Scientists frequently read scholarly journals in order to keep current in their disciplines and to apply the information to their research, teaching, and other activities. Scholarly journal articles are read far more frequently than any other type of material, such as trade journals, books, technical reports, etc. In fact, scientists in universities average 188 article readings per year and non-university scientists average 96 readings. Evidence from surveys of scientists from 1977 to 1996, indicate that the extent of reading has remained relatively constant over that period of time. The average amount of reading per article is about 500 to 1500 readings, depending on the field of science. Thus, a typical 1995 journal having 123 articles probably has well over 100,000 readings. However, it is important to recognize that scientific disciplines range in size from the 100's to the 100,000's, and journal readership reflects this phenomenon.

University scientists use the scholarly information for many purposes: over one-half of the readings are for current awareness or professional development, 75% for research, 41% for teaching, and 13% for administrative and other purposes (Tenopir & King, in press). Much of the information is highly important to teaching and research. Of the 188 annual readings per scientist, 13 are said to be absolutely essential to their teaching and 23 of them are absolutely essential to their research. University scientists whose research or teaching has been recognized through awards or special recognition read more than those who haven't (25-33% more reading, depending on their type of work). Economic value of scholarly journals can be considered in terms of purchase value of what one is willing to pay for the information, and use value in terms of outcomes from use of the information. Not only do these scientists pay for an average of 3.9 subscriptions per year, they also "pay" an average of 182 hours of their time reading this information. Their time is a critical resource and how they choose to use this time is an indicator of the value of the activity in which their time is spent. Indicators of the use value are that 95% of the readers indicated that some favorable outcomes resulted from reading the articles; nearly two-thirds indicated that their last reading resulted in improving their quality of research or teaching, and that their last reading helped them perform research or teaching better (33%), faster (19%), or resulted in using less time or money (16%).

Similar results were observed for non-university scientists (Griffiths & King, 1993). The primary purpose of reading scholarly articles was observed to be for: current awareness and professional development (30% of readings), research and development (57%), and giving advice, writing, or making presentations (13%). This information is rated very high in importance to all activities performed by the scientists. In fact, when compared with six other resources used for their work (e.g., laboratory instrumentation, computing, receiving advice from colleagues, etc.), information found in documents is found to be rated highest or second highest in importance for all activities. Award winners are found to read 53% more articles than non-award winners, and all other indicators of achievement are found to be correlated with amount of reading. When dollar values are observed, the purchase "value" of journals is found to be at least $6,000 per scientist and use value is estimated to be over $25,000. Other indicators of use value are that the amount of reading is correlated with five measures of productivity, dollar savings achieved from readings are estimated to be about $380 per reading, and a high proportion of readings are said to result in improved quality of research, and so on.

Over the years we have tried to demonstrate the usefulness and value of scholarly journals from a range of viewpoints and perspectives, all of which point to how essential they are to scientists and other professionals. With scholarly journals beginning to transform to electronic media, we feel that it is important that they maintain their quality and other attributes that contribute to current high levels of use, usefulness, and value. In fact, there is the potential that communication technologies might even lead to greater use, usefulness, and value in the future, if we can learn from past mistakes and build upon what we have learned.

One clear lesson from the past is that communication through scholarly journals involves a system of authors, primary publishers, secondary services, libraries, readers, and funders of these and other participants. As a system, there are systemic and economic interdependencies in which any economic or other type of action is likely to have a ripple effect on all participants. This has been clearly demonstrated with pricing policies in the past. Regardless of why prices began to escalate in the 1970's, the consequences are that personal subscriptions (in particular) began to decline; revenues dropped; publishers increased prices to libraries; readers relied more on their libraries (and others) as a source of articles at a substantial cost to their time; libraries began to cancel duplicate subscriptions and then subscriptions to expensive, but infrequently-read journals, and they relied on interlibrary borrowing and document delivery services to satisfy demand for these articles; and secondary services became more important as a means of identifying and locating needed articles. Meanwhile, amount of authorship and readership remains about the same over the last 20 years.

The unfortunate fact is that while publishers struggle, libraries pay more for fewer journals and have increasing costs of obtaining separate copies, and readers pay more in their time to obtain articles; the overall total cost of the journal system appears, if anything, to have increased with the costs having shifted among participants and processes. There may have been ways to have avoided this lose, lose, lose, lose situation. To examine alternatives to past economic policies, it is useful to understand past economic costs of publishing, effects of pricing policies, and changing distribution patterns.

Journal publishing costs are often characterized as having two components: very large, fixed costs associated with processes necessary to produce the first copy of journal issues; and small costs necessary to reproduce and distribute the issues. Unfortunately, publishers have been reluctant to make their costs known so that one has to gather snippets of data to know the extent of these costs. Evidence suggests that the costs of referencing, editing, composing, etc. scientific scholarly articles can be in the neighborhood of $4,000 per article when all direct and indirect costs are considered. Typical runoff and distribution costs are about $40 per subscription. Thus, such a journal with 100 articles would cost about $400,000 to prepare the first copy (ignoring non-article text). Because of the very high fixed cost, a journal with a small circulation of 1,000 would require a price of $400 to recover the fixed costs, plus $40 to recover reproduction and distribution costs. As circulation increases, the minimum unit costs would decrease rather rapidly: $80 at 10,000 subscriptions, and $44 at 100,000 subscriptions.

Scientific disciplines and corresponding scholarly information needs can range in size from a few hundred to hundreds of thousands of scientists. Thus, journals tend to "bundle" articles addressed to disciplines having a wide range of sizes. This means that journals serving small disciplines must find ways to reduce the first-copy costs or charge a high price for journals. Evidence suggests that both phenomenon occur to some degree. Even though the average circulation of journals is about 5,800 subscriptions per journal, about one-half of the journals have fewer than 1,900 subscribers and a substantial number are under 1,000 subscribers. This partially explains why so many journals have high prices. Also, costs have increased because the size of journals has increased substantially in number of articles, pages, and issues.

On the other hand, pricing strategies have contributed to the spiraling price increases, smaller circulation, and, in turn, even higher prices. It is abundantly clear that individuals and small organizations, in particular, are not willing to pay for relatively infrequently-read, high-priced journals because they have the option of going to libraries (and elsewhere) to get access to needed journals or articles less expensively. This subscription base is highly sensitive to price changes, because the small amount of reading means that the cost per reading is relatively high. In these instances, the cost per reading can be far less by going to the library. Yet, since journals in larger libraries have a relatively large (and increasing) readership, it often costs less (per reading) to subscribe than borrow; even though subscription prices are increasing. Thus, subscriptions to larger libraries are much less sensitive to price increases.

The cost of electronic journal publishing will be less than traditional paper, but not all that much. Most of the article processing costs will be the same, although slightly less, unless some of the value-added processes are discontinued or modified. The reproduction and distribution costs will be replaced by electronic storage and distribution, which is much less expensive. However, the paper reproduction and storage costs are small compared to the fixed costs for most scientific journals, and publishers must still recover the large first-copy costs. What may change are the ways in which articles are provided and how publishers charge for their services.

Internet and other communication technologies open up a whole range of possible new-article service provision. To begin with, publishers will be able to provide a database of journals, single journals, individual articles, or parts of articles. Various levels of information might be made available on examination including titles, abstracts, reviews of the article, accompanying data, appendices, and so on. Sets of articles could be sent automatically to users based on profiles of readers' interests. Quality of older articles can be rated by citation counts of authors (before or after publication), ratings made by readers, or ratings made by a panel of referees. Since any such changes will affect costs, information and service attributes, and use, pricing strategies must be established for each.

Product differentiation such as those mentioned above will require price discrimination as well. However, another form of price discrimination may also take place to avoid the spiraling prices observed over the last two decades. At least some of the problems could have been avoided if individuals or libraries serving small groups of individuals had been charged less than large libraries where readership of journals is high. It does not seem to make sense to charge a reader $500 for a journal that is read, say, 10 times (i.e., $50 per reading), and charge the same to a library where the journal is read, say, 100 times (i.e., $5 per reading). The individual will cancel the subscription and use a library's copy, thus denying revenue to the publisher, requiring the reader to lose time, and costing the library a small amount for its use. In paper journals, it would be best to charge the individual $40 for reproduction and distribution, plus some amount for reading, say, $5, or $90 total, thereby providing $40 revenue that could be applied to fixed costs. With electronic journals, nearly all of the $90 price could be applied to fixed costs, circulation and revenue would rise, and library subscription prices reduced (recognizing that some readership will also decline). Thus, some form of price discrimination based on actual or potential readership would seem to make sense. This kind of strategy would strive toward a win, win, win, win circumstance and can be achieved because electronic technology can also help keep count and provide reasonable payment mechanisms. Below are some details supporting this conclusion.

There has been very little data published on the cost of traditional journal publishing, let alone electronic journal publishing. Data that have been published can be misleading for two basic reasons:

(1) Frequently, costs are given as proportions of first-copy (pre-run), and runoff and distribution (e.g., 80% of the costs are first-copy and 20% runoff and distribution). The problem with this description is that proportions depend largely on circulation (and to a lesser degree on other parameters such as number of articles, pages, etc.). For example, assume the first-copy cost of a journal is $400,000, and runoff and distribution cost is $40 per subscription. A 1,000-subscription journal would have proportions of 91% first-copy, and 9% runoff and distribution. A 100,000-subscription journal would be exactly the opposite: 9% first-copy, and 91% runoff and distribution.

(2) Reported production cost data have ranged from $200 to $8,000 per article. It seems clear that such data do not represent an equivalent representation of what the costs cover.

To address these two problems, we have derived an economic cost model of journal publishing. The model is subdivided into four basic functions: (1) article processing, (2) journal runoff and distribution processing, (3) non-article processing, and (4) publishing support. These four functions are described briefly below.

Article processing includes all the activities that begin with receipt of an article manuscript and ends with "first-copy" composition, typesetting and/or engraving. Such activities include receipt processing, initial disposition decision-making, identifying reviewers/referees, review processing, subject editing, special graphic and other preparation, formatting, copy editing, processing author approval, redaction, and preparation of master images. Most of the activities are designed to enhance favorable attributes of article information and its presentation. Typical article processing costs of traditional scientific scholarly journals are about $2,000 per article or about $165 per article page published.

We believe that most of article processing activities performed for traditional journals should also be done for electronic journals. The largest cost savings achieved for electronic journals appear to be in the composition/typesetting costs and savings here would be at most in the 10% range. Advocates of electronic publishing suggest that savings can be achieved through Internet transmission and electronic input of manuscripts, but this can also be achieved with traditional publications. The duration of publication can be reduced by electronic transmission, but not a great deal because time delays are due largely to human intervention (i.e., edition, refereeing, etc.). The electronic journal advocates also quote very low per-article costs (i.e., typically in the range of $200 to $500 per article). It is not known why these costs are so low, but it may be that some resources are donated. However, a more plausible reason is that these low-cost journals do not incorporate all the "bells and whistles" that are found in more established journals. We have observed that small journals tend to cost less than large journals (i.e., unit costs are less), even for traditional journals.

For traditional journals, the runoff costs include plate-making, set-up, printing, collating, and binding. Typical costs come to about $25 per journal per year, recognizing that costs vary by number of articles, pages, issues, etc.; and quality of paper, binding, etc. Distribution costs include mail processing (i.e., inserting, sorting, sealing, labeling, etc.), postage, and subscription maintenance. Typical costs are about $15 per subscription, again, recognizing that number of issues, weight, etc., have a bearing on costs. These costs are much less for electronic journals. Electronic journals require electronic storage (but no runoff costs) and the distribution costs are far less than for traditional journals. However, cost savings attributable to electronic publication are not likely to be much more than about $35 per subscription. Thus, a $500 journal would only cost about 7% less, all other attributes remaining equal.

Non-article processing includes essentially the same activities as article processing, which include, but are not limited to covers, tables of content, editorials, book reviews, letters to the editor, and advertising pages. In the long-run it is not at all clear whether such information will be included in electronic journals. Currently, scientific scholarly journals annually average about 290 non-article pages (compared with 1,439 article pages). Costs of non-article processing comes to about $100 per non-article page (compared with $165 per article page). Publishing support activities include marketing and promotion (e.g., advertising, catalogs, etc.), administration (e.g., personnel, accounting, other resource allocation, etc.), managing rights and copyright protection, financing (e.g., new product research and development, interest payment, capitalization, payment of taxes, etc.), and other indirect costs (e.g., insurance, taxes, utilities, parking, travel, etc.). These costs tend to be about 40% of article and non-article processing costs and 21% of runoff and distribution costs. Thus, again, the support costs may not be much less for electronic journals than for traditional journals.

Together, a typical scientific scholarly journal would have a fixed cost of about $403,500, and a variable cost of about $40 per subscription (both in 1995 dollars) (Tenopir & King, 1997). The minimum price that would be necessary to recover publishing costs for various circulation levels would be as shown in the Figure 1. At 500 subscriptions, the minimum price would be $847, which drops nearly in half to $444 at 1,000 subscriptions. At 10,000 subscriptions, the minimum price would be $80. The principal point is that these costs will not decrease all that much with electronic publishing if all aspects of the journals remain the same. The price at 500 subscribers would decrease about $110 ($847 to $738), but would be about one-half the price at 10,000 subscriptions ($80 to $42).

It may be that Internet and complementary services will affect production in more important ways than in reducing costs of journals. The new technologies can provide the flexibility of multiple journal products such as selling a "database" of journals, single journals, single articles, pages of articles, or paragraphs from articles. Other types of products might include early preprints of articles and later edited versions; the quality of articles might be rated (by citation counts of authors, by past readers, by a panel of referees, etc.); sets of articles might be made available or automatically sent to readers based on profiles (like Selective Dissemination of Information); or various levels of article information might be made available for examination, such as titles, abstracts, reviews, entire articles, accompanying data, appendices, etc.

There is little doubt that journal prices have increased appreciably over the last 20 years (Tenopir & King, 1997). US scientific scholarly journal prices are estimated to have increased from an average of $39 in 1975 to $284 in 1995. Thus, the average prices increased by a factor of 7.3, or 2.6 times in constant dollars. Some of this price increase is attributable to an increase in size of journals (please see Table 1). However, these increases account for a relatively small proportion of the total increases. A much more plausible explanation is that early price increases in this 20-year period were attributable to large inflation, fluctuating international monetary exchange, and company and agency cutbacks in discretionary funds available for subscriptions. Once prices increased at a rate greater than inflation, subscriptions began to drop; particularly personal subscriptions. In fact, personal subscriptions of US scientists dropped from an average of 5.8 subscriptions per scientist in 1975 to 2.9 in 1995. While the total number of subscriptions increased during that period, the average circulation per title decreased. This resulted in annual losses of billions of dollars of revenue to publishers in that time period.

As a result of decreased revenue from reduced personal subscriptions, publishers raised institutional subscription prices even higher to make up for the lost revenue. We show that this strategy has been partially successful because institutional demand for journals is substantially less sensitive to price changes than personal subscription demand (Tenopir & King, 1997). For example, assuming a base of 2,500 subscriptions, increasing a price from $150 to $250 would potentially reduce individual subscriptions about 1,800; whereas institutional circulation would drop only about 200 subscriptions. The problem is that the pricing strategies that have led to the current "crisis" have resulted in less revenue to publishers and higher costs to individuals, libraries, and funders of the libraries and their users (i.e., lose, lose, lose, lose).

Electronic journal publishers can continue current pricing policies and merely charge for online subscriptions, albeit perhaps at a somewhat lower price. However, new technologies may result in flexibility of delivery, as mentioned above, which also suggests new pricing strategies and payment mechanisms (Varian, 1996; King, in press). Pricing of electronic journals is likely to migrate to various forms of price discrimination based on different amounts of output (i.e., several journals, a journal, article, page, etc.); delivery media (print, online, combination); service attributes (e.g., rush versus regular processing); and customization (e.g., preprint versus edited, etc.). Differential prices might also include classes of buyers based on potential readership. Payment mechanisms can also be adaptable to service provision and user needs (Sirbu, 1995).

Both authorship and readership of scholarly journals have remained relatively constant from 1975 to 1995. For example, there were about 0.118 articles published per scientist (in the US) per year in 1975 and this number was 0.101 articles in 1995. Even though the average number of subscriptions per scientist has dropped in half; evidence suggests that readership has not decreased (please see Table 2). Instead of readership declining, there has been a dramatic shift in source of articles read (please see Table 3). Most of this shift in reading sources is from personal subscription to library-provided journals. In 1977, about one-fourth of university readings were from library-provided journals (based on a national survey), but this rose to about 54% (based on a 1993 University of Tennessee survey). For non-university scientists the proportion rose from 10.4% to 37.4% (based on surveys of scientists in seven companies).

It is clear that libraries are used much more now for reading than in the past. Nearly all university scientists (97%) say that they have used their library's journal collection and 69% of non-university scientists indicate that they have. One reason for this is that increased prices have meant that scientists cancel subscriptions and rely on getting their articles from libraries. Evidence shows that they are willing to spend their time to go to the library instead of paying a high price for a journal, even though their organization might pay for it (Tenopir & King, 1997). They clearly balance their cost as evidenced by the fact that they subscribe to inexpensive journals that they frequently read, and go to the library to read expensive journals that they read relatively infrequently. In fact, from the perspective of cost per reading, they generally choose in an economically rational manner; that is, the average cost per reading to use a library journal is about $11.50 per reading. Thus, there must be at least 33 readings of a personal subscription of a $250 journal to make it worth while subscribing to the journal (considering the price, processing cost, and access cost). For a $1,000 journal, the amount of reading would have to be at least 128 readings. Yet, the average amount of reading from personal subscriptions is only 18.8 readings per journal per scientist.

With electronic journals, one can apply a similar cost analysis to recommend or provide guidelines to users concerning whether they should subscribe to a journal (either in paper or electronic) or obtain separate online copies. Similarly, if a library can display journals and distribute copies of articles by local area networks, one can demonstrate a similar cost trade-off between using these journals compared with obtaining personal subscriptions.

Libraries are also faced with the option of subscribing to journals or obtaining separate copies through interlibrary borrowing or document delivery (photocopy or electronic). The choice is straightforward-subscribe to frequently-read, inexpensive journals, and obtain separate copies of infrequently-read, very expensive journals. Typically, with all activities included, it costs a library about $70 per subscription to process a journal and an average of $1.40 per reading to shelve issues and/or photocopy an article (Tenopir & King, 1997). The total cost of interlibrary borrowing or using a document delivery service is in the $15 to $25 range per item obtained. One can calculate the breakeven amount of reading by knowing the price of a journal (and processing costs) and the cost of obtaining separate copies of articles. The problem is that the cost of electronic access varies dramatically among sources. For this reason, we provide a guide to assess breakeven points based on subscription price (plus processing costs mentioned above) and unit cost of obtaining separate copies (please refer to this guide in Table 4). As an example, a journal priced at $250 would require at least 17 readings if it costs $20 to obtain separate copies. At $5, the breakeven point is 89 and at $35, the breakeven point is 10 readings. Thus, at a given subscription price, the breakeven point goes down as the cost of obtaining copies increases. If the choice is between obtaining an electronic subscription and using electronic article delivery, the trade-off analysis is similar, recognizing that the processing costs are probably less for both options.

The average number of readings of library journals appears to be roughly 150 readings per journal (over its life). Examination of Table 4 shows that even at a price of $1,500, and average-read journal (150 readings) should be purchased if the cost of separate copies is more than $15 per copy. However, reading of library journals tends to be highly skewed; that is, a few journals are read a great deal and many journals are infrequently read. For example, several studies show that about one-half of the journals in libraries have fewer than 50 readings, only about 25% have more than 150 readings, and 15% have more than 250 readings.

Based on anticipated costs of electronic journals, we believe that some journals will lose traditional subscriptions and rely much more heavily on revenue from distribution of separates. These journals are likely to be the expensive journals that normally serve a small community of scholars. However, the publishers who do this must be willing to charge an amount that will cover the cost of both frequently and infrequently-read articles. This, of course, is what "bundling" of articles in traditional journals now does. Otherwise, high quality articles that are of interest to a small audience will be lost to the publishing process.

Sources: King, et al, 1981; Tenopir & King, 1997

Table 2
Average Number of Scientific Scholarly Journal Readings
Per Scientist Per Year by University and Non-University Scientists 

1977

1984

1993

1977

1984

1985

1991-1997

*Above the breakeven point a library should subscribe and below it, the library should rely on separate copies.

Griffiths, J.M. and King, D.W. (1993). Special Libraries: Increasing the information edge. Washington, DC: Special
Libraries Association.

King, D.W. (in press, April 1998). Some economic aspects of the Internet. Journal of the American Society for Information
Science.

King, D.W., McDonald, D.D., & Roderer, R.K. (1981). Scientific journals in the United States: Their production, use, and
economics. Hutchinson Ross Publishing: Stroudsbourg, PA.

Sirbu, M.A. (1995). Creating an open market for information. The Journal of Academic Librarianship. Managing Technology
Issue. Ed. By C.B. Lowery, pp. 467-471.

Tenopir, C., & King, D.W. (in press, Summer 1998). The transformation of scientific scholarly journals: Trends in the
economics, production, information seeking, and use.

Tenopir, C., & King, D.W. (1997). Trends in scientific scholarly publishing. Journal of Scholarly Publishing, 48(3), 135-170.

Varian, H.R. (1996). Differential pricing and efficiency.\