Probably the most common way to show values on a map is the "colored-region" or "choropleth" method. I'm not the first to notice, however, that this method presents some potentially serious perceptual risks (this was first pointed out to me by Stephen Few):
Other map types such as "colored dot" maps don't pose these same risks:
"Bubble maps" (a.k.a., "proportional symbol maps") also avoid the risks of choropleth maps:
You could also use bars, which would actually allow values to be estimated a bit more precisely than colored dots or bubbles. The downsides of bars, however, (there are always downsides...) are that they can make it tougher to spot patterns (clusters, gaps, etc.) when there are a lot of values, and the chart can look visually busier:
When I posted these examples on LinkedIn, a few people suggested that choropleth maps are O.K. as long as the map is showing "ratio" values, such as "Inhabitants per square kilometer." I think that the problems with choropleth maps would still exist when showing those types of values, though. For example, in a choropleth map of “Inhabitants per square kilometer” for East Asian countries, values for regions like Singapore and Hong Kong would be almost invisible, even though they would be some of the largest values in the whole map.
What’s going on, here?
A fundamental problem with choropleth maps is that they violate one of our most basic intuitions about how charts work, which is that bigger shapes represent bigger quantities. This intuition is why we find treemaps, bubble charts, pie charts, and many other chart types to be so obvious and easy to read. In fact, this intuition is so foundational in our minds that it would seem ridiculous to include an instruction in a bubble chart such as, “Note that larger bubbles represent larger quantities.” Well, yeah. Of course they do.
In a choropleth map, however, larger shapes (i.e., larger regions) don’t represent larger quantities. Even though we consciously know, for example, that small shapes don’t represent small quantities in a choropleth map, they really, really look like small quantities. This creates a “visual conflict” in our minds, forcing us to constantly remind ourselves that, even though shape size is the main thing that we should pay attention to in other chart types, we have to completely ignore it in this chart type and only focus on color, which slows down interpretation and increases the risk of misinterpretation. It’s kind of like a Stroop test.
Another way to look at this is that, in a choropleth map, there’s always a second variable that’s being shown in the chart—whether you want it to be shown or not—which is “Area in square kilometers/miles,” as represented by the size of each shape (i.e., each region). “Area in square kilometers/miles” is completely irrelevant if you're showing something like average household incomes or cancer rates, and yet, in a choropleth map, it’s always there, front and center, drawing the reader’s attention to a variable that has nothing to do with the purpose of the chart.
Now, I’m not saying that choropleth maps are “bad” or “useless”; I’m just pointing out that there are several alternative map types that can communicate the same insights as choropleths without any of the perceptual risks or cognitive gymnastics, so I’m not sure why a chart creator would choose to go with a choropleth after becoming aware of those risks. IMHO, the only time when it would be "safe" to use a choropleth would be when all the regions being shown happen to have roughly the same size and shape, but this is, of course, almost never the case for any area of the world at any scale (Asian countries, South African provinces, Californian counties, etc.).
BTW…
European friends! I'll be in your continent of the woods next month, and the week of Sep23-27 just opened up for me. If you'd like me to deliver my Practical Charts and/or Practical Dashboards courses in person to your team or speak at your event during that week, contact me asap!