Evidence is the basis of all science. Evidence should also be the basis of all policies. Concerning scholarly communication, one central piece of evidence is the cost of knowledge, i.e., how much are we spending on our communication system and the associated question of whether we get what we pay for. One point that has always irked me was that I couldn't find any reliable figures as to how much we (i.e., our university libraries) are actually spending on the journals we use. The data from the American Research Libraries is behind a paywall and I'm not even sure this data covers all the university libraries in the US.

The data for the German libraries are not behind a paywall, however. At the German Library Statistics, you can get all the info you want. You can even query their database, very cool! You can check every one of the 250 university libraries and how much they spent on what in which year. Here are a few things I found out:

German libraries spent in 2011

• 170 million € on books
• 130 million € on subscriptions

This amounts to an average of about 660k € in subscription costs for each library (I did not check the distribution to see if I should have calculated the median instead). Given a conservative estimate of publisher profits of around 30%, this suggests that each German library paid about 220k € to publishers' shareholders in 2011. Obviously, this will vary from library to library. For instance, our library here in Regensburg paid about 700k € in 2011 towards publishers' profits.

So far for some solid numbers and statistics. What keeps irritating me is that it appears to be tricky to get that sort of data from countries around the globe, or some global estimate. The estimates for scholarly publishing sales I have seen vary from les than 10b annually to 20b. Assuming the German data were representative of university libraries world-wide (maybe rounded down to $500k per library annually to account for the fact that Germany might be spending more on education than other countries), one can estimate that 9000 universities world-wide together spend about US$4.5b on subscriptions every year (again, using averages, not medians). Again given the conservative estimate of a 30% publishers' profit margin (and all costs remaining constant), a library-based communication system that were to replace scholarly journals would stand to save about $1.5b every year, globally. Clearly, these extrapolations are somewhat dodgy, but they should at least be in the right ballpark. They're a little less than half of what I used to reference, which might be explained by my previous sources probably including book sales.

What one could also see was that an average German library in 2011 subscribed to 2k print journals and 15k e-Journals, at an average cost of 34€ per title. Obviously, with the largest libraries subscribing to more than 60k e-titles, this must mean that this data includes more than just peer-reviewed scholarly journals. Either way, I can use these numbers to check the debate mentioned here before, whether or not per title costs have actually decreased in recent years. For German libraries this is easy to check: I can get the number of periodicals as well as the money spent on them. In 2008, the average subscription price was 48€ per title, so at least these two data points seem to corroborate the publishers' statement that per title costs are decreasing. Obviously, trends and absolute prices will vary from location to location or from field to field. For instance, in 2011, average cost for a printed periodical in the sciences was 540€ and the KIT library annually publishes their most expensive journals, with the most expensive one netting a whopping 20k € per year. Anyway, that's the big picture for German university libraries and for those so inclined, more refined statistics can be generated from the data there.

Obviously, these are not the numbers I would ideally like to rely on, but at least they appear to be more accurate than what I've used before.