Tim Gowers has a long post on the cost of Elsevier journals. My contribution to the discussion is to plot data he showed in tables relating Elsevier subscription costs (bundles of journals) to university enrolment.
Whether university enrollment is the right predictor for subscription costs is a good question, but one that should be directed at Gowers, not me! I just plotted the data.
Here are the data for American universities (dashed line in linear regression, dotted lines are 95% confidence intervals, and a couple of outliers labeled). You can click to enlarge:
And here are data for UK universities:
I’m interested in the differences in slope of these two lines.
Update: Due to a request, here’s both countries plotted on the same graph. I converted pounds to American dollars using the current exchange rate on XE.com.
Now we can see that despite huge differences in enrolment, the price being paid by Cambridge and Arizona for the bundle of Elsevier journals are... not that different.
And for the truly geeky, here’s the linear regression for each.
United States: Y = 326185.525 + 33.96366 × X, r = 0.93621, n = 23, p < 0.0001
United Kingdom: Y = 363131.99382 + 36.07912 × X, r = 0.46124, n = 19, p = 0.04684
More updates: And because Richard Van Noorden posted more data on university income in the UK, I get to make another graph:
“Because I was asked” update: Erika Saloman asked for this graph, showing cost as a function of university income per student:
Update, 28 April 2014: The Scholastica blog makes an infographic out of Gowers’s data. It’s not a terrible looking, but I like my graphs better than theirs.