If you've ever wondered about those deals where the large scientific publishers offer bundled discounts to libraries, wonder no more. There's a paper in PNAS whose authors used Freedom of Information Act requests to track down what various university libraries really paid for these deals, and it reveals that everyone paid something different.
Here's a comment in Nature on the study, which they can do with a straight face, since the Nature Publishing Group wasn't included in the study (although the authors seem to think, in retrospect, that they should have done so). These deals are always secret - the publishers make it a requirement not to disclose the terms. And that, as you might easily expect, benefits the publishers, since the library systems don't have a good way of finding out what the market price might be. The PNAS study reveals some odd discrepancies, with some universities getting noticeably better (and worse) deals than others. Wisconsin and Texas bargained hard, it appears, while BYU and Georgia could have done better for themselves.
As the article details, publishers used site licenses to take care of arbitrage opportunities, and the "Big Deal" bundles were used as incentives for the library systems and as tools for the publishers to figure out how much each customer might be willing to pay (using the print-based subscription data as a starting point). As you might have guessed, Elsevier comes out at the top of the pricing list when you just look at the dollar figures. On a cost-per-citation basis, though, they don't look so bad - in fact, they're the most cost-effective of the big publishers by that metric. (Sage and Taylor & Francis both look pretty bad in that table). For reference, the ACS bundle looks pretty decent, and it turns out that nearly 60% of the libraries that deal with the ACS choose the whole package (a high percentage compared to many other publishers). Interestingly, it turns out that some very wealthy schools (Harvard, MIT, Caltech) still don't take the full Elsevier bundle.
And the bundles are, naturally, a mixed bag. It's their whole purpose to be a mixed bag:
It would cost about $3.1 million at 2009 á la carte prices to buy all of the journals in Elsevier’s bundle, the “Freedom Collection.” The average research 1 university paid roughly $1.2 million, or 40% of the summed title-by- title prices, for access to the Freedom Collection. However, this bundle price is by no means equivalent to a 60% discount from journal-by-journal pricing. The Freedom Collection includes about 2,200 journals, many of which are expensive but rarely cited. The least cost-effective 1,100 journals contained in this collection supply fewer than 5% of the citations, but their prices add to more than 25% of the total of á la carte prices. A library that spent $1.2 million on Elsevier journals at listed catalog prices, selecting journals for cost-effectiveness, could obtain access to journals providing 79% of the citations to journals found in the Freedom Collection. Thus, for the average research 1 institution, the citation-scaled discount obtained from the Freedom Collection is about 21%.
Elsevier, though, drops its prices for smaller universities more quickly than many other publishers, and for Master's-level schools it's actually a better deal than many of the nonprofit publishers. We wouldn't know this, though, if these authors hadn't dug up all the info from FOIA requests, and I guess that's the take-home here: scientific publishing is a very opaque, inefficient market. And the publishers like it that way.