Data-Centric Book Review: How Charts Lie
Over the past three years I’ve been studying chart design and working to improve the visualizations I create. This has also made me much more critical and intentional of the charts that I see. If you know me, you know that I can’t resist pointing out a chart that has an inconsistent axis or colors that are so overwhelming that they distract from the actual data.
However, when making a chart, decisions have to be made. How many years of data should I include? What is the most important finding to highlight? What is the appropriate comparison data?
With each choice that is made, even in a good effort to present a clear finding, something goes missing. As Cairo says in his book, “Any chart is a simplification of reality, and it reveals as much as it hides.”
Often, even a clear and concise chart can lead to more questions, which isn’t a bad thing. But it’s worth regularly reviewing the ways in which you might be presenting a limited view and the ways in which the charts you see may be limited.
How Charts Lie: Getting Smarter about Visual Information walks you through the main ways in which charts can be misleading.
Alberto Cairo is well known in the data visualization field and has written several books on the topic. In this book, How Charts Lie, he provides visual examples of ways in which charts can be misleading.
The main ways Cairo claims charts lie are: being poorly designed, displaying dubious data, displaying insufficient data, concealing or confusing uncertainty, and suggesting misleading patterns.
Here are the five most interesting things I learned from the book:
1. 3D Charts are a poor design that can intentionally or unintentionally alter how the data is interpreted. Don’t use them. 2D charts are always clearer.
2. It is more important to use a scale that is appropriate for the data than for all scales to start at zero. A good example of this is average annual global temperatures. If you start the scale at zero it will look like there has barely been any change over time, but we know there has been a substantial increase. Showing a smaller range in the axis allows us to see the small changes that are having a major impact.
3. When talking about global or national data in percentages, it’s easy to forget that you are talking about people. For example, “10.9% of the world’s population” doesn’t sound like a lot, particularly if you are talking about something like poverty. But when you realize that 10.9% of the world is 783 million people, it feels a lot more real. It's ideal to include both percentages and numbers in these cases.
4. With maps, you have to be especially aware of whether numbers or percentages are being represented. Choropleth maps use shades of a color to indicate locations with more (darker) or less (lighter) of a variable present (i.e., people with advanced degrees, people in poverty, people with healthcare). The reason for caution is that often, a map showing the number of people that have some characteristic will look very similar to a population map. It is likely true that there are more people in New York City with advanced degrees compared to other cities, but there are also just more people living there in general. A more accurate view of the data would be to use percentages, which control for population size.
5. It is human nature to rationalize data we see to align with our beliefs. And unfortunately, as Cairo puts it, “the more intelligent we are and the more information we have access to, the more successful our rationalization are.” Florence Nightingale is a model of someone who went beyond accepting praise for her work in the medical field and investigated data with an open mind to understand what else could have been done differently to save more lives.