The Southern Policy Law Center has released its “The Year in Hate 2020” report (PDF) cataloguing hate group activity throughout the year. I don’t want to take away from the sixty pages of valuable content, but I think it’s worth taking a critical look at data visualization in any venue. Perhaps a closer look will raise our awareness of the content.
The first real graph is this line chart of the number of hate groups over 22 years. This is valuable context for the current climate and the kind of data you can only get from SPLC. Overall, it’s a fine graph.
For graphical insight purposes, I can pick a few nits. I usually prefer not to label every point, but since the data is not provided elsewhere that I can see, maybe it’s justified here. I also don’t normally care for showing points on the lines, especially when the points are at regular intervals.
Perhaps the most interesting topic of data visualization consideration is the y-axis baseline. As it is, the line starts near the baseline, and I find myself doing a double-take to decide if it’s zero-based or not. For that reason, I like the idea I first saw expressed by Francis Gagnon in the blog post A golden ratio for line charts with truncated y-axis to add a noticeable gap between the data line and the bottom of the graph if the y-axis doesn’t start at zero. Here’s a simple remake with those ideas taken into account (plus the numerous and tiny x-axis labels).
I left the y-axis gap at JMP’s default, even though it’s a little less of a gap than the original blog post suggested. A sharp-eyed reader may notice the line looks a bit different at the end. It looks like the 2020 count of 838 is not drawn at the correct y location in the original chart — it’s higher than 892, for instance. The value 838 is mentioned prominently in the text, so I can only imagine the chart was made with preliminary data and not properly updated for the latest number.
There are a number of maps in the report. This one is about locations of flyer distributions. Others show locations of hate groups.
These do seem to help to illustrate the geographic spread of hate group activities, but they also suffer from the unavoidable issues with maps such as taking population density into account. Looks like there’s a lot of overstriking in the northeast and probably some red dots hiding under the yellow dots in New Jersey. Maybe some hollow dots or transparency would help. I’m also not fond of Mercator projections for data, but it’s doesn’t matter too much for a US map.
What puzzles me about the following line chart is the caption in the lower left saying it “illustrates exponential growth.”
I’m not sure that “terrorgram” is. It’s not mentioned in the prose of the report, which is about hate groups on a platform called Telegram. Presumably, terrorgram refers to those groups or a related set. In any case I don’t see the exponential growth in the lines. Possibly, they meant that the sum is growing exponentially, but the graph doesn’t show the sum, so it doesn’t show the exponential growth, A stacked area chart will help see the sum while keeping some indications of the individual channel contributions.
You can argue that the total y-axis value (which I guess is channel subscribers) is doubling every three months for most of the graph though it still looks more linear near the end. Still worrisome growth, though!
I’ll leave this time series donut chart sequence from page 62 as a thought exercise for the reader.
My study has been based on the PDF version of the report. I notice the web version sometimes has different graphs, such as this nice interactive map of hate groups. Overall, it’s encouraging that the report contains many graphs (I’ve excerpted only a few) that handle data visualization basics well. For instance, all the bar charts have zero baselines and there’s little gratuitous coloring.