Costs of Resources in the Scientific Scholarly Journal System

Libraries are experiencing an untenable situation because spiraling prices are causing them spend more for fewer journals. While that fact dominates discussions among publishers, librarians, and scientists, it is useful to examine how many resources (e.g., labor, equipment, space, supplies) are expended in the overall journal system involving authors, publishers, secondary services, libraries, and readers. The systems approach ignores cash exchanged and focuses instead on the total resources expended in the journal system, normalized by the number of scientists. That approach represents the true cost of the journal system to society or the scientific community.

The amount of resources used to write articles is increasing, although not appreciably. Total publishing resources have risen moderately, while library journal-processing resources have fallen because fewer journals are acquired. Resources used to obtain separate copies of articles have risen, however, so that the overall cost of resources, again ignoring exchange of cash, has fallen slightly. The amount of resources applied to secondary services is thought to have risen slightly, but that has not been confirmed, and the readers' time expended has increased appreciably because of the additional time that must be spent traveling to libraries to obtain articles. Thus, the overall system resources expended per scientist and per use has increased, although not appreciably. The biggest change is that the resource expenditures are shifting among the participants.

Another systems view takes into account money exchanged among participants and the effects of price-and-demand relationships. Pricing policies have led to publishers losing subscribers, higher subscription budgets for fewer journals in libraries, readers paying more in their time per reading, and library and reader funding sources receiving less for their expenditures.

Two Reasons Scholarly Journal Prices Are So High

Scholarly journal prices have risen significantly for two compelling reasons: Scholarly journals are characterized by (1) very high fixed costs and (2) a relatively low, and decreasing, number of subscriptions to cover those costs. That is significant because fewer subscriptions require higher prices to recover costs. In fact, the 1995 median number of journal subscribers is approximately 1,900 subscribers, down from 2,900 in 1975. Unfortunately, as unit costs per subscriber decrease with an increasing number of subscribers, the unit cost per subscription does not begin to level out (at about $100) until at least five thousand subscribers. Below that amount, typical costs range from $840 per subscription at five hundred subscribers to $440 at one thousand subscribers and $200 at 2,500 subscribers. The problem is exacerbated by the fact that subscription prices have increased at a rate far higher than inflation, causing a reduction in the number of subscribers, which, in turn, forces the publishers to charge higher subscription prices.

Publishing Costs of Scholarly Journals

We have derived a publishing-cost model for print journals (King and Tenopir 1998) consisting of cost parameters (e.g., number of pages, issues, and subscriptions) and cost variables (e.g., editing cost per page). The cost parameters for scientific scholarly journals (1960-1995) were observed from a sample of journals; cost variables were estimated for the years 1975 and 1995 from data in the literature. The cost model for paper journals consists of five components:

1. first-copy article processing
2. non-article processing (covers, tables of contents, editorials, letters, and book reviews)
3. production (printing and binding)
4. distribution (wrapping, mailing, or electronic), and support (administration, promotion, and finance)

The model is not described here other than to present typical total costs using average cost parameters and cost variables.

For most publishers of scientific scholarly journals, processing and support — fixed costs, in most cases — are the largest part of total publishing costs. Those fixed costs of a typical journal total about $400,000, while production and distribution costs average $40 per subscriber. A publisher with 500 subscribers must charge $840 per subscription in order to recover costs, but need charge only $80 per subscription to recover the costs of reaching ten thousand subscribers. In that way, the price approaches an asymptote at the production and distribution cost of $40 per subscription. That model helps explain why some journals must charge a high price, while others can charge much less.

Scientific journals tend to address disciplines that range in size from hundreds to hundreds of thousands of practitioners. That wide range of potential subscribers means a wide range of prices. Incorrect pricing can lead to extremely large losses, because the fixed costs are so large. For example, if a journal with the typical costs outlined above charges $200 a subscription, but has only one thousand subscribers, the publisher would lose $240,000 a year. In fact, in all of science the price and demand relationships are such that the downside risks tend to far outweigh the upside gains.

Price and Demand Sensitivities

Since 1960, journal prices have generally increased at a rate faster than inflation (King, McDonald, and Roderer 1981; Tenopir and King1997). However, the rate of increase was greatest from 1975 to 1995. The average subscription price of $39 in 1975 rose to $284 in 1995 — an increase factor of 7.3 (or 2.6 in constant dollars). The increase is in large part attributable to increase in journal size, number of issues published, and other cost parameters (Tenopir and King 1997). However, size increase and inflation account for only half of the price increases. While the cost of publishing resources such as paper, labor, and capitalization of equipment also explain some of the increases, they do not represent nearly the increase in costs that is often stated.

The accelerated price rises began in the late 1970s, triggered by inflation, fluctuating foreign exchange rates, and other factors. As a result, personal subscriptions began to decline because of their high sensitivity to price changes. In fact, they declined from 5.8 subscriptions per scientist in 1977 to approximately 2.9 in recent years. That, in turn, caused a precipitous decline in revenues. To cover costs and recover revenue, publishers increased subscription prices to libraries, where the demand is much less sensitive to price changes.

As personal subscriptions declined, readers shifted to libraries as the source for much of their reading. In universities in 1977, 25 percent of readings were from library-provided journals; that increased to 54 percent in 1993. Elsewhere, those proportions were 10 percent in 1977 and 56 percent during the period 1994-98. That transfer of the source of scholarly journal articles from personal subscriptions to libraries was accomplished at a sacrifice in readers' time, since they were now required to visit libraries more frequently. Because library prices increased at an accelerated rate, libraries began to cancel duplicate subscriptions and also rely much more heavily on interlibrary borrowing and document delivery to replace expensive and infrequently read journals. The 5.7 million scientists in 1995 who personally subscribed to an average of 2.9 fewer journals represented an annual loss of billions of dollars in revenue for publishers that was recovered by raising the price of library subscriptions.

To explain how price sensitivities occur with personal and library subscriptions (Tenopir and King 1997), we developed cost models of alternative ways of obtaining articles: (1) readers subscribing or using the library, and (2) libraries purchasing or relying on obtaining separate copies. For both readers and libraries, we established a break-even point of the amount of reading below which it is less expensive for a reader to use the library (or for a library to obtain separate copies) and above which it is less expensive to subscribe as either a reader or a library. Thus, over a range of prices we established break-even points for both personal subscriptions and library subscriptions. (We recognize, of course, that to both libraries and individuals, those figures are unrealistic because they do not take into account whose budget is being affected. However, for institutions that ultimately provide the funding for both, those are useful exercises.)

Using the distribution of reading by scientists and the reading of library journals, we established the number of journals affected by increasing prices. For example, at a personal subscription price of $50, it would be cost-effective, on average, for a scientist to subscribe to ten of the eighteen journals in which at least one article is read. At a personal subscription price of $150, it is still cost-effective for a scientist to subscribe to one journal. Thus, the amount of reading of individual journals by scientists is very important in determining the sensitivity of demand to price changes.

For the library, however, the total amount of reading of library journals is much higher and, as a result, demand is much less sensitive to price changes. When applying break-even points in a library setting, it is cost-effective to pay a subscription price of $100 to purchase 87 percent of the library-provided journals read by users. Even at a subscription price of $250, it is cost- effective to purchase 76 percent of the journals, while at a subscription cost of $1,000, nearly half of the journals should be purchased.

If personal and library subscription bases are 2,500 subscribers at $150, a price increase to $250 would decrease the number of personal subscriptions to 719 (a loss of 1,781 subscribers), but library subscriptions would only drop to 2,284 (a loss of 216 subscriptions). That is the reason that the price increases over the years that have severely affected personal subscriptions have not crippled publishers who have been able to increase subscription prices to libraries at a greater rate than inflation without seeing a dramatic decline in the library market.

Costs of Electronic Journals

Electronic journals today still comprise a relatively small percentage of scholarly journal publications, but change is coming quickly, as demonstrated by the self-proclaimed, authoritative source of scholarly electronic journals being developed by CIC, a consortium of university libraries for the Big Ten and the University of Chicago. Two types of electronic journals are replicates of paper journals. Some publishers duplicate the journals in parallel, offering both electronic and paper copies; other electronic journals offer no paper counterpart, but nevertheless replicate the features of print journals. A third type of electronic journal includes enhanced features that add value to traditional journals, and may be an electronic counterpart to a paper journal, or an electronic-only journal. A variety of secondary sources are available to assist the user in identifying those journals including the University of Houston index [formerly http://info.lib.uh.edu/wj/] supplemented by Charles W. Bailey's Scholarly Electronic Bibliography. The University of California at San Diego also hosts a broader list of electronic journals, NewJour.

Electronic-only journals typically save reproduction and distribution costs of $40 per subscription as well as some costs associated with paper issues such as color processing, cover stock, etc. However, those savings are partially offset by the costs of electronic storage, software, and typically higher labor costs. Holmes (1997) shows that electronic-only publishing saves only $7.14 per subscriber — not enough to make an appreciable dent in subscription costs.

The reason for the similarity in cost structure is that most publishing activities are common to both paper and electronic publishing. Publishing-overhead costs are common to both media, as are the costs for support activities such as editorial work, sales, etc. The major difference between the two media is in production costs — the costs of paper production are replaced by the costs of electronic storage for online access and disk production for CD-ROM. Paper distribution is also replaced by online or CD-ROM distribution; thus electronic production and distribution costs are much lower than paper costs. However, those costs represent a relatively small percentage of total costs for low-circulation journals such as those serving low-population disciplines.

Parallel paper and electronic publications typically cost more than paper alone because systems-related costs are added to the traditional paper publication costs. However, we reiterate that those additional costs are not appreciable. The literature gives numerous examples of the low cost of article processing for electronic journals, including the CIC in its annual report. Nevertheless, it is not clear that all costs are included in those figures, including the costs of preparing intellectual content, computer and telecommunications infrastructure costs, and storage costs. Furthermore, most of the costs are for small journals. We have reexamined both the 1970s and 1980s data and discovered that the unit cost of small journals is less than that of larger journals, contrary to the notion of economies of scale. Evidence suggests that the fixed costs per article tend to be relatively low for small journals, rise significantly for medium-sized journals, and fall again for the largest journals.

The wide range of value-added processes that may be offered with exclusively electronic journals will improve communication, but raise the cost significantly. For example, publishers will be able to provide a database of archival journals, single journals, individual articles, or selected sections of articles. Various levels of information can be made available for reader examination, including titles, abstracts, reviews of the article, accompanying data, and appendices. Sets of articles on specific topics can be sent automatically to users, based on a profile of reader interests. Quality of older articles can be rated by citation counts of authors (before or after publication), ratings offered by readers, or ratings garnered from a panel of invited referees. Multimedia (e.g., sound, motion, and extended graphics) and interactivity among users (e.g., between readers and data, among readers, and between readers and authors) are features that can be added to electronic journals. Pricing strategies will need to be established for each feature, since such changes will affect costs, information and service attributes, and use.

The cost savings to buyers of exclusively electronic journals that have approximately 500 circulation is certainly less than 5 percent. Journals of 10,000 circulation could save approximately 50 percent if the savings were passed on to subscribers. It is useful to examine those costs in terms of buyer type (individual readers or libraries) and the extent to which the journals are read.

Choices between paper and electronic journals depend to some degree on price (circulation) and frequency of reading. Assume that a journal with a circulation of five thousand is priced at $120, and it costs readers approximately $11 to process and store the journal. The cost per reading then varies by the number of readings.

Some journals priced that low are frequently subscribed to by individuals. It makes sense to provide such a journal by subscription to the frequent readers and provide electronic copies of articles for the infrequently read articles (ordered by the readers or their library, if the article is not easily located).

Assuming that a library purchases the $120 journal (five thousand circulation) and incurs processing costs of $68 and per use cost of $1.34 (Tenopir and King 1997; Griffiths and King 1993), the cost per reading would be:

At that low price, it may be hard for electronic journals to compete on a cost-per-reading basis for frequently read paper journals in libraries.

Those low costs are what must be considered in comparing electronic and paper access to traditional journals. It well may be that large-circulation journals will continue to have a paper-publication niche that serves scientists who thoroughly read the journal, even considering favorable attributes of electronic publishing.

A 500-circulation journal might be priced at about $840. With the same $11 cost for processing and storage, it would cost a reader the following to subscribe:

In that case it might be less expensive to use electronic access considering the likely number of readings. With library subscriptions, the comparable costs are:

Depending on electronic-access costs, it may be less expensive for most libraries to obtain electronic access. Thus, there is little economic incentive to publish small-circulation journals in paper whenever electronic access costs are low enough.

Success in the future will also depend in part on inexpensive access to separate copies of articles and the ability to easily identify and locate needed articles by — at least initially — the use of traditional bibliographic databases.