Graphicacy and how charts can lie

I have been enjoying reading Alberto Cairo’s excellent 2019 book How Charts Lie, which is subtitled Getting Smarter About Visual Information. It is all about graphicacy, the analog of literacy, which the Merriam-Webster dictionary defines as:

“the ability to understand, use, or generate graphic images (such as maps and diagrams).”

Mr. Cairo’s book has the following chapter titles:

1]     How Charts Work

2]     Charts That Lie by Being Poorly Designed

3]     Charts That Lie by Displaying Dubious Data

4]     Charts That Lie by Displaying Insufficient Data

5]     Charts That Lie by Concealing or Confusing Uncertainty

6]    Charts That Lie by Suggesting Misleading Patterns

(7)] Conclusion  Don’t Lie to Yourself (or to Others) with Charts

You can download the graphics from that book here. You can read the text here at Google Books. Alberto’s blog has a February 14, 2020 post on ‘How Charts Lie’: a few edits to the first print edition.

In Chapter 2 (Charts That Lie by Being Poorly Designed), his first example on page 54 as shown above, is what at first glance appears to be a plot having time as the horizontal scale with two lines on a common vertical scale one increasing and one decreasing. Instead it is nefarious nonsense. When you glance at the number in the upper right (328,000) you see it is only 35% of the one at the lower right (935,573). On October 1, 2015 at Politifact Linda Qiu described how a Chart shown at Planned Parenthood hearing is misleading and ‘ethically wrong.’ Cancer Screening was cherry picked as a topic that decreased. The largest category (which also increased) really was STI/STD Testing and Treatment.

  

In Chapter 3 (Charts That Lie by Displaying Dubious Data) on page 99 he uses the example shown above, where the red state of Kansas anomalously has the highest page views per person on Pornhub. Then he explains that Kansans likely really aren’t naughty. Northern Kansas is the geographical center for the U.S., and people whose location cannot be determined are assigned there by default. On February 10, 2019 I had blogged about another example - A misleading bar chart with inflated fear percentages. SlideHeroes had misread tables in a survey and wrongly used a cumulative percent that included people who had refused to answer (Didn’t Know). 

In Chapter 4 on page 121 (Charts That Lie by Displaying Insufficient Data) he uses a hypothetical example about smallpox and vaccination that I replotted as shown above. More people died because of the vaccine than did from smallpox. But that bar chart omits mentioning that actually 99 times as many people were vaccinated than not vaccinated. We just are shown the numerator for what should have been expressed as a fraction dividing by a very different denominator.

As shown above, I have replotted the graphic from page 122 to display calculation of the probability for dying. The percentage if you take the vaccine (0.01%) is forty times smaller than that (0.4%) if you do not. But the calculation calls for a lot of information - it involves knowing six percentages.   

As shown above, in my plot of his chart from page 123, we also can compare the number who died based on the figures for a mostly vaccinated population (139) with the disaster which would have happened had there instead been no vaccination (4000).   

In Chapter 6 (Charts That Lie by Suggesting Misleading Patterns) on page 168 Mr. Cairo first has a chart showing how employment rose after January 2017, when Donald Trump was inaugurated. Then he has another chart revealing that rise simply is a continuation of the same long rise going back to 2011 during the Obama administration.  

Back on January 28, 2011 I blogged about how Michelle Bachmann pinned my bogometer! In her Tea Party Express reply to Obama’s State of the Union address she showed a bar chart of the unemployment rate which suggested a jump under Obama. But there was a smooth curve going from Bush to Obama.