Pick an index, any index.

Over at The Scholarly Kitchen, Philip Davis takes the ARL to task for comparing their serials expenditures with the Consumer Price Index:

By adopting the CPI as a general frame of reference, almost any industry that requires huge professional worker input will look like it is spiraling out of control. Perhaps this is the reason the ARL uses the Consumer Price Index as a reference for journal prices when it could have used the Higher Education Price Index, the Producer Price Index, or an index which more closely resembles professional knowledge production.
The CPI is an excellent tool for collective salary bargaining, for estimating who should be eligible for food stamps or free school lunches. It is a very bad tool for measuring the purchasing power of libraries or justifying a reinvention of the journal publication system.

Since I’ve just played around with updating the famous graph to which Davis takes exception, I thought I’d better take a closer look at the alternative indices he suggests.
From the Commonfund 2008 HEPI Report (pdf; linked from here) I extracted historical HEPI and CPI data from 1976 to 2003, and from the ARL stats interface at U Virginia I extracted the median values for serials expenditures (EXPSER), total salaries expenditures (TOTSAL) and total expenditures (TOTEXP) for the same period (it was limitations in the ARL data range that dictated the time period). I also extracted Producer Price Index data for “all commodities” (PPI ALL) over the same period from the Bureau of Labor Statistics. There are lots of choices for PPI data, but most of them don’t go back as far as 1976. (I did try a couple of industries that I thought required “huge professional worker input”, such as hospitals and book publishers, but the data weren’t available for all the years I wanted — and by eyeball it didn’t look as though they showed much greater increase than the all commodities index.)
Plotting percent cumulative change against time we see:


There isn’t a lot of difference between the HEPI and the CPI, and the all commodities PPI index shows even less increase. Davis suggests that salaries, professional worker input, are at least part of the reason why the CPI is a poor choice for comparison with serials costs, but (to the extent that the HEPI is a better “professional worker weighted” measure) the data do not bear him out. Neither does his claim regarding librarian salaries fit the data I have to hand:

If we plotted academic librarian salaries against the CPI, we could claim that the profession was in crisis, that salary growth was unsustainable, and that the system was simply broken.

It’s clear from the data, though, that library salary expenditures have outstripped the HEPI and CPI, but not by as much as total expenses and not by nearly as much as serials costs.
Remember, too, that this is still only part of the story: “serials” includes a great many publications whose costs have not increased at the same rate as the scholarly literature. The Abridged Index Medicus data I got from EBSCO only cover 1990 onwards, so I reworked the comparison to include the AIM data:


I used the AIM data because comparison with a much larger data set, broken down by individual discipline, showed that the AIM data gave what looks like a reasonable “middle value” — and as you can see, scholarly journal price increases outstrip all others, including total serials, by a considerable margin.
Note also that there’s little difference between “total salaries” and “professional salaries” — the professional salary data series (SALPRF) only goes back to 1986, which is why I’ve included it in this second graph.
None of this is to say that the CPI is the ideal comparison index against which to measure increases in the cost of the scholarly literature. It seems from the comparisons above, though, that there’s not much difference for this particular purpose between the CPI and the HEPI. While I don’t have data for publishing industry salaries, library salaries hew fairly closely to the HEPI and to total library expenditures. It therefore doesn’t seem that salaries have much to do with the much-bruited discrepancy between “general cost of living/doing business/whatever” increases and the rise and rise of the cost of scholarly literature.
If you want the data I used, the spreadsheet is here.

Leave a Reply

Your email address will not be published. Required fields are marked *