journal and publication costs, corrected figures

[David Goodman <dgoodman@princeton.edu>]

I gave some figures for this earlier, but these are considerably more
reliable numbers, -- but see below for notes about limitations of the
analysis.

Cost of serials to libraries:

Columbia: $7.14 million serials expenditure, 5838 articles, $1223 each
Cornell: $6.20 million, 4804 articles, $1290 each
Dartmouth, $3.89 million, 1034 articles, $3720 each
Princeton, $5.45 million, 2692 articles, $2025 each
Stanford, $11.26 million, 5838 articles, $ 1930 each
Yale, $10 million, 4648 articles, $2169 each

Potential savings if articles cost $500:
Columbia: $3.71 million
Cornell, $3.36 million
Dartmouth, $2.42 million
Princeton, $3.31 million
Stanford, $7.33 million
Yale, $6.44 million

If articles cost $1500, Columbia & Cornell would pay more, the others
less.

I did not adust for the cost of submission fes now being paid by the
researchers. The cost given is just the cost to the library.

Number of articles: number of articles published from the university
listed in 2001, according to ISI. Includes Science, social science, and
humanities sections.  Does not include what isn't in themError here may be
large in number of humanities articles etc, because of the spotty coverage
of ISI, but this should be balanced by the low cost of journals in those
fields.

Includes main university 
#NAME?
For Princeton, does include Plasma Physics Lab, for others, will people
who know the local setup please verify. My guess at the likely error here
from my not checking properly is plus or minus 25%, except for Princeton,
which should be correct.
 
o The budgets are the 2000-2001 ARL numbers. (www.arl.org)
o I did not check whether they include the med schools in all cases.
o Whether they include electronic resources is unclear. 
o Whether they include electronic resources other than journals is 
  unclear. 
o The ARL data starts with whatever the library reported for serials, 
  using the 1995 standard, adjusted by me , based on what the library 
  reported for"computer files" and for "other materials"
o There was no category for electronic journals.

I make no claim to have done this adjustment right.  (The new standard
will clarify this in the future.) My guess at the likely error here from
my misinterpretation is plus or minus 25%. My guess at the possible error
is of course larger.

Note:
Division by field:
Columbia:  5126 sci, 540 soc sci, 172 human.
Cornell,   4253 sci, 417 soc sci, 134 human
Dartmouth,  688 sci, 279 soc sci,  67 human
Princeton, 2169 sci, 251 soc sci, 272 human
Stanford,  5126 sci, 540 soc sci, 172 human
Yale,      3857 sci, 522 soc sci, 269 human

These numbers depend a lot on the vagaries of ISI-- and of course is only
journal articles.  There is some very small overlap in psychology between
sci and soc sci, (not considered) otherwise almost none.  I did not
attempt to obtain figures on the percent of serials cost in each subject.

If you use, please understand that any impression of precion is purely
spurious.

Dr. David Goodman
Princeton University Library
and
Palmer School of Library & Information Science, Long Island University
dgoodman@princeton.edu