I hate this graph. OK, sure, I don’t hate hate hate hate it: it’s not a 3-d exploding pie chart or anything. It’s not misleading, it’s just extremely difficult to read. Basically, you have to go back and forth between the colors and the labels and the countries and read it like a table. OK, so here’s the table:
Average Hours Per Day Spent in Each Activity Work, Unpaid Eating, Personal Country study work sleeping care Leisure Other France 4 3 11 1 2 2 Germany 4 3 10 1 3 3 Japan 6 2 10 1 2 2 Britain 4 3 10 1 3 3 USA 5 3 10 1 3 2 Turkey 4 3 11 1 3 2
Hmm, that didn’t work too well. Let’s try subtracting the average from each column (for these six countries, the average (unweighted by population) time spent are 4.6 hours on paid work and study, 3.1 hours on unpaid work, 10.2 hours eating and sleeping, etc.):
% Excess Hours Per Day Spent in Each Activity (compared to avg over all countries) Work, Unpaid Eating, Personal Country study work sleeping care Leisure Other France -10% 0% +10% +50% -20% -20% Germany -10% 0% 0% -10% +10% +20% Japan +40% -20% 0% 0% -20% 0% Britain 0% 0% 0% -10% +10% +10% USA 0% 0% 0% -20% +10% 0% Turkey -10% 10% 0% -20% +10% -10%
OK, the Japanese spent more time at work and the French spend more time grooming. Beyond that, I don’t see these numbers as particularly “stereotype confirming” (in Yglesias’s words). But I’m not fully up on my pop culture. What is the stereotype about Turkish people? I have the impression that in Dashiell Hammett’s day they were called “Turks” and the detective was likely to be waylaid by one of them in a dark alley (this counts as “other activities,” I believe), but I’m sure there are some new stereotypes I’m not aware of. Blogging counts as “unpaid work,” right?
Anyway, my first thought was that the above ugly graph should be redone as a line plot, Here’s what I came up with after an hour of work (yeah, yeah, I must have a lot of real work to do if I’m willing to put in this level of procrastination. On the upside, I’m pretty high on the procrastination ladder if I spend an hour on an R script as a way of taking a break!):
Click to see the full-sized version.
I could’ve done this a little better–in particular, the text is hard to read–but it’s basically what I was envisioning. [See P.S. below for something better.] Also, I don’t really know what to make of the ordering of the countries or the ordering of the categories on the x-axis–I just copied what the Economist graph did.
Why do I like my display better? I like it because you can directly compare within a country–to see which activities are done more and which are done less, compared to the average. And you can also compare between countries to see where people spend more time on any particular activity. This between-country comparison would be clearer if we put all the lines on the same graph, but that looks a bit busy to me and I’m happier with the separate line plots. If you had data on a lot of countries I could see batching them (for example, the lines for northern European countries on one plot, the lines for Southern European countries on another, and other plots for English-speaking countries, east Asian countries, south Asian countries, Middle Eastern/North Africa, sub-Saharan Africa, and Latin America).
I can see where the Economist’s graphics designers were coming from with their plots. In any country, the categories add to 24 hours, and the circle plot enforces that constraint. (They could’ve made pie charts but everyone knows how bad that is.) But there are a lot of categories so they needed colors and a legend. And the circle arcs are hard to compare so they needed to put in the exact numbers. The result, though, doesn’t work for me. I mean, sure, maybe it was fine–Matthew Yglesias is more in the target audience of the Economist than I am, and he liked the graph–but I think it could’ve been much better. And I’m sure that if a graphics designer worked with me on it, the graph could be better still.
At some point this would represent a bit too much effort spent on one particular graph in a weekly newspaper. But if we have enough good examples of these, they could represent a template that could be used all over.
P.S. I was dissatisfied with my graph above because of lack of readability of the labels. So I spent another
45 minuteshour to make this:
Wow! All the information, it’s clear and readable, and I got it in under 600 x 250 resolution on a png. I like it.
P.P.S. Here’s the R code I used to make the graphs.
P.P.P.S. See here for yet another version.