The conventional model.

Cost estimates for the conventional model were prepared for each of the 154 cell biology journals in the Science Citation Index (SCI; ISI 2006a). Under the conventional model (the current subscription model), scientific journals have three primary sources of revenue from colleges and universities: subscriptions (print and online), page charges, and submission fees.

We began by searching the online library catalogs of the 24 institutions in the sample, noting which cell biology journals were accessible in print or online format at each institution. Journals accessible only after a delay of months or years were excluded. In calculating costs under the conventional model, we were interested only in those forms of access that resulted in subscription fees or other journal-related expenditures. Free online access to a journal after 12 months, for example, does not represent an expenditure on the part of the institution.

In calculating subscription costs, we assessed each journal at its 2006 list price, accounting for the type of access (print, online, or both) and the relevant characteristics of the subscribing institution (enrollment, Carnegie classification, number of life sciences faculty and staff, etc.). If multiple subscription options were available, we chose the least expensive option consistent with the information presented in the library's catalog. We made two exceptions to the list-price rule, however. First, institutions with current access to more than half the cell biology journals offered by any of four major publishers (Blackwell, Elsevier, Springer, and Wiley) were credited with a 33 percent discount for those titles. Although the terms of large-scale licensing arrangements are generally confidential, our experience with both consortial and single-institution licensing suggests that a 33 percent discount is not unreasonable. Second, the cost of each journal received through the BioOne collection was calculated as the total cost of the collection (for the institution) divided by the number of journals in the collection.

Page charges were assessed at 2006 rates based on the number of pages published in cell biology journals by the authors at each institution (ISI 2006a). Data for a five-year period (2000 through 2004) were divided by five to minimize the effects of annual variations in publication rates. We accounted for variable page charges (e.g., no charge for the first eight pages but $100 per page thereafter) and assessed each page charge at the lowest possible rate (the member rate, for journals published by scholarly societies). Each page charge was assigned to the first author's institution.

Only three journals in the sample charged submission fees. We assessed these fees on the assumption that each published article was the result of 1.6 submissions, an assumption consistent with recent publication data (APS 2007).

The PLoS open-access model.

The nonprofit PLoS publishes several open-access journals in biology and medicine. Their flagship journal, PLoS Biology, has a 2005 impact factor of 14.7, placing it first among the 64 journals in SCI's general biology category. Likewise, PLoS Medicine ranks seventh in the general medicine category (ISI 2006b). PLoS is the best-known open-access publisher, and the only one to have established top-ranked journals in the life sciences. For these reasons, we developed our initial open-access cost estimates in accordance with the publication fee schedule currently in use at PLoS Biology and PLoS Medicine. Our PLoS-model estimates are the amounts that would be paid if all publishers abolished their subscription charges, made their journals freely available online, and charged a publication fee of $2500 per article (the current PLoS fee).

PLoS-model cost estimates were prepared for each of the 154 cell biology journals indexed in SCI (ISI 2006a). Specifically, we used SCI to identify those articles written by authors at the 24 colleges and universities in our sample. Five-year publishing totals (2000 through 2004) were divided by five to arrive at annual figures. We limited our results to research articles and notes, excluding editorials, book reviews, and similar items. Each $2500 publication fee was assigned to the first author's institution, a practice consistent with the billing policies of most open-access publishers. Previous research has shown that this method produces results nearly identical to those achieved when an equal fraction of the cost is allocated to each author's institution (Walters 2007).

The equal-revenue open-access model.

The equal-revenue model is based on the assumption that neither commercial nor nonprofit publishers would willingly accept less money than they currently receive—that within the system as a whole, total open-access publication fees must equal conventional (current) subscription revenue. Unlike the PLoS model, the equal-revenue model is likely to be regarded as realistic by those who contend that open-access publishing offers no incentive for publishers to accept reduced income (Ewing 2004, Mabe 2004, Morris 2005). Although we did not account for the loss of subscription income from non-academic sources, the available evidence suggests that personal subscriptions generate relatively little revenue for most scientific journals (Tenopir and King 2000). Likewise, we did not account for any possible increase in advertising revenue associated with higher readership in an open-access environment. Subscription income from nonacademic institutions, while potentially substantial, could not be included in our cost estimates because of the absence of reliable data on corporate subscriptions.

The cost for each institution under the equal-revenue model was computed as 4.96 times the equivalent PLoS-model cost, since the total revenue generated by the conventional model ($40,737,702) is 4.96 times the total revenue generated by the PLoS model ($8,210,000). Table 1 shows these calculations, which account for the number of colleges and universities within each institutional size category. For example, the 25 institutions with 2,365,001 to 2,865,000 library volumes were assumed, on average, to have the same subscription costs and publishing productivity as the University of Oregon.

Table 1.

Calculation of aggregate journal costs (publishers' revenue) under the conventional (subscription-based) pricing model and the Public Library of Science open-access model.

As table 1 indicates, 1107 colleges and universities did not report the number of volumes in their libraries when responding to the Academic Libraries Survey. Because the institutions that declined to report their library holdings may have had various reasons for doing so (very small collections, lack of current information, institutional policies that prohibit disclosure, etc.), we evaluated a separate random sample of 20 nonresponding institutions to ensure that their publishing productivity and library holdings were consistent with the data for National American University, the institution representing the “not reported” category in our analysis.

The PLoS open-access model.

Within the field of cell biology, a wholesale switch from the conventional model to the PLoS model would result in lower journal expenditures for nearly all of the institutions in our sample; only one institution would pay more. For instance, the University of Illinois would pay publication fees equal to just 40 percent of its current journal expenditures (see table 2). Smaller schools would achieve even greater savings, and 11 of the 24 institutions in the sample (representing 89 percent of all colleges and universities in the United States) would pay no publication fees whatsoever in the field of cell biology. Those institutions, like the general public, would receive online access to all 154 cell biology journals without incurring costs of any kind (other than the cost of Internet access).

Table 2.

Institutional journal costs and availability under three pricing models: the conventional (subscription-based) model, the Public Library of Science open-access model, and the equal-revenue open-access model.

In more general terms, a switch to the PLoS model would have two major effects. The first is an overall reduction in sytemwide costs. As noted earlier, the PLoS fee schedule ($2500 per article) generates only 20 percent as much revenue as the conventional model. Consequently, it requires aggregate payments equal to only 20 percent of current subscription expenditures.

The second effect is a redistribution of costs. Because institutional disparities in research output are far greater than institutional disparities in library holdings, any shift from a pricing model based on subscriptions to a model based on publishing productivity will reduce the proportion of the total cost paid by most institutions and increase the proportion paid by the largest research universities. Within this sample, two universities (Harvard and Michigan) would each pay a greater share of the total systemwide cost under the PLoS model. The other 22 institutions would pay not only lower dollar amounts but a lower proportion of the total cost.

The preeminence of Harvard and Michigan in the life sciences has a major impact on the distribution of open-access journal costs within this sample. For instance, Harvard authors publish an average of 202 papers in SCI-indexed cell biology journals each year—roughly 6 percent of the total national research output in this area, and more than the authors at the other 23 institutions combined (ISI 2006a). Both Harvard and Michigan are among the top 15 recipients of funding from the National Institutes of Health (NIH 2007b). Likewise, a recent guide to graduate programs ranked Harvard 2nd and Michigan 12th in the Biological Sciences (PhD) category. Harvard was ranked 1st, and Michigan 11th, in the Medical (Research) category (US News and World Report 2006).

The data in table 2 suggest a three-group typology of academic institutions. Specifically, Harvard and Michigan—the universities that would pay a larger share of the cost under the PLoS model—can be regarded as primary producers of research in the field of cell biology. They would gain access to additional journals in an open-access environment, but not in proportion to their share of the total cost. Sixteen other institutions in the sample—those with significant expenditures under the conventional model but savings of 50 percent or more under the PLoS model—can be regarded chiefly as consumers of cell biology research. They are the institutions that stand to gain the most from the widespread adoption of open-access pricing. Finally, six colleges in the sample—those with costs of $200 or less under both the conventional model and the PLoS model—are neither producers nor consumers of research in cell biology (see Table 1). The impact of open access on these institutions is difficult to predict. On the one hand, these colleges currently operate without access to any of the 154 cell biology journals indexed in SCI. On the other hand, free online access might encourage faculty and students at these institutions to make greater use of the scientific literature. That is, improved access to scholarly work may itself encourage greater research productivity.

Our PLoS-model results suggest that fewer than 10 universities in the United States can be regarded as primary producers of cell biology research. (That is, fewer than 10 schools fall into the same size ranges as Harvard and Michigan; see table 1.) About one-fourth of all postsecondary institutions are net consumers of cell biology research, and roughly three-fourths are neither producers nor consumers—that is, they neither subscribe to cell biology journals nor publish in them.

The equal-revenue open-access model.

As mentioned earlier, universities such as Harvard and Michigan can expect to pay a far higher proportion of the total systemwide cost under any pricing model in which expenditures are linked to research productivity. Under the equal-revenue model, their absolute costs would increase dramatically as well, by 901 percent at Harvard and by 374 percent at Michigan. As table 2 shows, a switch from the conventional model to the equal-revenue model would bring increased journal costs for 10 of the institutions in the sample: the 8 largest universities—those with more than 2.5 million library volumes—plus Brandeis and Carnegie Mellon. In terms of library size, those 10 institutions represent about 5 percent of the colleges and universities in the United States (see table 1).

The other 95 percent of all postsecondary institutions would achieve cost savings under the equal-revenue model. Even top liberal arts colleges such as Grinnell would spend far less on cell biology journals than they currently do. This finding is especially significant when considered in combination with the potential for improved access to the scientific literature. For example, the 26,000 students at Florida International University would gain access to all 154 of the SCI-indexed cell biology journals rather than the 68 that are currently available to them. At the same time, Florida International would cut its cell biology journal costs by 89 percent.

How many institutions would pay more?

The financial viability of the equal-revenue model is likely to depend on the number of institutions that would experience substantial increases in their journal expenditures. Our sample data suggest that 5 percent of US colleges and universities would pay more than they currently do, and that a handful (perhaps a dozen) would pay far more. At the same time, the idiosyncrasies of this particular sample may mask some of the more general relationships between library size (as a proxy for research productivity) and institutional journal costs. For example, the University of North Texas has a library twice as large as Carnegie Mellon's but would experience a 53 percent decline in journal costs, whereas Carnegie Mellon's would increase 3 percent. The University of North Texas is bigger, but Carnegie Mellon's faculty are more productive.

To illustrate the more general relationship between institutional size and open-access journal costs, we used bivariate regression to estimate the number of institutions that would pay more under the equal-revenue model. Specifically, we used institutional size (number of library volumes) to predict equal-revenue cost as a percentage of conventional-model cost.

The ideal measure of institutional size would be effective as a predictor, valid over the entire range of institutions, and publicly available. This last criterion facilitates the estimation of open-access journal costs for those colleges and universities that do not have convenient access to their own institutional publication data through SCI. We considered a number of predictor variables, including full-time enrollment, graduate enrollment, federal research funding, medical school ranking, and biology program ranking, before choosing the number of library volumes (NCES 2007) as the only measure that met all three criteria. Within the sample of 24 institutions, the number of library volumes is correlated with both current subscription costs (polynomial r² = 0.81) and publishing productivity (polynomial r² = 0.85).

As figure 1 shows, the best-fitting regression model is polynomial. Specifically, y = 3.6354x² + 1.8841x + 24.869, where x is the number of library volumes, in millions, and y is equal-revenue cost as a percentage of conventional-model cost. The regression model has high predictive power (r² = 0.84), and y equals 100 when x equals 4.29 million. This suggests that only those institutions with more than 4.29 million library volumes—30 of the 4221 colleges and universities in the United States—would pay more for cell biology journals under the equal-revenue model.

Figure 1.

Expected relationship between the cost to colleges and universities under the conventional (subscription-based) journal pricing model and under the equal-revenue open-access model, by library size (number of volumes). Most institutions with fewer than 4.29 million library volumes can expect lower costs under the equal-revenue model.

Because the PLoS-model cost for each institution is 20 percent of the equivalent equal-revenue cost, we can further predict that only those institutions with more than 11.18 million library volumes would pay more for cell biology journals under the PLoS model (since y equals 500 when x equals 11.18 in the equal-revenue equation). Only two US universities, Harvard and Yale, have libraries that large.