and draw attention to smaller yet significant changes in the data. However, when used inappropriately, they can also exaggerate those small changes by giving the impression that they are more drastic than they actually are. The pair of charts below uses the same data, yet at first glance show very different stories.
On this first chart, the truncated axis gives the impression of a series of erratic spikes with a really pronounced dip around 2001. But the second chart, using a full axis, shows the trend has actually been more stable, only varying by 14 percentage points over the 30 year time series.
Bar and column charts should always have zeroed axes, but for line charts consider whether the scale used is relevant to the subject: graphs showing temperature changes over summer are a good example of where a truncated axis makes sense, while the above is not.
More potential to mislead comes in the form of missing Y axes, especially when combined with a truncated scale. You may have spotted some of these in election leaflets coming through your letterbox.
The lack of Y axis in Figure 2A and using a small font size for the value labels give the impression that the number of children taking part in reading activities from higher income families is at least triple that of those from the lowest income quintile, when the difference is less than double (27 percentage points).
For this type of chart, take notice of whether the size of the bars is relative to the values shown. If the proportions don’t seem realistic, they probably aren’t.
Next time a chart catches your eye, make sure you think about what you’re looking at, and whether you can trust the source and chart design. In the same way you don’t believe every news story you see online, don’t always take your charts at face value.
Follow me on Twitter: @Emma_Bowey