Via Andrew Sullivan this morning, I stumbled across this interesting infographic that displays relative gas consumption, both on a per-state and per-capita basis. Like everyone I have been watching the unfolding disaster in the Gulf of Mexico with the sort of horrified fascination usually reserved for car accidents. In the end it’s all about our own addiction to oil– but are some areas of the country more addicted than others? Are the differences class-based, culture-based, or geographically-based? These are some of the things which the chart hints at very nicely. But as usual, there are issues.
Just to get a little Freshman-year-art-school here on the chart, it’s lovely. But why choose those three colors? Are they related somehow? They’re a secondary triad, if you call that orange. Do they posess some signifying characteristics in our culture, like, say, stoplight colors green-yellow-red? No, not really. If they were green-yellow-red then at least there would be a built-in coding of the information along lines we are used to. I saw a similar approach in color-coding food to teach children about nutrition and calorie intake, where foods were classified as green, yellow, or red according to how freely you could eat them without thinking about the consequences. Vegetables were green, fruits were yellow, and carbs, fats, and sugars were red.
It’s useful when doing a chart like this to set up colors which are related but vary along one attribute, either light-to-dark, warm-to-cool, or saturated-to-dull, to better communicated degrees of amount. One side effect of this is usually to make the chart more useful to people who are color blind. I know a couple people with this problem, and it’s made me more aware of the situation. Just for fun, I’ve desaturated the image so you can get a vague idea of how someone with sever color blindness might see it (it’s not an accurate representation by any means but you get the idea):
You can see how, really, all three colors are fairly close in value– that is, they differ by hue but not very much by brightness.
I also wonder about the bar graph along the bottom. When something is made of discrete units, and the units are color-coded in that manner, we ususally expect the color coding to signify units of different types which together comprise a set. But here each circle denotes one barrel of oil, and they are differentiated by a fairly arbitraty quantity of nine barrels– anything below that is green, or “low”, and anything above is coded as “moderate”, with another, more sensible marker at the “average” point of 11, and anything above that marked as “high.” So I might read my own state as having eight barrels of green oil and one barrel of purple oil.
Since the only real metric being displayed here is quantity, a simple bar graph might suffice. Since the boundaries between low, moderate, and high are fairly arbitrary anyway, it might do to make each bar gradiated a bit for emphasis, but to me the dots are just misleading.
I’ve seen a lot of state-by-state graphs lately that use size to illustrate quantity. Some are more successful than others. This one suffers from lack of context. Many of the states appear relatively the same size as in a normal map. If a state uses ten percent more oil than average, should it be only ten percent larger? Or should the deviations from the mean be exaggerated to more easily see them?
One quick fix would be to put an outline-only map of the U.S. at its normal proportions underneath the size-adjusted, filled-in states. This would give us an immediate point of reference not only for the newly-adjusted relative sizes between states, but also for each state’s adjusted size vs. its normal size.