RE: A word on calculating costs

[<MARIE.MCVEIGH@thomson.com>]

I read Phil's report and his spreadsheet with great interest - but most of
debate following has centered on his cost-assumptions, quickly leaving
behind the question of the volume of articles produced.  The total cost to
any given institution would be a function of both these things:  total
articles x cost per article.  May I weigh-in with some examination of the
first part of the equation?

If I understand Phil Davis's calculations correctly:  if you use the
figure 20,000 to estimate the size of the journal universe, then, yes, the
8900+ journals indexed by ISI is about 45%.  However, we don't have any
bias towards indexing the most "prolific" journals.  If you look at the
journals we do index from any given publisher and the journals we do not
index from that publisher, there is no difference in the volume of
materials produced per journal.  Given any two journals in a subject
and/or from a region, we prefer the one that is more cited per article,
not the one that produces more articles.

Because ISI's selection is largely independent of the volume of articles
published, I would guess that ISI's coverage is close to being
representative of the population as a whole.  This would make 92% an
overestimate of the proportion of indexed articles compared to all
articles.  If ISI's coverage is, indeed, repesentative, then the
percentage of covered articles compared to all articles would be closer to
45%.

However, 92% might be pretty close to the proportion of Cornell-authored
articles that are indexed by ISI.  I think David Goodman made that point -
that ISI coverage might be quite closely matched to the journals in which
most authors at ARL Universities will choose to publish.  I think Phil's
calculation assumed that the ARL scholars' articles would show a uniform
densite across all available journals (20,000), and the number of articles
in any given journal would depend only on the size of the journal's
production.  However, among the 20,000 journals will be a lot of local,
regional, and specialized journals. Such a journal may publish many
hundreds of articles a year, but will not have published ANYTHING from an
a scholar from Cornell.

How, then, can you estimate the volume of articles published by scholars
at Cornell, based on the number of articles indexed by ISI with "Cornell"
in the address field?  Is it entirely unjustified to use Bradford's Law?  
Assume that 80% of Cornell's "core" production is included in the top 20%
of journals.  If this 20% co-incides with ISI's coverage (and it would be
pretty close in most subjects), then ISI would cover at least 80% of the
published output from Cornell.  This shifts the total cost upward - since
it results in a larger estimate of Cornell's output.

Marie McVeigh