83  Pie Charts — Type 1

Different Observations, Same Variables

Figure 81.1 tells us why pie charts and heat maps, are not optimal for visualization of quantitative information. These types of plots both rely on encoding elements that are poorly decoded by the viewer. Here I’ll consider the two types of pie charts, when they are used and some great alternatives.

The first type of pie chart we’ll encounter is what I call the different observations, same variables variant. An example is given in ?fig-Elections2017MakebarChartSplit.

In the 2017 federal elections in Germany, 16 bundesländer (states) took part in voting for 5 well-established parties (the incumband CDU and the Bavarian sister party CSU, SPD, Linke, Grüne and FDP), one young party (AfD) and a smattering of other regional parties. Here different observations (i.e. voters) of the same variables (i.e. political parties) are plotted. The pie slices are colored according to the party colors, making interpretation at least somewhat intuitive for the informed viewer. The pie charts are arranged according to CDU/CSU results, instead of alphabetically, allowing us to include another layer of information.

Figure 83.1: 16 pie charts, each consisting of 6 differnt colors is overwhelming for the reader. It’s also difficult to make comparisons between distant sub-plots and with slices that are in different orientations.

The classic solution to a collection of different observations, same variables pie charts is a stack of horizontal, proportional bar charts, Figure 83.2). Here, many of the inefficiencies of decoding a pie chart are alleviated. Each chart begins and ends at the same point and the only difficulty is comparing segments in the middle, which begin and end at differnt points on the scale (i.e. Common, but unaligned scale). It’s a deficiency we can live with since it does improve readability considerably.

Instead of ordering the bars according to CDU/CSU results, we’ll use AfD results. The astonishing results of the young, right-wing, AfD party was mostly unexpected and Figure 83.2 allows us to see that, for example, they achieved approximately 25% of the vote in Sachsen.

Figure 83.2: 16 stacked horizontal proportional bar charts contains a lot of information in an easy to read format.

Anoter level of information that is easy to convey here is the distinction between the alte, formerly West German, and the neue, formerly East German, Bundesländer, depicted in Figure 83.3. This allows us to see a clear distinction between the “old” and “new” states. Reunification didn’t occur that long ago and it appears that distinctions between the former German states still persist.

Figure 83.3: It’s easier to introduce another variable via small multiples when our plots are already well organized.

We’ve already move quit a bit away from pie charts, which means we can start thinking about introducing other interesting variables. For example, pie charts are really terrible for looking at a change over time. Allowing for a few exceptions, time is typically mapped onto the x axis and drawn with a line. Figure 83.4 allows us to compare the 2017 election results with those from 2013. Here, I’ve oped to arrange the individual sub-plots in alphabetical order.

Figure 83.4: time setup.