Andrew, thanks, I'd forgotten about your article; it does have some useful numbers (though 8 years old - in particular, Phys Rev B's numbers have changed somewhat, and the profit you mentioned was quite atypical for us...). However, on the issue of publication expenses vs R&D expenditures - there's a third variable which I don't see in the ARL or NSF numbers - the total number of institutions world-wide that are participating in research. I believe this number has been growing - is there any statistical collection of total expenditures over all institutions, rather than the sort of per-institution data that ARL has?
Andrew Odlyzko wrote:
University libraries have already lost some of their importance. [...] My opinion is that this is unrealistic, and that the decline in the relative share of resources devoted to libraries resulted from their decreasing importance. The increasing availability of phone, fax, email, interlibrary loan, and other methods of obtaining information, and the inability of any single library to satisfy scholars' needs, may mean that scholars do not need the library as much, and as a result do not fight for it.
The main focus of your "tragic loss" article was the obsolescence of paper, and the resulting consequences. One consequence which was perhaps not widely anticipated is expanded access to research journal content - now available from the desktop instead of having to go to the library. And the increased availability that consortium deals and other special arrangements are providing. So the library as a physical facility is less useful, but as a provider of information, surely the utility of every library has grown over the past 8 years? Are the other things you mention (phone, fax, email, etc.) really a substitute for traditional scholarly communication? The total number of institutions (N) comes into the equation assuming some number of research dollars per institution (R). The number of articles published (P) varies as total research funding (P = c * R * N for some constant c) so for every institution to have access to every article in the paper world meant spending something proportional to P: per-institution spending S is then S = c2 * R * N and total spending is S*N = c2 * R * N^2 i.e. if research spending per institution is level (or only growing with inflation) per-institution spending in the print era still had to increase because total worldwide research spending was growing - and total publication spending was increasing as the square of the number of research institutions world-wide. Obviously, libraries could not keep up, hence the crisis. But what the electronic era gives us is a gift - having access to electronic articles does not mean actually having a physical copy: the only physical copies an institution has to have are those specifically downloaded by its researchers, which will grow only as R (dollars spent per institution), not P (total articles published). The most intensive electronic activity per institution would be searching, which grows only logarithmically with P for properly indexed searches. So there is no longer anything that forces per-institution spending to vary as P - it's possible to drop below the R*N total research spending curve now, and still have access to everything. In the long run, when paper is really gone, library spending should be only some constant fraction of R (S = c3 * R) and total spending becomes S*N = c3 * R * N, growing only linearly with total worldwide research spending. I believe this shift from N^2 to N growth in total publication spending is what we are in the middle of. As the transition takes effect, it's going to mean a huge improvement in accessibility for the foreseeable future. Arthur