July 29th, 2021
Visualizing survey results, explained
On Tuesday, my co-worker Gregor and I published an article that was a bit more provocative than what we normally write. In it, we argue against diverging bar charts, that are often used to visualize survey results. We recommend 100% stacked bar charts instead. So what could be more appropriate than to show some survey results in our chart this week? And yes, you guessed right, as a 100% stacked bar chart.
The data comes from YouGov, a British market research firm that asks Internet users for their opinion about…anything. Like, what counts as “fake news”? (Turns out, people have quite different opinions about that.) Or: how young are “young people”? (It’s beautiful to see that you can only deny the truth for so long.) Shortly before christmas last year, they asked people how much they like certain chocolates. Here are the results:
Most people seem to think that chocolate is a great idea in general, so we only see a few Dislikes in this chart – except when it comes to Bounty and Snickers. 🥥 and 🥜 seem to be controversial ingredients: If you offer your friends a Bounty or Snickers, it’s more likely that they be like “OMG YES PLEASE” than when you offer them Twix, Mars or Milky Way. But it’s also more likely that they will react with utter disgust than when you offer them any other candy from the Celebrations selection. It’s a risky gift.
The most important choice when showing survey results with 100% stacked bar charts is about colors. We want to tone down the Neutrals and the Don’t knows, and in most cases, we want to make the extrema clearly stand out.
Fun fact: If you do that and if you have an uneven number of categories (five, in our case), Datawrapper will recognize that you want to visualize survey results. It will center-align the labels of the middle category (Neutrals) and right-align the labels of the two categories to the right. This helps to communicate that the chart is showing two opposed positions.
This feature exists since last week, together with two other new ones:
First, you will find a “Stack percentages (normalize values to 100%)” checkbox in the “Refine” tab. That means you can upload absolute values (“350 people liked Bounty, 150 disliked it”) and then convert it to percentages directly in Datawrapper (“70% liked Bounty, 30% disliked it”). Hover over the chart up there and click “Edit this chart” and then go to the “Check & Describe” step to see what I mean.
And second: Grid lines! Turning them on makes it easier for readers to see, for example, that 75% of the survey respondents liked or very much liked Twix. Or that a quarter disliked or very much disliked Bounty.
As always, we’re happy to hear how you like the changes in our tool. Get in touch with me at firstname.lastname@example.org. And see you again next Thursday.