Data, data, data - on journal prices!

I was forwarded info about the latest release of the Journal Cost-Effectiveness project, put together by  Ted Bergstrom and Preston McAfee.  They are researchers who have become quite heavily involved in the fight against the practices of overcharging by commercial publishers.  Because of their emphasis on the continued rising prices of subscriptions on top of very large profits, their data is divided by for- and non-profit entities.  Their methodology is well-described and transparent, and they provide the complete set of data for all years (2009-2011) for downloading.

The data set includes price (2010 prices) per article, price per citation (citations in 2009 papers to articles published 2004-2008), composite price index ("geometric mean of the Price Per Article and the Price Per Citation"), a relative price index (relative to mean of non-profit subscription price within same subject), and a ranked "value" based on the relative price index (<1.25 is "good", 1.25-2.00 is "medium", >=2.0 is "bad").

I used SPSS to see the distribution of the measures by profit status (for- or non-profit):

Doing a t-test of the differences in the mean showed that the differences were, of course, all statistically significant.  But that doesn't do justice to the extreme differences between these two groups.  Also important is the amount of variation within each group.  This time I looked only at mean price per article for each of the 3 "value" categories.  Here's an error chart:

This demonstrates not only the extreme differences between the price per article for non- and for-profit publishers, but also the differences in variation.

Now, same data, only by value and clustered by profit status:

Notice that the variation in the "good" value category is the least (very tiny bars), is somewhat larger in the "medium" category, but much larger in the "bad" category.

This is a look at just one aspect of one measure of this very nice data set.  I'm going to have fun looking at the other aspects of this and the other measures.  But what I'd really like to do is merge this with our own list of journals and see the value we're getting.  Of course, the prices are based on the published subscription prices for libraries (for tiered prices, they used that for a large academic library), and may not match ours.  I am disappointed that they did not include all of the data, including prices, number of articles, and number of citations.  Maybe they'll provide it if I ask...