Line charts for distributions, not time

The best of Datawrapper charts 2018, part 1

The last month of the year has started, which gives us a great reason to look back at what happened between the beginning of 2018 and…now. Turns out, lots of our users have built great charts and maps! I selected some of them and got in touch with their creators to ask them about their work. In the following weeks, I will share with you some of the (what I consider) most innovative Datawrapper charts I’ve seen this year.

Today we start with two line charts. Doesn’t sound so innovative, right? But in the following two charts, each line shows a distribution instead of a development over time:

The first one is by data journalist Elena Erdmann from the German newsroom ZEIT Online. Besides her job, she’s part of the great initiative Journocode, which tries to “close the gap between journalism and data science”. Have a look at their website & advent calendar, if you don’t know them yet. You can find Elena’s chart in its original article here.

Elena’s chart shows the age of German mothers when they gave birth to their kids. Every curve shows one year between 1972 (dark green) and 2016 (red). We can read, for example, that for every thousand of 22-year-old women, 122 kids were born in 1972. 44 years later in 2016, only 32 kids were born to a thousand of 22-year-old women. The peak of the curve moved. German mothers give birth later than ten, twenty or forty years ago.

You might notice that in Elena’s chart, the area below some curves is greater than below others. That’s because she uses absolute numbers: In some years, fewer babies were born than in others. Let’s look at a distribution chart that shows shares – there, the area below each curve is the same:

This second chart is by Pramit Bhattacharya, the head of the data journalism unit “Plain Facts” at New Delhi-based Mint. You can find the chart above (and four other ones!) in the original article here, a collaboration with Pramit’s former colleague Tadit Kundu.

Here we see the power of relative distribution charts. Not only do we learn that Indian women have their first intercourse earlier than men.[1]  We also see that uneducated Indians experience their first sex in a smaller age window: The peaks of their distribution curves are higher than the peaks of the distribution curves that represent educated people.

Both Elena and Pramit use distribution to draw a more nuanced picture of reality than a simple average number could possibly achieve. We don’t just see the most common behavior, but also the outliers. Why is that important? Because many of us are outliers. Elena and Pramit communicate to us readers that there is not that one age when it’s “normal” to have sex or to have a child for the first time in our societies, but lots of options – and there are people who share our non-average experiences.

Next week, we will look at dozens of dots. See you then!

  1. Tadit and Pramit explain the reason for these distributions in their article: “People with higher levels of education tend to stay in college longer, and hence get married later. The age at first intercourse for such people therefore tends to be later.”↩︎