All piecharts are 'wrong' but some are useful
When it comes to piecharts, I belong to the team-don't. But every rule has an exception. And this post is about one such exception. In rare and specific instances, piecharts can be useful. Below I present such an instance.
This is a bit of an involved figure. The dataset, sourced from the ACLED Project, comes from my recently accepted paper in the AJAE. It includes the total number of incidents of violence during the 1997-2020 period. There are four different types of perpetrators, also known as conflict actors. These are State forces, Rebel groups, Political militias, and Identity militias. The aim is to illustrate their relative geographic prevalence.
Presenting everything on the same map can be tricky. But piecharts save the day. By design, a piechart is a weighted mix of colors assigned to each of its elements. So, I assign four distinct colors to the four conflict actors. In this instance, we don't care much about the exact numbers that go into piecharts. All we care about is whether any given color dominates the other colors.
And it works like a charm. We can see, for example, the prevalence of violent acts staged by political militias in Southern Africa and Madagascar. As well as those staged by rebel groups in peripheral parts of Nigeria and DRC, and the parts of Somalia surrounding its capital city of Mogadishu.
Replication material is available here.