The Inefficient Market: Contradiction between Social and Economic Logic
Human knowledge appears to function on the same probability structure as the rest of living nature and society. It is a structure dominated by highly skewed, exponential distributions that arise from the operation of two basic multiplicative, often interactive stochastic processes: inhomogeneity and contagion. Because all probabilities must add to one, the result is essentially a zero sum game in which the success of the few necessitates the failure of the many. The library market for ST journals is characterized by an imbalance between two such distributions, one for ST value and the other for cost.
Two principles dominate the library market for ST journals. The first is societal and is exemplified by the U.S. scientific and technical associations. Associations have long played an important role in American society, and their importance was noted by none other than Alexis de Tocqueville (1969, 189-95, 513-25), who in his famous book Democracy In America called particular attention to the significance of American intellectual and moral associations. The importance of U.S. scientific and technical associations in the library market for ST serials is that they are the primary social organizations of those involved in science and technology, and it is largely through the medium of their publications that the ST elite communicate the results of their work.
In a landmark article Cole and Cole (1972) analyzed the comparative contributions of the various social strata of scientists to scientific progress, using the work of Jose Ortega y Gasset to set the null hypothesis that science was advanced by the work of many average scientists making minor contributions. Cole and Cole rejected this hypothesis. To the contrary, they found that scientific progress is mainly the work of a small elite concentrated at a relatively few institutions and that even the minor discoveries come primarily from the top strata of the scientific community.
Using Price's modification of Lotka's law, Cole and Cole estimated that roughly 50% of all scientific papers are produced by approximately 10% of the scientists, and they analyzed the extent to which the 10% of the scientists who produce 50% of the research depend on the other 90% of the scientists responsible for the remaining 50%. To do this, Cole and Cole examined the references made by 84 university physicists in their most highly cited papers in the 1965 SCI to a random sample of 385 physicists drawn from those referenced in these 84 papers. Of the 385 cited physicists, 72% were affiliated with universities, and 60% of these were located at the nine physics departments given the highest rating of "Distinguished" in the Cartter 1964 assessment of quality in U.S. graduate education. The nine top departments represented 10.5% of the 86 physics programs rated that year, and at the risk of redundancy their names in descending rank order were California at Berkeley, Cal Tech, Harvard, Princeton, Stanford, MIT, Columbia, Illinois at Urbana–Champaign, and Cornell.
In general, Cole and Cole found among those highly cited many who were members of the National Academy of Sciences and winners of the Nobel prize as well as other awards, and there was a tendency for the elite to cite the elite and to be cited by those lower down the science stratification system. It is this elite that lies at the basis of ST value, which they transfer to U.S. association serials by publishing their most important work in them. Because of the nature of the probability structure governing science and technology, it is the elite's success that reduces the success of others, causing zero classes in citations and library use.
The other principle dominating the library market for ST serials is commercial, and this principle is epitomized by the commercial publishers, both domestic and foreign. It would be easy to end this paper by praising the societal principle and denouncing the profit motive, but unfortunately the picture appears to be much more complex. The separation of ST value from cost may be mainly the result of the interaction of the science and technology social stratification system with the economics of journal publishing rather than the profit motive.
The conceptual model proposed by Noll and Steinmueller (1992) for the analysis of scientific journal prices is extremely interesting from this perspective. According to these authors, scholarly journals have essentially the same cost structure as any other media product in that this cost structure consists of two parts: (1) "first copy cost," which includes all activity associated with producing the basic information that the product contains; and (2) the costs of actually printing and distributing the publication, which for scholarly journals are a relatively small part of the total costs. Given this cost structure, Noll and Steinmueller point out that circulation is a key variable and that prices must be higher for journals with low circulation because the fixed first copy cost must be recovered from a smaller number of subscribers. They further posit that a fundamental factor in determining the demand for academic journals is the need of the faculty to publish for promotion and salary purposes.
In the view of Noll and Steinmueller, scholarly journals exist within social hierarchies, and as more faculty seek publication outlets, it becomes increasingly difficult to be published in the "best" journals. The result is that a smaller and smaller proportion of scholars succeed in publishing at the top of the hierarchy. In their model, both publishers and scholars respond to this situation by seeking to create new publishing outlets that form new hierarchies by narrowing the scope of journals, because it is better to be at the top of a small, new hierarchy than at the bottom of an established, large one. Noll and Steinmueller state that the type of market that develops in this regime is one of "monopolistic competition," where each producer supplies a somewhat different product from all others, attempting to serve a relatively small subset of the overall market. However, they point out that as more and more such journals are launched, the subscription base of all the competing serials is reduced, forcing price rises to cover the high fixed (first copy) costs of serials. Noll and Steinmueller then state (p. 34):
In their opinion, the performance of the journals market under such conditions is "socially undesirable and economically inefficient" (p. 35) for the following reasons: (1) secondary journals tend to publish many articles that are not particularly important and that could not sustain sufficient circulation to keep a journal viable; and (2) as average circulation declines and prices go up to cover first copy costs, a wider spread develops between the marginal cost of a journal and its price. The situation is loaded with social dynamite with possible catastrophic consequences, for in a model developed by Quandt (1996), if in the presence of myopic behavior publishers try to maintain their revenue by raising prices to compensate for falling subscriptions, and libraries cancel journals on the basis of cost and importance, there could be a dramatic implosion in subscriptions, ranging from 13% to 96%.
Credence was given to Noll and Steinmueller's model by Bensman's (1996) analysis of the library market for chemistry journals, using the database compiled as a result of the SRP pilot project in 1993. Regression tests with this database had shown that for every type of publisher—association and commercial, domestic and foreign—serials prices tended to go down in line with the increase in the number of libraries holding them. Taking all types of publishers together, prices decreased in terms of standardized regression coefficients either 0.40 or 0.47 standard deviations for every one standard deviation increase in holding libraries, depending on which measure of scientific value was utilized. U.S. publishers' prices were much more sensitive to this variable than foreign ones, but this could have been an artifact of the holding libraries being predominantly located in the U.S. For the present analysis, the database was supplemented with the use data contemporaneously gathered at the University of Illinois at Urbana–Champaign (UIUC) Chemistry Library.
The analysis was done in two stages. First, the reasons why libraries purchase and hold serials were investigated by regressing the OCLC library holdings of the chemistry journals on their price and scientific value measured sequentially in terms of LSU faculty ratings, total ISI citations, and total UIUC Chemistry Library use. Preliminary analyses for setting up the regression equations uncovered outliers of three basic types (besides the consistent appearance of one title as such for unexplainable reasons).
One type of outlier was subject-related and could have been the result of the set of journals under analysis being defined by a survey of the Department of Chemistry without taking into account the Department of Biochemistry. This type occurred when LSU faculty ratings and total ISI citations were utilized as the measures of scientific value, and it required the exclusion of the Annual Review of Biochemistry, Free Radical Research Communications, and the Journal of Biological Chemistry as outliers.
The second type of outlier related to all three measures being research oriented. Use in the UIUC Chemistry Library was so, because this library supports a large research-doctorate program that has been consistently ranked at the top of such programs in chemistry from 1910 through 1993. This characteristic led to the considerable underprediction of the number of libraries holding Chemical and Engineering News, the main informational publication of the American Chemical Society, and the Journal of Chemical Education, whose purpose is primarily instructional.
The final type of outlier included one title, and it was the most interesting, because it was due to an extreme imbalance between price and scientific value. This title was the Journal of the American Chemical Society, which had to be excluded as an outlier from the model that used LSU faculty score as the measure of scientific value as well as from the model that used total UIUC use in this role. The reason for this was that the Journal of the American Chemical Society was such a bargain in terms of price and scientific value that, instead of the 1,381 libraries actually holding it, the model with LSU faculty score predicted that it should be held by 4,429 libraries, and the one with total UIUC use, by 5,688—in both cases, probably more libraries that needed it than exist in reality.
All three models showed that scientific value plays a much more important role than price in determining why libraries purchase and hold serials. With respect to LSU faculty score, the standardized regression coefficients showed that for each move of one standard deviation upwards in scientific value, library holdings went up 0.64 standard deviations; for each move of one standard deviation upwards in price, library holdings went down 0.15 standard deviations. The model accounted for 39.7% of the variance in library holdings, and by itself LSU faculty score accounted for 39.5% of the variance beyond that caused by price, whereas by itself price caused only 2.3% of the variance beyond that caused by LSU faculty score.
The same picture was obtained when total ISI citations were used as the measure of scientific value. Here the standardized regression coefficients showed that for each move of one standard deviation upwards in scientific value, library holdings increased 0.61 standard deviations, whereas for each move upwards of one standard deviation in price, library holdings decreased 0.24 standard deviations. This model accounted for 33.2% of the variance in library holdings, and by itself total ISI citations accounted for 33.0% of the variance beyond that caused by price, whereas by itself price accounted for only 5.2% of the variance beyond that caused by total ISI citations.
These findings were corroborated by the model using total UIUC Chemistry Library use as the measure of scientific value. With this model, the standardized regression coefficients revealed that for each move of one standard deviation upwards in scientific value, library holdings rose 0.61 standard deviations, whereas for each move upwards of one standard deviation in price, library holdings fell 0.23 standard deviations. This model accounted for 33.9% of the variance in library holdings, and by itself total UIUC use accounted for 33.6% of the variance beyond that caused by price, whereas by itself price accounted for merely 4.7% of the variance beyond that caused by total UIUC use.
The next stage of the analysis was devoted to investigating the dichotomy between the serials of the U.S. associations on the one hand and those of the commercial publishers, both domestic and foreign, on the other. It was done in two steps. First, the means of these types of publishers were compared on the independent variables of the above models—price, LSU faculty ratings, total ISI citations, and total UIUC Chemistry Library use. In all cases, the serials of the U.S. associations represented far better purchases for libraries than those of the commercial publishers. In terms of price, the mean was $575.76 for the serials of the U.S. associations, $878.67 for titles of U.S. commercial publishers, and $1,565.24 for those of the foreign commercial ones. In terms of LSU faculty score, the mean was 189.7 for the titles of the U.S. associations, 53.6 for the serials of U.S. commercial publishers, and 72.5 for those of the foreign commercial publishers. In terms of total ISI citations, the mean was 30,073.8 for the serials of the U.S. associations, 6,541.8 for those of the U.S. commercial publishers, and 8,975.6 for the foreign commercial titles. In terms of total UIUC Chemistry Library use, the mean was 1,440.2 for the U.S. association serials, 175.0 for the U.S. commercial ones, and 391.2 for the foreign commercial titles. ISI impact factor was also tested as part of the investigation of the independent variables, and only in terms of mean ISI impact factor did the commercial serials do well against the U.S. association titles. However, interestingly enough, it was only the differences between the impact factor means that were not statistically significant, once again calling into question the validity of this measure.
In the second step, the dependent variable—library holdings—was examined, and not unexpectedly, the serials of the U.S. associations fared much better here than those of the commercial publishers. The mean library holdings were 743.0 for a U.S. association serial, 387.5 for a U.S. commercial one, and 316.7 for a foreign commercial title. The differences between these means were also statistically significant. Evidence exists of a positive relationship between library holdings and the scholarly value of serials in other subjects besides chemistry; for example, Wallace and Boyce (1989) found significant correlations ranging from 0.49 to 0.74 between OCLC holdings and total citations in business, clinical psychology, ecology, genetics and heredity, and international relations.
On the basis of the Noll and Steinmueller model, it is now possible to say that the logic of the social stratification system of science and technology works contrary to the economic logic of the ST journal market, enabling better ST information to be delivered at a cheaper price and forcing worse ST information to be delivered at a higher price. Even if the U.S. associations were operating on the principle of profit maximization, they would still be able to underprice the commercial publishers. However, for whatever the reason, the propensity of commercial publishers to charge more for their products is causing zero classes of unknown magnitude not only in library monographic purchases but perhaps even in the launching of good ST serials.
One of the major bases of modern investment theory is the "efficient market hypothesis" (EMH). According to this hypothesis, a stock market is efficient when share prices reflect all the information relevant to the companies underlying them (Bodie, Kane, and Marcus 1996, 338–41). From this perspective, the library market for ST journals is extremely inefficient. Libraries pay the highest prices for serials with the lowest ST value and the lowest probability of being used. The inefficiency of the library market for ST serials is demonstrated by the three regression models above, where from 60.3% to 66.8% of the variance in library holdings is left unexplained by the two variables—price and scientific value— which should be dominant in their determination.
A number of basic factors might play an important role in causing this inefficiency besides academic politics. First, academic libraries tend to compare themselves on the basis of the ARL Library Index. However, this index is a measure of the relative size of libraries, and it does not attempt to measure the quality of collections or success in meeting the needs of the users (Kyrillidou and Blixrud, 20 March 1997). The ARL Library Index is constructed from five data elements— number of volumes held, number of volumes added, number of current serials received, total operating expenditures, and number of staff—and, insofar as the number and cost of ST serials are factored into this index, it may be a measure of not which is the better library but of which is the greater fool.
Second, there persists a belief among librarians that publishers charge according to the quality of their serials, and this belief found expression in the following statement quoted by Dougherty and Barr (1988, 8):
On the contrary, Bensman (1996) used regression analyses of chemistry serials to demonstrate that the primary determinant in pricing by all types of publishers—association and commercial, domestic and foreign—is size and not quality, which plays virtually no role.
Third, many librarians think that the commercial publishers possess monopoly power (Stoller, Christopherson, and Miranda 1996), but this does not appear to be the case if so many of their serials can be canceled with extremely small losses in ST value. On the contrary, careful consideration of the evidence presented here might lead to the conclusion that many of the publishers—who do possess monopoly power due to the quality of their products—actually charge less for these products.
A fourth factor possibly causing inefficiency in the library market for ST serials is more subtle, and it is that librarians appear to feel a professional obligation to preserve all human knowledge, no matter its importance. In studies of serials cancellations by five ARL libraries in the Midwest Chrzastowski and Schmidt (1993; 1996) noted that libraries tended to cancel English-language science titles that had a higher-than-average subscription price and that they alone held. The authors attributed this phenomenon to the operation of the Matthew Effect, which was leading to the development of similar core collections of high-use and frequently cited serials titles. Chrzastowski and Schmidt regretted this loss of "diversity," calling for cooperative collection development to address the issue of cancellations. However, the progress of science and technology requires high levels of consensus with agreed- upon paradigms and established channels of communication. By their very uniqueness, the canceled titles demonstrated their unimportance, and the relatively high cost of these titles more than likely resulted from this lack of importance.
In a discussion of the EMH, Frankfurter and McGoun (1996, 60–61) specify three versions of market efficiency. In all three versions the library market for ST serials can be considered inefficient. The first is "informational efficiency." This has been defined above, and we have shown that the prices of ST journals certainly do not reflect the information that is available or easily obtained as to their underlying ST value. The second version is "allocational efficiency," and Frankfurter and McGoun state that "markets are allocationally efficient, if investment projects are financed at the marginal productivity of capital." However, due to the highly skewed distributions of ST value, the marginal utility of journals diminishes at a rapid rate, while at the same time the prices of these same journals are forced up by the need to recover first copy costs from the fewer number of subscriptions dictated by their lower ST value. If one considers the serials holdings of a library as part of its capital, then libraries are investing money at a point where the cost of journals far exceeds their productivity in terms of usable human knowledge, making the library market for ST serials allocationally inefficient. The third version is "operational efficiency." According to Frankfurter and McGoun, operational efficiency is "concerned with the ease and speed by which capital markets make the meeting of buyers and sellers possible," and this in turn depends upon the level of market liquidity. From this perspective, the library market for ST serials can be seen as operationally inefficient, because libraries have a large part of their resources tied up in fixed costs for serials of questionable ST value, reducing their liquidity and flexibility in meeting new informational needs.
Options for Resolving the Crisis
The crisis predicted by Price a generation ago is now here. New serials are being established at a rapid rate. For example, the 35th edition of Ulrich's for 1997 ( vol. 1, vii-viii) reports adding nearly 6,000 titles; the 36th edition for 1998 ( vol. 1, vii- viii) increased its coverage by about 7,000 titles, of which more than 3,571 were known to have begun publication since January 1, 1995.
Moreover, the U.S. Periodicals Price Index (USPPI) for 1998 (Alexander and Dingley 1998) shows that inflation continues unabated. Excluding Russian translations, the average price of a U.S. periodical rose 34.3% from $149.46 in 1995 to $200.74 in 1998. In addition, the 1998 USPPI reveals that the subject category Chemistry & Physics remains at the heart of the problem. The average price of the serials in this category increased 38.4% from $767.96 in 1995 to $1,062.49 in 1998. Once again, with the exclusion of Russian translations, the average price of Chemistry & Physics serials was the highest, being 5.3 times higher than the average price of all 1998 USPPI serials of $200.74 and twice as high as the second-most expensive subject category, Medicine, which averaged $524.65. With the elimination of Russian translations, the Chemistry & Physics subject category accounted for only 4.6% of the 3,729 titles in the 1998 USPPI sample but 24.1% of the $748,559.46 that this sample cost.
Three basic factors appear to be driving the inflationary spiral in the prices of ST serials. First, there is the normal exponential growth of science and technology, which manifests itself in the increasing number and size of ST serials. As Price pointed out, this growth outstrips that of the socioeconomic institutions supporting science and technology. It alone could cause the prices of ST serials to rise faster than other prices in society.
Second, as the Noll and Steinmueller model indicates, the problem of normal ST growth is probably compounded by the dysfunctions introduced by the university promotion and tenure system. Publishers are serving the need of the faculty to publish by providing ever-more specialized outlets, and the constant shrinkage of the markets for such outlets makes it ever-more difficult to cover first copy costs. The result is the production of much ST literature whose cost of production cannot be justified by its potential use. These features have been well documented in two studies done under the sponsorship of the ARL by Economics Consulting Services, Inc. (1989) and Okerson (1989), who located the roots of the crisis of the serials system in, among others, the growth in the volume of published research, the promotion and tenure system, the effect of first copy costs, the "twigging" of journals into subspecialties, and the role of the commercial publishers.
Finally, the inflating mass of ST literature appears to be fast approaching its socioeconomic limits as is witnessed by the growing inability of library budgets to finance it. Libraries are being forced to make the transition from ownership to access, and the resulting cancellations are stressing the entire system by making it ever more difficult to cover first copy costs. Even a relatively few cancellations can have an enormous impact on the system, because the market for many ST serials is small. Of the 154 serials in the database compiled as a result of the 1993 SRP pilot project with the LSU Department of Chemistry, 75% had library holdings of 528 or fewer, and the median library holding of the sample was only 318.5.
Unfortunately, there does not seem to be either an easy or a rational solution to the crisis of the ST serials system in its present form. Technology does not provide an answer and may even exacerbate the situation because it places additional stress on the system without altering the economic fundamentals of ST publishing.
This consequence of technology is apparent in the factors which the American Chemical Society (ACS) takes into account when it prices the newly established online versions of its 26 journals. Over the last 5 years, the ACS has spent millions of dollars on electronic publishing. During a recent interview (Wilkinson 1998, 18), Bovenschulte, ACS Publications Division director, stated that it was costing the ACS approximately $2 million per year to support Web publishing through investments in new staff, training in computers and networks, additional servers, and backup systems. Bovenschulte estimated that such expenses were likely to increase, and he noted that the ACS still had all the costs associated with print.
In a recent talk, Garson (1997a), chief technology officer of the ACS Publications Division, analyzed the economics of ST publishing. He particularly emphasized the impact of the new technology. Garson located the primary cause of the rapid escalation of ST journal prices in the explosive increase in the amount of material being published. According to him, despite substantial improvements in productivity, publishing costs are directly proportional to the amount of material published. Noting that scientific publishers are under tremendous pressure to publish, Garson then gave the following statistics on the growth of ACS publishing for the period from 1980 to 1996: the number of articles rose from around 8,000 per year to over 18,000; the average number of pages per article increased from around 5 to about 7.25; and the number of pages published by the ACS increased from approximately 40,000 to about 135,000. He also pointed out that much—even most—of the material rejected for publication under the ACS system of strict peer review eventually was published elsewhere.
1
Garson then divided ACS journal publishing expenses into two categories, giving their percentages for 1996 as follows: (1) first copy costs (peer review and external editors, technical editing, database building and composition, marketing and sales, research and development)—84.3%; and (2) paper, printing, and distribution costs or expenses directly associated with distribution of the information—15.7%. Of the first copy costs, database building and composition was by far the largest, comprising 43.4% of total production costs, and this fact is important because, as he indicated, database creation is necessary to produce both print and electronic products in a cost-effective manner. The second-highest first copy cost—peer review and external editors, comprising 19.3% of total costs—is also of interest, because it reflects the emphasis of the ACS on the quality of the science it publishes.
Garson held out no hope for a technological solution to the ST serials crisis through electronic journals. By his estimate, even with electronic journals, first copy costs would still be 80% or more, while the format threatened a reduction in the number of copies that could be sold to cover these costs. In a response to questions about his talk, Garson (1997b) noted that ACS first copy costs of 80% or more might be much higher than others due to the association's heavier investment in front-end database creation and rigorous peer review. As evidence of this, for example, Cox (1997), managing director of Carfax Publishing, estimated average first copy costs of scholarly journals at between 60% and 70%.
From the viewpoint of the ACS, electronic journals pose a threat to publishers, because the costs of producing them are the same or even greater than print versions, while they reduce the number of needed subscriptions due to their accessibility from remote points. In his interview (Wilkinson 1998, 13), Bovenschulte noted that the ACS had already experienced an erosion in print subscription numbers over the past five to seven years due to individual subscriptions being replaced by sitewide institutional subscriptions as well as readers and libraries wanting the bells and whistles that electronic journals provide. In his opinion, "electronic journals are going to take over in the long run, no sooner than five years, but no longer than 10," and he declared that, as a result, "scientific publishers—and maybe journal publishers in general—are going to give up counting print subscriptions as a measure of journal success."
The points made by Bovenschulte and Garson were driven home by Durniak (1997) in a written justification of why the ACS had to charge more for its electronic products than its print ones. His rationale basically boiled down to two points: (1) the ACS had to invest more into the first copy cost of database creation to provide such enhancements as powerful search software, hypertext links, interactive molecular models and animations, etc.; and (2) the number of subscriptions would decline as ACS members canceled their personal print subscriptions with the availability of electronic versions from their libraries, and as large organizations eliminated the now redundant duplicate subscriptions to ACS journals. The hostile response to the ACS policy of pricing electronic journals prompted Durniak to declare ruefully (Wilkinson 1998, 16), "Pioneers are the ones that always have the arrows in their chests."
Moreover, similiar to technology, neither cooperative collection development nor consortia offer solutions to the crisis embracing the ST serials system in its present form. The simple fact of the matter is that important ST information is largely restricted to a relatively few serials, and research has consistently shown that these serials not only dominate internal library use but also interlibrary loan. Evidence of this phenomenon appeared in the analysis of the documents supplied by UnCover to LSU from July 1, 1994, to June 30, 1996, from titles classed in the 33 curriculum cores, where a crucial role was played by serials either highly rated by the university's faculty or having high impact factors.
Because of this, all university libraries must have basically the same set of journals—a condition dictated by their being a part of the same social stratification system—something that was demonstrated here by the high intercorrelations among LSU chemistry faculty ratings, ISI citations, and UIUC Chemistry Library use. Moreover, the required set of journals—as the ACS publication data show—has a tendency to grow exponentially. Therefore, in the long run, no library will agree to waste scarce resources on titles that have a low probability of being either used internally or borrowed by other libraries, particularly because such titles are most likely to be subject to inflationary pressures due to a shrinking subscription base from cancellations by other libraries.
Consortia offer no solution to the crisis due to first copy costs. As libraries reduce the number of subscriptions by banding together to buy expensive titles, they force publishers to raise prices to recover these costs. This holds particularly true for electronic journals that are more easily shared, and it is behind the ACS policy of imposing a 25% premium over collective print expenditures on consortia for Web access to those of its journals to which any member library subscribes. When this offer was made to Sanville (29 April 1998) of OhioLINK, it provoked him to describe the premium as "huge" and the acceptance of anything remotely resembling the ACS proposal as "a disastrous message to the community—both publishers and libraries."
Because technology, cooperative collection development, and consortia offer no solutions to the crisis of the ST serials system in its present form, the only option that may be open to librarians is to alter the market structure of the system. ARL has come to this realization, and through its Scholarly Publishing and Academic Resources Coalition (SPARC) project it is seeking "to create a more competitive marketplace for research information by providing opportunities for new publishing ventures, endorsing new publications and information products, and recruiting authors, editors, and advisory board members" (Case 1998, 1). Cognizant of one of the major causes for the dysfunction in the ST information market, the ARL is also working for "the decoupling of the academic credentialing process from formal publication" (p. 5).
Altering the market structure of the ST information system would entail a continuation of the transfer from ownership to access by taking advantage of the ability of the new technology to provide information speedily. In today's parlance, the electronic journal and the fax machine offer libraries the opportunity—through rapid document delivery—to buy "just- in-time" instead of "just-in-case." Under the constant pressure of a dysfunctional ST serials system, librarians will be compelled to rationalize the cost structures of their institutions by carefully placing high cost, low ST value serials on document delivery, essentially removing them from the subscription system to a free market. However, such a transfer raises the crucial question of who should assume the financial risk of the zero and other low use classes—libraries or publishers. This question was raised by ACS director of publications Marks, who in an interview dealt with the issue of single-article ordering in the following manner (Borman 1995, 49):
Nevertheless, the real cost of a journal is not its subscription price but its cost-per-use. This was demonstrated here by the cost accounting method applied to the price and use data of the serials held by the UIUC Chemistry Library. For example, whereas the subscription price of the Journal of the American Chemical Society was $1,055.00, its cost-per-use was calculated to be $2.45. In contrast, another title had a subscription price of $2,331.01 but a cost-per-use of $4,079.27. The true costs of the secondary ST literature that have been for the most part hidden under the present subscription system will have to manifest themselves in the free market of document delivery.
One must keep firmly in mind that what is described here is not a solution to the crisis but its violent denouement and the transition of the ST serials system to a different economic basis. Such a transition can result in the dramatic implosion modeled by Quandt (1996), because it threatens the very revenue base of many publishers. This menace is evident from the data presented by Garson (1997) on the journal revenues of the ACS in 1996. Of these revenues, 81.3% were derived from subscriptions, and only 15.8% came from reprints, page charges, microfilm, back issues, or copyright royalties. Ninety percent of the subscription revenues came from institutions, and 10% from ACS members.
Given these facts and the probability structure of human knowledge, it is highly doubtful whether many publishers can recover their lost subscription subsidies from library budgets through the sale of documents. The transition will not be a pleasant one, because there are powerful forces with vested interests in the existing ST serials system. At LSU, in the famous words of Pogo Possum (Kelly 1982, 224) "We have met the enemy and he is us." After being shown an Evaluator run that brought his pet journal up for cancellation—a foreign commercial title with an extremely narrow subject focus—one LSU professor shouted, "When it comes to my journal, your damn statistics mean nothing!" He then threatened to run to the Provost's Office. The journal in question had a faculty score of 10, suggesting perhaps that 1 faculty member had selected it and assigned it top priority. In his outburst the LSU professor initially directed his anger against his own colleagues for not supporting this journal, and he declared that not only should the journal be kept on subscription but that the university should pour more money into the subspecialty covered by the journal as a natural path to national preeminence. The LSU professor later admitted that he had his own personal subscription to the journal, which indicates that more issues are involved than simple access to information.
It is a terrible thing to be caught in an inefficient market that suddenly becomes efficient. No less a personage than Sir Isaac Newton found this out. When reportedly asked to comment on the South Sea Bubble in whose collapse he subsequently lost his shirt (Carswell 1960, 131, 199), Newton is said to have replied that he could calculate the motions of the heavenly bodies but not the madness of people.