Chapter 18—
The Library and the University Press
Two Views of Costs and Problems in Scholarly Publishing

Susan Rosenblatt and Sandra Whisler

Introduction

The conflicts becoming apparent in scholarly communication have been anticipated for almost two decades. In 1982 Patricia Battin wrote:

During the decade of the 1970's, librarians faced declining budgets, increasing volume of publication, relentless inflation, space constraints, soaring labor costs, a horrifying recognition of the enormous preservation problems in our stacks, increasing devastation of our collections by both casual and professional theft, and continuing pressure from scholars for rapid access to a growing body of literature. It is ironic that both librarians and publishers introduced computer applications into libraries and publishing houses to save the book, not to replace it. Both were looking for ways to reduce labor costs rather than for visionary attempts to redefine the process of scholarly communication.... The former coalition shattered and publishers, scholars and librarians became adversaries in a new and unprecedented struggle to survive in the new environment, each trying in his or her own way to preserve the past and each seeing the other as adversary to that objective.^[1]

Library Costs

Library Materials: Print

The results of the economic crisis in scholarly publishing were documented statistically in 1992 in University Libraries and Scholarly Communication.^[2] Some of the principal findings included the fact that although materials and binding expenditures remain a relatively constant percentage of total library expenses, there has been a hidden, but significant, change in the ratio of books to serials expenses. Although materials expenditures have steadily risen, the average numbers of volumes added to library collections annually continue to decline. Not only are libraries spending more and receiving fewer items in absolute terms, but also libraries are collecting

an ever smaller percentage of the world's annual output of scholarly publications. Since 1974, even increases in university press outputs have outstripped increases in library acquisition rates.

Moreover, the study documents that some of the fields experiencing the greatest increases in their share of the total output are precisely those with the highest average per-volume hardcover prices: business, law, medicine, science, and technology. According to the report, science had the highest average prices; social sciences and business experienced price increase rates closer to the GNP deflator (p. xix).

Another finding was that serials prices consistently increase faster than general inflation. Serials had an overall annual inflation rate of more than 11% from 1986 to 1990. Prices of scientific and technical journals rose at the highest rates (13.5% per year, on average, from 1970 to 1990), and the most expensive serials experienced the largest relative price increases. In contrast, book prices inflated at 7.2% per year, while the general inflation rate averaged approximately 6.1%. In some institutions, science journals could comprise only 29% of the total number of journal subscriptions but consumed 65% of the serials budget. According to the Mellon report, "three European commercial publishers (Elsevier, Pergamon, and Springer ...) accounted for 43% of the increase in serials expenditures at one university between 1986 and 1987" (p. xxi). The report does not introduce the question of the extent to which these inflation rates in the prices of scientific journals reflect increasing costs of production, expansion in content, price gouging, or the market value of the information itself-a value that might extend well beyond the university.

Brian Hawkins's 1996 study of library acquisition budgets of 89 schools finds that although budgets nearly tripled from 1981 to 1995 and increased by an average of 82% when corrected for inflation using the Consumer Price Index, the average library in the study lost 38% of its buying power. In the 15 years covered by his study, the inflation rate for acquisitions was consistently in the midteens. Confirming the Mellon study, he finds that the costs of some science journals increase more than 20% per year. He also notes that the trend line for average increases in library acquisition budgets is downward, accelerating the rate of decline in volumes added to collections.^[3]

Harrassowitz regularly alerts libraries to subscription pricing information so that its customers can plan in advance to adjust purchasing patterns to stay within budget. In November 1996, Harrassowitz provided firm 1997-98 subscription pricing for six publishers publishing the majority of the STM (science, technology, and medicine) journals.^[4] The announced price increases ranged from 1.2% to 22%, averaging 11%. According to Harrassowitz's analysis, libraries categorized as "General Academic/including Sci-Tech" could expect average price increases from the six publishers of almost 14%.

Peter Brueggeman from the Scripps Institution of Oceanography (SIO) Library at UCSD has discussed the problem from the perspective of a science li-

brary.^[5] During the five-year period from 1992 to 1996, journal subscription costs at SIO rose 57% but the recurring collections budget increased 2.3%. Brueggeman singles out Elsevier and Pergamon for particular analysis: "Elsevier titles had a 28% increase between 1995 and 1996 and a 32% increase between 1992 and 1993. Pergamon titles had a 29% price increase between 1995 and 1996 and a 17% price increase between 1992 and 1993."

Various authors have demonstrated that not only do the most expensive journals experience the highest rates of inflation, but they are also among the most used. Chrzastowski and Olesko found that over a period of eight years, the cost of acquiring the ten most used chemistry journals increased 159% compared to an increase of 137% for the 100 most used journals.^[6] During the same period, usage of the top ten journals increased 60% compared to an increase of 41% for the top 100 journals.

Library budgets that inflate more slowly than the rate of inflation for scholarly journals will cause a steady decline in the number of titles held in each library. Because libraries generally cancel journals on the basis of use, high-use, high-inflating titles may be protected. This protection results in a gradual homogenization of collections among libraries. Lesser-used titles, many with low prices and low inflation rates, will be crowded out faster than the general rate of decline in library subscriptions.

Figure 18.1 illustrates a hypothetical scenario. This scenario assumes that the collections budget is inflated by 4% per year. However, the average rate of inflation in the cost of scholarly publications is greater. The graph shows that if science journals, because they demonstrate high usage patterns, are canceled more slowly than other titles, then science journals will eventually crowd out other journals. In the example, the budget for science journals is allowed to inflate at approximately 8% per year (slightly less than one-half the actual inflation rate, but twice the rate of inflation in the total serials budget). Other, lesser-used journals, with lower subscription prices and lower rates of inflation, therefore must be canceled more rapidly in order for the collections budget to be balanced. Within a few years, the crowding-out effect from protection of high-use/high-price/high-inflation journals is quite noticeable. While no particular library may implement a budget strategy exactly like that depicted, all libraries tend to retain subscriptions to the highest use journals and to cancel first the lesser-used journals. Although the curve may change as the time line lengthens or shortens, the eventual result will be similar to that shown.

Library Materials: Electronic

As yet, there is no evidence that the emergence of electronic journals will improve the fundamental economic problems in the cycle of scholarly communication. The basic premise of publishers is that they must protect their current revenue base and secure guarantees to cover future inflation and increases in content.

[Full Size]

Figure 18.1.
Crowding-out Effect

Thus, publishers frequently structure their initial subscription pricing for digital journals upon the actual cost of paper subscriptions acquired by the institution with which the publisher is negotiating. The proposed base subscription rate may include all subscriptions: library, departmental, personal, and other types identified with the campus, thereby greatly increasing the price that the library would have to pay to receive the digital journals. Clearly, publishers are concerned that network availability of electronic journals on the campus network will undermine nonlibrary subscriptions to the print versions.

In early 1996, Ann Okerson reported that

In general electronic licenses so far have cost on average 1/3 more than print equivalents.... Publishers are setting surcharges of as much as 35% on electronic journals, and libraries simply do not have the capacity to pay such monies without canceling a corresponding number of the journals of that particular publisher or dipping into other publishers' journals.^[7]

Moreover, during license negotiations for certain electronic journals, libraries may be asked to consent to such provisos as the following:

1. That there be multiyear price increase guarantees to compensate for inflation, often at somewhat lower rates than historical rates of price increase for print materials

2. That there be upward price adjustments for increases in content, often capped at lower rates than typical for print journals

3. That the publisher be protected against declines in revenue through cancellation

4. That fair use rights typical for print journals be abrogated for the digital journals

Maintaining a combination of print and electronic subscriptions for a multiyear period without incurring substantial new marginal costs for electronic versions and ensuring a cap on inflation are attractive to libraries. But neither "feature" of these new licenses will alter the basic economic difficulty facing libraries: inflation in the price of scholarly information outstrips libraries' ability to pay. In fact, by being locked into multiyear agreements that ensure price increases to particular publishers, libraries hasten the rate at which other journals and monographs are crowded out of the market.

Not all scientific publishers have negotiated as described above. For example, the American Physical Society and the American Mathematical Society offer electronic versions of their journals free to subscribers to the print editions. Clearly, publishers must find revenue streams that will enable them to survive, and the pricing structures for both print and digital journals are the key to those revenue streams. To base a pricing structure for electronic publishing on the costly print model will not be economically viable in the long run (it may, in fact, be unsustainable in the short term as well). Libraries' declining budgets will result inevitably in cancellations to avoid the structural problems associated with doubledigit inflation, thereby undermining the publishers as well.

The current economic model for scholarly publication cannot be sustained. Continued escalation in the prices for scholarly journals, stagnation in library budgets, and isolation of the creators and consumers of scholarly information (the faculty) from the effects of the economy could lead to the collapse of the system of scholarly communication itself.

Operations Costs in Libraries

Library operations costs associated with printed scholarly journals include the costs to acquire, process, preserve, and circulate journals. Each library's costs differ based on the organizational structure, degree of centralization or decentralization of processes, differentials in salary and benefit scales, effectiveness of automated systems, success at reengineering internal operations, and other factors.

University Libraries and Scholarly Communication reports that "salaries as a percentage of total library expenditures have declined over the past two decades, while other operating expenditures (heavily reflecting computerization) have risen

markedly" (p. xxii). The report infers that the increases in other operating expenditures reflect automation of technical service operations such as acquisition, cataloging, serials control, and circulation. It also notes that despite the decline in salaries as a percentage of total library expenses and the increase in other expenditures, "the number of volumes added per staff member has declined" (p. xxii), implying that there has not been a measurable staff efficiency gain from the investments in automation. In fact, on average, library staff increased by a total of 7% from 1970 to 1985 and by 6% from 1985 to 1991. Thus the rise in the nonsalary operations portion of the total operating expenses did not occur through staff reductions but rather as a result of budget augmentation for nonsalary items.

Presumably, greater efficiency in processing and circulation coupled with declining acquisitions should have resulted either in staff reductions or in substantial shifts of personnel away from the back room of technical processing to and into service to faculty and students, but it is not possible to discern from ARL statistics whether this is so. The ARL did not report service transactions until 1991, so one cannot discern changes in user demand for the earlier periods. Between 1991 and 1996, the ARL reports steady increases in interlibrary borrowing, library instruction, reference transactions, and total circulation. During the same period, total staff has declined by 2%.^[8]

The inability to learn from the ARL reports or other reliable studies how libraries might be changing their staff allocation among operations and services reflects a serious flaw common to almost all analyses of library costs relating to both collections and operations. It is not obvious to what extent nonsalary investments, for example in automated systems, have actually improved processing productivity or the quality of services rendered by staff; nor is it clear whether or to what degree these investments have moderated the rate of rise of operations costs.

Library rankings typically reflect inputs such as budget, volumes acquired, number of serial subscriptions maintained, size of staff; or operational statistics such as the number of circulation transactions, titles cataloged, hours of opening, or items borrowed through interlibrary services. Ironically, the ARL Index ranks research libraries in part on the number of staff they employ; improving productivity and reducing staff accordingly would have the paradoxical effect of reducing a library's ranking vis-à-vis its peers. Developing measurements of service outcomes related to the mission of the institution would be more helpful as comparative data. For example, how do a library's collections and services contribute to research quality, faculty productivity, or student learning? The problem of defining productivity of knowledge workers was mentioned 30 years ago by Peter Drucker^[9] and is further examined by Manuel Castells in his recent book The Rise of the Network Society.^[10] Library operations represent a clear example of this productivity paradox.

William Massy and Robert Zemsky, discussing the use of information technology to enhance academic productivity in general, remark on its transformational potential, calling it a "modern industrial revolution for the university" that can

create economies of scale, deliver broad information access at low marginal cost, and allow for mass customization.^[11] The analysis they provide for the academy at large would appear to be even more pertinent to libraries, many of whose functions are of a processing nature similar to those in industry and whose services can also be generalized to a greater degree than is possible for teaching and research.

Massy and Zemsky suggest that although capital investments in technology to enhance productivity will increase the ratio of capital cost to labor cost, they may not actually reduce overall costs. But the writers argue that those investments will save money in the long term because over time labor costs rise with productivity gains and technology costs decline.

The primary purposes of automating processing operations in libraries have been to reduce the rate of rise of labor costs and to improve timeliness and accuracy of information. From the point of view of faculty and students, the service improvements are the most important result of automation. For example, on-line catalogs and automated circulation services provide users with more rapid access to information about the library's collections, reduce errors in borrowing records, and support more timely inventory control. Use of on-line indexing and abstracting services rather than the print versions preserves the scarce time of scholars and effectively extends the library's walls throughout the network.

Despite the efficiencies that automation has brought, labor costs to perform library processing operations such as ordering and receiving, cataloging, maintenance of the physical inventory, and certain user services including interlibrary lending and borrowing remain substantial. A transition to electronic publishing of journals (accompanied by the elimination of print subscriptions) could enable libraries to reduce or eliminate many of the costs of these labor-intensive operations. The freed-up resources might then be moved into higher priority services, necessary capital investments in technology, or provision of technology-based information resources. The benefits to end users could also be significant: less time spent in finding and retrieving physical issues of journals.

In the very long term, restructuring of library operations in response to electronic scholarly publishing could, in theory, result in major improvements to the quality of library services and also reduce operations costs. However, to maximize operations costs reductions, libraries will need to define better the desired outcomes of their operations investments, measure those outcomes effectively, and engage in rigorous reengineering of processes.

Several studies have attempted to quantify typical costs of acquiring journals. In a study funded by CLR (The Council on Library Resources), Bruce Kingma found the average fixed cost of purchasing a journal subscription to be $62.96.^[12] In discussing the economics of JSTOR, Bowen estimates the costs of processing, check-in, and binding to be approximately $40.00.^[13] In 1996, the library of the University of California, Berkeley estimated the physical processing costs, including check-in of individual issues, bindery preparation, and binding, for print serial subscriptions received and housed in the main library to be as low as $17.47 for

a quarterly journal and up to $113.08 for a weekly journal. Berkeley's figures exclude the costs of ordering and order maintenance under the assumption that those costs will not differ significantly for electronic journals. The figures also exclude staff benefit costs and overhead and therefore understate the true cost to the university of receiving print subscriptions. Assuming an average annual processing cost of $50.00 per print serial subscription, a research library subscribing to 50,000 titles may incur an operations cost of $2.5 million per year simply to acquire journals.

Once the library acquires these journals, it begins to incur the costs of making them available to students, faculty, and other users. In the late 1980s, Michael Cooper reviewed the costs of alternative book storage strategies.^[14] He found that circulation costs ranged from a low of $.53 per transaction in a medium-sized, open-stack research library to a high of $9.36 per transaction from a remote storage facility. Adjusted for inflation of 3% per year, these costs would range from approximately $.67 to $11.86 per transaction today. Berkeley calculates that an average circulation transaction costs approximately $1.07, and Bowen's estimate is $1.00. According to ARL Statistics, 1995-96, the mean number of initial circulations per library was 452,428. Using a circulation transaction cost of $1.00, the average ARL library spent almost $500,000 to circulate materials during the fiscal year 1995/96.

A review of the costs of acquiring and circulating print journals indicates that a transition from print to electronic journals would eventually reduce annual library operations costs related to providing the university community with the fruits of recent scholarship, but it is not clear how much of these savings might be offset by costs of technology infrastructure and equipment replacement. Large recurring expenses in support of historical print collections would continue but gradually diminish over time as the aging of the collection reduced the rate of usage. The long-term cost reductions could be substantial in the sciences where currency of information is of utmost importance. Costs associated with traditional operations and physical facilities might be more rapidly reduced were high-use print collections converted to digital form. Ultimately, the shift from labor-intensive processing operations to capital investments in electronic content (current journals and retrospective conversion of high-use print collections) might bring about the kinds of effects envisioned by Massy and Zemsky.

However, caution must be exercised in forecasting these types of savings. Despite the potential for long-range cost reductions, savings are unlikely to occur to any significant degree in the short term. The pace of transition from print to digital journals is moving slowly, and only those publishers with a strong financial base will be likely to succeed in quickly providing on-line access. As noted above and in the section of this paper relating to publishers' cost structures, there is no clearly viable economic path to move from print to digital publishing. Moreover, because user acceptance of digital journals may not occur rapidly and because of the many uncertainties about archiving digital information, libraries will need to

maintain print collections-historical and prospective-into the foreseeable future, requiring that investments in operations and facilities be maintained.

Interlibrary borrowing and lending is a growing cost within research libraries,^[15] and its rate of increase promises to escalate as the inflation-generated rate of serials cancellations escalates. According to the ARL, faculty and students borrowed more than twice as many items through interlibrary loan in 1996 as they did in 1986. The University of California Libraries recently reported an annual increase approaching 10% per year. Interlibrary services are labor-intensive operations; in 1993, the ARL conducted a cost study that determined the average cost of a borrowing transaction to be $18.62 and that of a lending transaction to be $10.93. The average ARL university library processed 17,804 interlibrary borrowing transactions and 33,397 interlibrary lending transactions during 1995-96, incurring an average annual cost of approximately $700,000. Given the rate of rise of interlibrary resource sharing transactions as well as the rate of rise of labor costs, research libraries are likely to experience increasing interlibrary borrowing and lending costs of about 10% per year. The rate of rise of interlibrary lending costs could be reduced through use of on-line journals rather than print journals; but if traditional print-based fair use practices are abrogated in the on-line environment, publishers might create pay-per-view contracts that would actually increase costs beyond the costs of manual interlibrary loans. Thus there are unknown cost implications in interlibrary resource sharing of digital information.

Capital Costs

Capital assets in libraries are of three basic types: buildings, collections, and equipment. Expenditures for the most costly of these assets, buildings, are typically not a part of library budgets and therefore are not generally considered by librarians in their discussions of library costs. This paper will not attempt to discuss capital costs for library buildings in any depth except to cite several relevant studies. In the late 1980s Cooper estimated the construction cost per volume housed in an on-campus open-stack library to range from $4.33 for compact shelving to $15.84 for traditional open stacks; he calculated the construction cost per volume of a remote regional storage facility to be $2.78.^[16] Bowen uses Cooper's construction costs, adjusted for inflation, and Malcolm Getz's life cycle estimates to calculate an annual storage cost of $3.07 per volume.^[17] Lemberg's research substantiates Bowen's premises regarding the capital cost that might be avoided through digitization of high-use materials.^[18] He demonstrates that, even considering the capital costs of technology necessary to realize a digital document access system, research libraries as a system could accrue substantial savings over time if documents are stored and delivered electronically rather than in print form.

Extrapolating from Bowen's estimate of an annual storage cost of $3.07 per volume, a research library subscribing to 50,000 journal titles per year, each of which constitutes one volume, accrues $153,000 in new storage costs each year.

Over 10 years the cumulative cost to house the volumes received through the 50,000 subscriptions would exceed $8 million.

The growing dependence on information technologies to deliver scholarly information requires that universities make new investments in capital equipment and allocate recurring operations resources to the maintenance of that equipment and the network infrastructure. Although universities have invested heavily in network technologies, the true costs are still inadequately understood, and it is clear that increasing dependence on digital, rather than print, scholarly information will require that reliable funding for technology be developed. While capital costs for print libraries entail buildings, whose construction costs fall within known ranges and whose life cycle is long, and collections, the long-term costs of which can be rather reliably estimated, capital costs for the digital library are distributed across the campus and indeed the world. As yet, there is no clear formula to indicate how much initial capital investment in technology might be required to deliver a given number of digital documents to a given size academic community. Moreover, the life cycle for capital assets relating to delivery of digital library content is typically very short, perhaps shorter than five years. Thus funding allocations must be made frequently and regularly to ensure continued access to digital information. The Berkeley library, for example, estimates that annual equipment replacement costs for the existing installed base of equipment would be approximately $650,000, assuming a five-year life cycle. But the library has never had an explicit budget to support that expense, so investments in computer equipment, networking, and equipment replacement have been made through periodic redirection of the operating budget. Similar technology funding and renewal problems exist across the campus. Berkeley's situation is not unusual, and further work needs to be done to understand more fully the capital cost differentials between the physical plant investments required for print collections and the network investments required to make digital information available to the campus community.M

It is possible that if libraries and their parent institutions, universities, could avoid some of the capital and operations costs associated with print-based dissemination of scholarly publications, these resources could be reallocated to capital investments in technology, provision of additional information resources available to the academic community, service improvements within libraries, and restoration of control of the system of scholarly publishing to universities and scholarly societies rather than the commercial sector.

The Economics of Electronic Publishing: A View from the University of California Press

The market realities described in the first portion of this paper are sobering, but the basic outlines have been well known to libraries and scholarly publishers for more than a decade. This section discusses the realities for nonprofit journal pub-

lishers (university presses and scholarly societies) as a way of answering the question "So why don't publishers just reduce their prices-at least for electronic publications?" Although the focus is on nonprofit presses, the basic economics are equally true for commercial publishers, except that they require profits and have the considerable advantage of greater access to capital to fund innovation.

The largest constraint on all publishers' ability to radically change the price structure for electronic publications is the first-copy costs, which commonly range from 70% to 85% of the print price (see Table 18.1 for an example of first-copy costs for University of California Press journals).

These first-copy costs will remain whether the format is electronic, paper, or both. Any pricing model must provide sufficient income to cover these costs in addition to the unique costs associated with publishing in any particular medium. Publishers are not wedded to maintaining print revenues per se but to maintaining enough revenues to cover their first-copy and unique-format costs and to cover the costs of the technological shift. In the transition period, when both print and electronic editions must be produced, this objective will inevitably result in prices that are higher than print-only prices. Whether wholly electronic publications are, in the long run, more economical will depend on the costs of producing a uniquely electronic product and on the size of the market. If substantially fewer libraries subscribe to electronic publications than subscribed to their print predecessors, the cost per subscription will inevitably increase in order to cover a larger share of first-copy costs.^[19]

Electronic Pricing Models

There are a number of models for pricing electronic resources. But all of them boil down to various ways of obtaining revenue to cover the same set of costs. They all ultimately depend on the same formula of first-copy costs plus print costs plus electronic costs. Table 18.2 shows humanities journal x.

Electronic Access Provided "Free" Publishers that are providing electronic access "free" with print subscriptions are, in fact, subsidizing the costs of the electronic edition out of the surplus revenues generated by the print publication; the print publication already covers the first-copy costs allocated to each subscription. For relatively high-priced scientific journals with high first-copy costs, this subsidization can be done without inflating the price too substantially; the uniquely electronic costs are then subsidized by all institutional subscribers and hidden as a percentage of the total cost of publication. Because the basic subscription price is high enough, relatively modest additional increases will also cover the cost of lost individual subscriptions (since individual subscriptions typically cover the run-on costs of producing additional issues but make only a partial contribution to first-copy costs). This approach has the added advantage of sidestepping for now the problems of negotiating prices and guarantees with libraries (and the associated

overhead costs). However, it does not contribute to developing commonly understood and agreed upon cost recovery models that will cover the costs of electronic scholarly communication in the long run.

Extra Charge for Electronic Access, Bundled with Paper An electronic edition that is provided with a print subscription for an extra charge is essentially the same as the previous cost recovery model, but the increase to cover electronic costs is made explicit. This approach may be especially necessary for journals whose base rate is low and whose markup for electronic costs cannot be covered by a typical inflationary increase. This model still has the advantage, for publishers, of spreading the cost over all institutional subscribers and of simplifying licensing negotiations.

Negotiated Price by Library Based on Paper Subscription Base Some publishers take the basic institutional print subscription base and guarantee this revenue for a period of years (typically three). Publishers are willing to guarantee limits to inflationary increases for this period in exchange for the guaranteed income and protection from cancellations to help cover transition costs. Again, this approach works better with higher priced journals for which the added costs of electronic publishing are a smaller proportion of the total cost.

Separate Price and Availability for Electronic and Paper, with an Incentive for Bundling Offering paper and electronic editions separately but with an incentive for bundling is the method deployed by SCAN and by Project MUSE. This model offers more flexibility to libraries, because libraries are allowed to cancel print and take only electronic or to select among the publications offered. Discount incen-

tives encourage maintaining paper and electronic subscriptions (a strategy used by both projects) and ordering larger groups of journals (the entire list for MUSE; discipline clusters for SCAN). The advantage to this approach is that the costs of electronic publishing are made clear. (See the revenues section below for a discussion of the adequacy of this model for supporting humanities publishing in the long run and of the impact of consortia discounts.)

In all these models, the ultimate economic effect in the transition period is the same: costs for libraries go up. Publishers must cover their first-copy costs; continue to provide paper editions for individuals, many libraries, and international markets; and generate revenue to cover the infrastructure and overhead costs of electronic innovation. For nonprofit publishers, at least, these costs must all be supported by the revenues from current journal subscriptions.

Electronic Costs

It is likely, in the long run, that eliminating print editions entirely will reduce costs somewhat for some kinds of journals. However, for journals that are trying fully to exploit the new capabilities offered by electronic technologies, it seems likely that the additional costs of generating links, specialized formats, and so on will continue to cost as much, or nearly as much, as the cost of printing and binding.^[20] But even for simpler humanities journals, the experience at the University of California Press raises questions about the assumption that ongoing electronic costs will be substantially lower.

Covering Costs of Development The University of California Press's original economic model assumed that the development costs were largely one-time expenses and that there was a single learning curve and set of skills to master, after which electronic publishing would be largely routinized; additional expenses would be easily absorbed by the margin generated by the savings in the paper edition. On the basis of the past three years, it seems apparent that this assumption was flawed. UC Press dedicated 3,500 staff hours on the SCAN project in 1994 (gopher site development); 4,100 hours in 1995 (WWW site development); and 3,700 hours in 1996 (largely on WWW development and on laying the groundwork for SGML implementation). It is apparent from ongoing trends in technological innovation that Internet technology and expectations for electronic publishing will continue to evolve very rapidly for at least the next 20 years. The Press's "bad luck" in initially developing for an outmoded platform (gopher) is an inevitable occurrence over the long term for electronic publishing projects. As a result, it seems foolhardy to assume that substantially less investment will be necessary for technical research, experimentation, and site redesign and revision in the future. Any viable economic model for the University of California Press must thus assume one or two technical FTE positions as part of ongoing overhead. (Please note that these positions will not include file server maintenance and enhancement, since the

costs of file service for the SCAN project are presently borne by University of California/Berkeley Library.)

The SCAN project has experienced ongoing instability in technical staff at the library and at UC Press. Being located in a region with such a strong hightechnology industry has actually proven to be a disadvantage, since current and potential employees can make so much more money at other jobs. This situation results in long staff vacancies and repeated training on the specifics of the project. In this way, the project again faces not one but rather a continual series of learning curves.

There is a third implication to this vision of a constantly evolving future. Functionality and continually changing platforms, combined with the Press's commitment to archiving and to long-term responsibility for viable electronic access, demand implementation of a coding system that is totally portable and highly functional. As a result, the commitment to SGML seems more and more wise as time goes on. This commitment leads the Press to reject image-based solutions like Acrobat, which would require less work and which would be faster to implement but which do not have long-term migration paths. Having experienced the painful process of completely receding individual files, the Press does not want to face the same problem with a much larger set of files in the future. The necessity and the difficulty of repeated conversions of legacy text is sadly underestimated by many publishers and librarians. Scaleability, an important and underrated issue in any case, becomes even more vital in a scenario in which larger and larger amounts of material must be converted each time the technological environment evolves.

Electronic publishing is adding new duties (and requiring new resources) within the Press, without removing present duties. For example, the Press has added .5 FTE in the journals production staff (a 25% increase) to handle liaison with suppliers, scanning and archiving of all images being published, archiving of electronic files, and routine file conversion duties. This position will clearly grow into a full-time position as all the journals are mounted on-line; only the slowness of the on-line implementation permits the luxury of this low staffing level. The seven people working on Project MUSE or the seven people working on The Astrophysical Journal Electronic Edition confirm this assumption. In addition, clearing electronic rights for images in already published books and journals and maintaining an ongoing rights database creates a new staff responsibility, since many rights holders are requiring renewal of rights and payments every five to ten years. The need for technical customer support is still unknown but surely represents some portion of an FTE.

Marketing is another area requiring addition of new expertise and staff. Successfully selling electronic product requires a series of changes within the publishing organization. The marketing necessary to launch a new print journal successfully or to sell a book is expensive and time-consuming, but the approaches and tasks are familiar and can be performed by existing marketing staff as part of their

existing marketing jobs. In contrast, successfully establishing a customer base of licensed libraries for electronic product requires new skills and abilities, a substantial staff commitment, a higher level of staff expertise and authority, and substantial involvement from the licensing libraries. Marketing electronic services requires all the brochures and ads that print publications do. In addition, it requires substantial publicity efforts, a travel schedule to perform demonstrations at a wide range of library and end user meetings, and participation in appropriate LISTSERVs. There must be at least one staff member who has the requisite knowledge and authority and who can dedicate a large portion of time to outreach, negotiations, and liaison with potential and actual license customers and subscription agents. There are also demands for ongoing customer relations work, including the provision of quarterly or annual use reporting. The Press has found it very difficult to fit those new functions into its traditional marketing and distribution job descriptions and workloads. As the Press moves more seriously into electronic publication of current season books, it will surely need to hire a new person to market on-line books; these functions cannot possibly be integrated into the already busy jobs of books marketing professionals with their focus on current season bookstore sales.

In short, the Press anticipates a permanent addition of at least three or four full-time staff to the overhead of the publishing operation. For now, some of these positions are covered by the Mellon Foundation grant, and some of them have been deferred (to the detriment of the project), but in the long run the electronic publishing model must absorb this additional $200,000 in annual costs.

Finally, UC Press and the UC Library have just begun to step up to the costs of long-term archiving, including periodic refreshing of technology and the requisite reconversion of files-another argument for structured standardized coding of text.

Income for Electronic Product

Unfortunately, in a period when electronic publishing generates additional costs that must be funded, several trends apparent in the emerging purchase patterns of electronic products limit the income available to support publication costs and create further pressures on publishers to increase prices.

Slowness to Adopt University presses attempting to sell electronic product directly (as opposed to bundling it automatically in the paper price and offering "free" access) are finding that sales to universities are progressing more slowly than projected. Project MUSE sales, for example, are at 378 after two years; sales to MIT's electronic-only journals hover at around 100; in no case are there more than 50 library subscriptions. There are under 25 subscriptions to the online edition of The Cigarette Papers at the University of California/San Francisco Library's Brown and Williamson Tobacco site after nine months (http://www.library.ucsf.edu/tobacco/cigpapers/ ). Sales to SCAN are a handful (although access

has been restricted for less than one month at the time this paper is written). Even for publications for which no additional charge is being made, library adoptions are slow in coming. The Astrophysical Journal Electronic Edition, for example, has 130 libraries licensed to date. There are, of course, good reasons for this slowness: libraries face the same difficulties in building infrastructure, funding, and staff expertise that publishers do. But the low sales nevertheless make funding the transition more difficult, because publishers can't count on sales income from the electronic product to help to cover the costs of electronic publication. The growth curves to which publishers are accustomed from launching paper journals (even in this age of low library adoptions) are too optimistic when applied to electronic publications. This slowness has real consequences for funding electronic innovation.

New Discount Structures Emerging business practices and discount expectations lessen the income per subscribing institution (at the same time that the efforts necessary to obtain that subscription are intensified). The expectations of consortia for deep discounting (both for number of consortia members and for adopting a bundle of publications) can go as high as 40% for academic institutions, with nontraditional markets receiving even deeper discounts. If 70-85% of the list price represents first-copy costs, a 40% discount means that these subscriptions are no longer carrying their full share of the first-copy costs. Deep discounting cannot be a long-term pricing strategy.

In addition, other consortial demands (for example, demands that inflationary increases not exceed a certain percentage for several years or that access be provided to high schools free of charge) further lessen the ability of publishers to fund electronic innovation out of electronic product sales. Again, it is easy to empathize with these library positions and to understand why they are evolving. But these efforts by libraries to control costs actually have an inflationary pressure on overall prices, since the base price must increase to make up the losses.

Loss of Subscriptions Publishers are also worried about losing subscriptions. Some losses will surely happen. At a minimum, subscriptions will be reduced by another major wave (or waves) of cancellations as libraries try to cope with the ongoing costs of paper and electronic subscriptions from the major commercial science publishers and by the loss of any duplicate subscriptions still remaining on campuses. In addition, publishers are haunted by the potential for substantial shrinkage of individual subscriptions or society memberships as more and more scholars have "free" access from their campuses, though loss of individual subscriptions is less sure than library cancellations. (By December 1996, almost 60% of SCAN uses were coming from U.S. non-.edu addresses as more and more people obtain access from home workstations; it is possible that individuals will pay for the convenience of noncampus access, just as they now do for nonlibrary print access.) Nevertheless, because individual subscriptions play an increasingly important role in financing many journals (especially journals launched within the past

ten years, when library support has been so eroded), widespread cancellation would have a substantial impact that would force journal prices higher.

Possible Increases in Sales Two possible new revenue sources may somewhat balance the losses in income described above, although both are highly speculative at this point. First, publishers may obtain new institutional markets and wider distribution as consortia bring institutions like junior colleges and high schools to scholarly publications. Project MUSE has begun to see this trend. It is not clear, however, that these customers will be long-term subscribers. Given the present nature of scholarship, many of these new subscribers may conclude that any amount of money is too much to pay after two or three years of low use statistics, especially when on-demand access by article becomes widely available. There will be a substantial market for scholarship at junior college, high school, and public libraries only when the possibility of wider audiences through the Internet fundamentally changes the ways in which scholars write and present their work-a change that will surely take many years to materialize. Other publishers are more optimistic about this potential source of income.

Second, a substantial revenue stream may exist in sale of individual chapters and articles to scholars whose institutions do not have access, who do not have an institutional base, or who are willing to pay a few dollars for the convenience of immediate access at their workstations (people who are now presumably asking their research assistants to make photocopies in the stacks). And there may be substantial sales among the general public. This new product may represent enough income to relieve some of the pressure on journal finances, if the process can be entirely automated (at $6 or $7 per article, there is no room for the cost of an employee ever touching the transaction). This solution needs substantial traffic, because it takes seven or eight article sales to cover the first-copy costs of one typical humanities subscription.

Of course, the ability to purchase single chapters or articles will also diminish subscription revenues, as some libraries choose to fill user needs on demand and to cancel their present subscriptions. It is too soon to tell what the mix of new audiences and subscription cancellations will be, and whether the revenue stream from new sources will replace that from canceled subscriptions.

Aggregators So far, the models we have examined have all assumed that the publisher is providing access to electronic resources. Publishers could, of course, avoid many of these costs by providing electronic files to aggregators and leaving marketing, file service, file conversion, and archiving to outside suppliers who would provide a single point of purchase for libraries and individuals. This scheme offers a number of advantages from a library point of view. The instant connection between search engine and ordering ability that the larger services like UnCover and OCLC offer may potentially bring more end users.

But from a publishing point of view, this model has two very large disadvantages. The first is strategic. In an electronic world, one of the major values that

publishers have to offer is the branding value of our imprints as symbols of excellence resulting from peer review and gatekeeping functions, which will be ever more valuable in the time-starved world of the Internet. This brand identity is inevitably diluted in an aggregated world, especially if the aggregator is handling marketing and distribution.

Second, and more relevant to the discussion at hand, it is hard to see how the royalties most typically offered by aggregators (for institutional licenses or for ondemand use) can begin to replace the revenue lost from direct subscriptions. A 30-40% royalty does not cover first-copy costs of 80%. Only by retaining the entire fee can publishers hope to generate enough revenue for on-demand sales to make a sufficient contribution to the costs of publication. A wide-scale move to aggregation would have the effect of making the first-copy costs for the few remaining subscriptions very large indeed, in addition to reducing the perceived value of what we sell (yes, it is possible for a humanities quarterly to cost $1,200 annually!).

The University of California Press and most other nonprofit scholarly publishers would like nothing better than to price electronic products substantially lower than print. However, the low margins under which they operate, the demands of users that print continue to be provided, the high first-copy costs typical of scholarly publishing, the need to fund the development of electronic product, and the expenses of producing full-featured electronic publications all mitigate against low prices, at least during the transition period.

Conclusion

The university press and the library face economic pressures that neither can address alone. So long as journal prices escalate more rapidly than library collection budgets, libraries will continue to reduce serial subscriptions to balance the collections budget. These reductions will adversely affect the revenues to university presses. Pressure from science, technology, medicine, and business faculties to retain high-cost, high-use journals will crowd out less-used scholarly journals, many of which are published by university presses. Because libraries must continue to provide access to and preserve print inventories, housing them in large physical plants that must be maintained, they will be unable to implement large-scale, costreducing changes in operations to free up resources for investments in technology. The trends noted in University Libraries and Scholarly Communication and in Hawkins's paper will result in a catastrophic decline in the system of scholarly communication unless there is a fundamental shift in the way in which its processes, products, and costs are analyzed. Each of the two partners, the library and the press, serves as an inadequate unit of analysis for the system of scholarly communication as a whole.

Sandra Braman's description of the three stages in the conceptualization of the information society provides a useful context in which to view today's problems of

press and library within the system of scholarly communication.^[21] In her conceptualization, the first stage of the information economy is recognized by the increasing importance of information sector industries. In the second stage, certain forms of information never before recognized as commodities, become so recognized. In this stage, political controversy about information's value as a public good versus its market value as a commodity is highlighted. The rising commercialization of scholarly publishing and the declining ability of libraries to provide access to scholarly information may be interpreted as a second-stage information society phenomenon.

Braman postulates that the third stage of the information society produces a more sophisticated understanding of the flow of information: the flow may replace the market as the primary feature of the information economy. This stage represents a paradigm shift in which the information economy operates in a qualitatively different manner than in the two previous stages. According to Braman: "key insights of this perspective include identification of a new unit of analysis, the project, involving multiple interdependent organizations, as more useful than either the industry or the firm for analytical purposes"(p. 112). She further describes the third-stage conceptualization of the information economy as including a production chain, or "harmonized production flows," including information creation, processing, storage, transportation, distribution, destruction, seeking, and use, in short, all the stages of the system of scholarly communication from author to user, including the library. In the third stage, networked information economy, economic viability stems not from maximizing profit or economic stability within each component of the system, but rather through building long-term relationships and a stable system or flow of information.

Michael Hammer makes a similar point with respect to industrial or business reengineering but applicable to the operations of libraries and presses as well.^[22] He notes that automation and other reengineering efforts frequently have not yielded the improvements that companies desire. He believes that heavy investments in information technology deliver disappointing results because the technology is used primarily to speed up traditional business processes, along with their inefficiencies, rather than to transform them. "Instead of embedding outdated processes in silicon and software, we should obliterate them and start over. We should ... use the power of modern information technology to radically redesign our business processes in order to achieve dramatic improvements in their performance" (p. 104).

Both Braman and Hammer emphasize the disquieting qualities that characterize this kind of paradigm shift implied by the third stage of the information economy and by radical reengineering. According to Hammer,

Reengineering cannot be planned meticulously and accomplished in small and cautious steps. It's an all-or-nothing proposition with an uncertain result.... At the heart of reengineering is the notion of discontinuous thinking-of recognizing and

breaking away from the outdated rules and fundamental assumptions that underlie operations. Unless we change these rules, we are merely rearranging the deck chairs on the Titanic. We cannot achieve breakthroughs in performance by cutting fat or automating existing processes. Rather, we must challenge old assumptions and shed the old rules that made the business under perform in the first place ... Reengineering requires looking at the fundamental processes of the business from a cross-functional perspective.

Manuel Castells takes a different approach, suggesting that technology-driven productivity increases in the informational economy have not thus far been evident. His thesis is that technology-driven productivity increases were steady in the industrial sector between 1950 and 1973, but since 1973 productivity, particularly in the service sector, has stagnated despite the intensive investment in technology. He suggests three factors that appear to be relevant to the library and press sector as well as to the service sectors of the economy in general. These factors include the following.

1. Diffusion: before technological innovation can improve productivity markedly, it must have permeated the whole economy, including business, culture, and institutions.

2. Measuring productivity: Service industries traditionally find it difficult to calculate productivity statistically; thus the lack of observable productivity enhancements may in part be a symptom of the absence of relevant measures.

3. The changing informational economy: Productivity cannot easily be measured because of the broad scope of its transformation under the impact of information technology and related organizational change.

If Castells, Braman, and Hammer are correct, then libraries and presses, alone or together, cannot implement technological solutions that can transform the processes, productivity, and economics of scholarly publishing.

The Mellon projects have been useful in introducing two players in the information flow to the problems of the other, and in forging collaborative relationships to aid in sustaining the system of scholarly communication. These cooperative projects between university libraries and presses have helped participants begin to understand the system of scholarly publishing as an information flow rather than as separate operational processes. But their effectiveness is limited because, outside the parameters of the projects, the partners must still maintain their separate identities and economic bases.

A fuller exploration of the potential of transforming the flow of scholarly information would incorporate a more integrated approach, including the creators of the information, the university administration, and the information consumers as well as the publisher and the library. In this approach, costs and subsidies of the entire process of scholarly communication could be better understood and resources made more flexibly available to support it. For example, it might be possible to view operational and capital savings to libraries resulting from a transition

to electronic publication as resources ultimately available to sustain the publication chain, or consumers could be asked to pay some or all of the costs of creating, storing, archiving, and delivering scholarly information. A critical flaw in the current system is the existence of a part of the gift economy, in the form of the library, within a monetary economy for commercial publishers. Because the consumers of the information within the university do not pay for it, they and the campus administration see the library as a "problem" when it cannot provide the information needed within the budget allotted.

A key problem in securing the future of scholarly communication is that both presses and libraries are undercapitalized. Although libraries incur huge capital costs over time in both inventory and facilities, they are not free individually nor as parts of the system of scholarly communication to reallocate present or future capital expenditures to investments in new modes of publication. However, such reallocation, if it occurs at all, will take place very slowly because the transition to digital publication will also be slow. It is possible that a more rapid transition to electronic publishing would reduce libraries' recurring operations costs, thereby enabling them to invest greater resources in information itself. But a more rapid transition is feasible for presses only if there is a rise in demand for digital publications from libraries and from end users or a substantial increase in subsidies from their parent universities. Presses can offer electronic publications, but they cannot change the demand patterns of their customers-libraries-nor the usage patterns of the end consumers in order to hasten a transition from print to electronic dissemination. As long as a substantial portion of their market demands print (or fails to purchase electronic product), presses will be forced to incur the resulting expenses, which, in being passed on to libraries as costs that inflate more rapidly than budgets, will reduce the purchases of scholarly publications.

Ironically, in the present environment, universities tend to take budgetary actions that worsen the economics of scholarly communication as experienced by both libraries and presses. University administrators increasingly interpret any subsidy of university presses as a failure of the press itself as a business; as university subsidies are withdrawn, presses must increase prices, which reduces demand and exacerbates the worsening fiscal situation for the presses. But in the networked economy where everyone can be an author and publisher, the value added by presses (for example, gatekeeping, editorial enhancement, distribution) may be more important than ever in helping consumers select relevant, high-quality information. At the same time, university administrators see the library as a black hole whose costs steadily rise faster than general inflation.^[23] Since library materials budgets grow more slowly than inflation in the costs of scholarly publications, the inevitable result is reduced purchasing of scholarly publications of all types, but particularly of university press materials, which in general are of lesser commercial value in the commodity market. Unless the system as a whole changes, both university presses and university libraries will continue to decline, but at accelerated rates.

Although it is not possible to envision with certainty exactly how a successful transition from the present system to a more sustainable system might occur, one plausible scenario would be for universities themselves to invest capital resources more heavily in university-based information flows and new forms of scholarly publication as well as place increased market pressures on the commercial sector. If universities were to make strategic capital and staffing investments in university presses during the short term, the presses could be more likely to make a successful and rapid transition to electronic publication. At the same time, intensive university efforts (i.e., investments) to recover scientific, technical, medical, and business publishing from the private sector could be made to reduce the crowding out of university press publications by for-profit publishers. These efforts to recover scholarly publishing could be accompanied by libraries' placing strong market pressures on commercial publishers through cancellation of journals whose prices rise faster than the average rates for scholarly journals in general. The investments in these two areas: converting publication processes to electronic form and returning commercial scholarly publishing to the university could be recovered over time through reductions in capital investments in library buildings. Ultimately, the university itself would encompass most of the information flow in scholarly communication through its networked capability. That information having commodity value outside the academy could be sold in the marketplace and the revenues used as a subsidy to the system itself.

Another way of accomplishing a harmonization of the scholarly information economy was suggested by Hawkins:^[24] the independent nonprofit corporation model in which universities and colleges would invest together in a new organization that would serve as a broker, negotiator, service provider, and focus for philanthropy. It would leverage individual resources by creating a common investment pool.

However the solution to the problem of the economic crisis in scholarly communication is approached, there must be a fundamental change in how the process as a whole is conceived and how intellectual property rights of both authors and universities are managed. Such a change cannot be made unilaterally by university libraries and presses but will require the strategic involvement and commitment of university administrators and faculty within the university and among universities. Patricia Battin, envisioning an integrated scholarly information flow, said almost ten years ago:

Commitment to new cooperative interinstitutional mechanisms for sharing infrastructure costs-such as networks, print collections, and database development and access-in the recognition that continuing to view information technologies and services as a bargaining chip in the competition for students and faculty is, in the end, a counterproductive strategy for higher education. If the scholarly world is to maintain control of and access to its knowledge, both new and old, new cooperative ventures must be organized for the management of knowledge itself, rather than the ownership of formats.^[25]