Re: Graphing the Bergstrom and McAfee Journal Pricing Data

[Phil Davis <pmd8@cornell.edu>]

Jan Veltrop wrote [snip]:

"My suggestion would be to make tables and graphs per discipline and 
taking into account journal types in order to have more material for 
meaningful comparisons."

Phil responds:

Bergstrom and McAfee DO factor in the price differences between subject 
fields, and this is reflected in their Relative Cost Index

"Relative Cost Index: The relative cost index (RCI) is the Composite Price 
Index (CPI) divided by the median CPI of non-profit journals in the same 
subject category. Journals that have multiple subject listings are 
factored into the average CPI for each field it belongs to, and its RPI is 
its CPI divided by the average of the average CPIs for each field."

In addition, they DO create subject calculations, as reflected in their 
summary statistics: http://www.hss.caltech.edu/~mcafee/Journal/Summary.pdf

While Jan's suggests breaking the journal field down into smaller analysis 
groups in order to control for subject, publisher, journal type, page 
length, citations, usage, et cetera, we are left with categories of one 
journal.  In other words, we digress to argue that each journal is unique 
and therefore cannot be compared to any other journal.  This is not very 
helpful.  One only needs to review over twenty years of pricing research 
that analyzes journals -- in some cases, down to price per character -- to 
understand that the system is composed of massive pricing inequalities.

I'll respond to Jan's question of 'Value' in a separate post.

--Phil Davis

At 05:39 PM 12/5/2005, Jan Velterop wrote in full:
> Some shortcomings of the otherwise interesting data and graphs are that
> the data do not (not easily anyway) separate different journal types, such
> as review journals, research-journals, magazines with scientific content
> (such as e.g. Nature and Science), and mixtures of the above, and the
> graphs do not seem to take into account the very different portfolios of
> the various publishers in terms of disciplines covered. 'Value' is a
> relative concept. Average article length, for instance, and things like
> citation behaviour and habits with regard to downloads/usage from the
> publisher's site (take Physics as an example), or even download/usage
> habits per se (the value of publishing an article is not just in its
> usage, and sometimes almost not at all in its usage) vary considerably in
> different disciplines with its consequences on price per article and on
> just about every other variable in the dataset. For example, a portfolio
> consisting mainly of medical review journals and one consisting of mainly
> mathematics research journals will have extremely different
> characteristics, which are reflected in prices, citations, usage, et
> cetera, rendering direct comparisons between the two pretty useless.
>
> My suggestion would be to make tables and graphs per discipline and taking
> into account journal types in order to have more material for meaningful
> comparisons.
>
> Jan Velterop