Academic Databases and the Art of the Overcharge

A fair price is a fine thing. It allows customers to buy with confidence, without worry of overcharge, and it allows vendors to sell knowing the margin will allow them to prosper. In some markets, determining fair pricing is straightforward. Consider how easy it is to check online prices for a pair of shoes. Or to check the price of a house by reviewing what comparable properties have sold for. But in other markets, comparable pricing is not easy to find. The academic library database market is one that holds its secrets close, making it hard for a library to know whether the price it pays is fair.

The database market is served by commercial and non-commercial vendors. They sell essential electronic tools used in academic research, teaching, and patient care. Like most research libraries, our organizations, Iowa State University and the University of Utah, negotiate pricing for a range of different databases. This includes bibliographic databases as well as specialized databases that serve niche communities of chemists, engineers, physicians, political scientists, sociologists, and historians.

Unfortunately, like our peers, we conduct our database negotiations in the dark. Vendors hide pricing behind confidentiality clauses in database agreements, restricting libraries from sharing what they pay. As we and others have pointed out, the information asymmetry this creates greatly advantages vendors during negotiations (Thornton & Brundy, 2020; Bergstrom et al, 2014). As in the academic journal market, lopsided access to information in the database market allows vendors to deploy a strategy of price discrimination, charging not what is fair, but what the customer is willing to pay.

To help libraries avoid price discrimination, we gathered research library pricing for three popular academic databases: SciFinder from Chemical Abstract Services (a division of the American Chemical Society); Scopus from Elsevier; and Clarivate’s Web of Science. The pricing largely came through public records requests to member libraries of the Association of Research Libraries. A few libraries, not subject to confidentiality restrictions, shared pricing with us directly. Pricing, contracts, and key provisions are described in our recent data article (Brundy & Thornton, 2024a), and the full dataset is available in Iowa State’s open data repository, DataShare (Brundy & Thornton, 2024b).

Using this data, we will examine a selection of pricing that demonstrates the range of prices paid by libraries and compare pricing across different institutional factors. We will conclude with tips on how to use pricing data in your library’s next negotiation.

Web of Science

Unlike Scopus and SciFinder Scholar, which are stand-alone products, Web of Science is best understood as a suite of products and services. This makes direct pricing comparisons very difficult, since libraries may subscribe to different combinations of Web of Science products and services. Clarivate, the owner of Web of Science, adds further complexity by not applying a consistent pricing structure to its agreements. For example, one library might receive a bundled price for multiple products, while another might see separate prices for the same products. When pricing is bundled, individual product costs are invisible, making it impossible to understand how the final price was determined.

Because of Clarivate’s inconsistent and complicated approach to pricing, we included an overview pdf document (“Web of Science Product and Service Details”) in the dataset that contains, for each institution, a table showing individual subscribed products, pricing, whether pricing is bundled or split out, and the total price for all products and services. The products and services pricing in the analysis below comes from this overview document.

Figure 1 shows 2022 Web of Science product pricing for six research libraries that license directly with Clarivate. The products in Figure 1 were selected for the sake of this comparison and do not represent all Web of Science products and services subscribed to by each library or detailed in the dataset.

Key Findings from Figure 1:

• Pricing discrimination is present. For example, Texas Tech pays 2.9 times more than the University of Houston for Journal and Highly Cited Data. Texas Tech pays 4.4 times more than the University of Virginia for Citation Connection. And the University of Oklahoma pays twice as much as Texas State for Science Citation Index.
• Bundling limits the ability to compare pricing. The University of New Mexico and Texas State both subscribe to Citation Connection and Journal and Highly Cited Data. Since both receive only a bundled price, it is not possible to compare the pricing of individual items.

To allow us to explore how pricing may or may not correlate to institutional characteristics, Figure 2 shows spending on Web of Science per full-time student and spending on Web of Science as a percentage of total library expenditures. If pricing is based on full-time students, we would expect the price per full-time student to remain relatively stable regardless of the size of the institution. (Full-time students and total library expenditures data comes from the 2022 ARL Statistics.)

Key Findings from Figure 2:

• Web of Science pricing variability is not explained by an institution’s total number of full-time students. Texas Tech, which has approximately 26 percent more students than Texas State, spends the most per full-time student, at $13.14, while Texas State, at $4.88, spends the least. But the University of New Mexico, whose headcount is less than half that of either of them, pays a price-per-student of $11.07
• Web of Science pricing variability is not explained by total library expenditures. Texas Tech spends $488,444 to license Web of Science, 1.38 percent of its total expenditures. Virginia spends $257,660—0.54 percent of its total.

Scopus

Scopus, produced by Elsevier, is an abstract and citation database typically used by researchers to conduct literature reviews and analyze citation impact. Since it is a stand-alone product, its pricing is less complicated than Web of Science. Figure 3 shows pricing for libraries that license directly with Elsevier. Similar to Figure 2, it shows spending on Scopus per full-time student and as a percentage of total library expenditures.

Key Findings from Figure 3:

• Scopus pricing variability by the total number of full-time students is extreme for some libraries. University of Houston spends the most per full time student at $6.08. University of Florida spends the least at $1.81. Florida has approximately 41 percent more students than Houston, but Houston pays 2.39 times more than Florida.
• Scopus pricing variability is not explained by total library expenditures. Texas State spends $70,642, 0.32 percent of its budget, to license Scopus. University of Houston spends $202,980, 0.90 percent. This suggests willingness to pay, rather than ability to pay, determines Scopus pricing.

In explaining pricing, vendors, at times, will connect pricing with research productivity. Thus, another method of assessing whether pricing is fair is to analyze it relative to research publication output. A key question is whether high research output institutions are paying more or less per research publication. The preliminary assumption is that institutions with higher publication output would pay more per publication than institutions with lower publication output.

Figure 4 analyzes article output for seven institutions within a pricing band from $70K to $278K. One limitation of the publications data is that it does not include affiliates that may be associated with the primary institution. For example, totals for the University of Utah do not include the University of Utah Hospital, which is an affiliate of the University of Utah. Another limitation is the small sample size, which limits our ability to generalize the findings to a broader population.

Key Findings from Figure 4:

• The data shows an inverse relationship between publication output and price per publication—institutions with higher research output typically paid less per publication than those with lower research output. The findings suggest that research output doesn’t explain the pricing differences among the institutions. Thus, more evidence of pricing discrimination.
• Institutions with higher research expenditures typically paid less as a percentage of pricing compared to total research expenditures. For example, Texas Tech, which had a relatively low total R&D expenditure, had a ratio of price to total research expenditures of 0.09 percent. Florida, which had one of the highest research expenditures, had a ratio of price to total R&D expenditure of 0.01 percent, approximately 9 times smaller than Texas Tech’s.

SciFinder

SciFinder is a database produced by the American Chemical Society that is typically used by researchers to conduct literature reviews and find chemical information. Similar to Figure 2 for Web of Science and Figure 3 for Scopus, Figure 5 compares 2022 SciFinder pricing per full-time student and spending on SciFinder as a percentage of the overall library budget. With the exception of the University of Buffalo, which licenses through NERL (NorthEast Research Libraries), all universities in Figure 5 license directly.

Key Findings from Figure 5:

• SciFinder pricing variability, as with Web of Science and Scopus, is not explained by an institution’s total number of full-time students. Iowa State spends the most per full-time student at $7.26. Illinois spends the least at $3.37. Illinois has approximately 69 percent more students than Iowa State, but Iowa State pays 1.27 times more than Illinois.
• SciFinder pricing variability is not explained by total library expenditures. Illinois spends $156,930, 0.28 percent of its budget, to license Web of Science. Iowa State spends $200,060, 0.96 percent. This again suggests willingness to pay, rather than ability to pay, is driving pricing.

Similar to Figure 4, Figure 6 analyzes seven institutions with SciFinder pricing ranging from $140K to $203K, looking at price relative to research productivity.

Key Findings from Figure 6:

• Similar to the Scopus findings in Figure 4, institutions with higher research output typically paid less per publication than those with lower research output. Thus, an inverse relationship between article output and price per article exists. The findings suggest that research output doesn’t explain the pricing differences among the institutions—more evidence of pricing discrimination.
• Institutions with higher research expenditures typically paid less as a percentage of pricing compared to total research expenditures. For example, the University of Illinois’s ratio of price to total research expenditures was 0.02 percent vs. 0.09 percent for University of Oregon, whose research expenditures are approximately 4.8 times lower.

Putting Pricing Data to Work

Clarivate, Elsevier, and the American Chemical Society are comfortable pursuing a strategy of pricing discrimination. But there is no reason libraries should allow this approach.

As we have done, we recommend libraries review their pricing for all three databases. Find comparables and determine whether local pricing is fair. If a library finds it is paying a high price relative to peers, it is possible to use comparable pricing to negotiate a discount. In 2024, Iowa State was able to use SciFinder pricing from the dataset to show we were being overcharged relative to peers. We negotiated a 12.5 percent price reduction.

Clarivate’s approach to Web of Science pricing presents unique challenges. We recommend libraries review the Products and Services tables included in our dataset, as well as their own products and services pricing, to get a better understanding of Clarivate’s different pricing strategies. If Web of Science product pricing is bundled, libraries should ask to have it split out to allow easier comparison. If pricing is already split out, libraries should compare individual item pricing to similar institutions in the dataset.

We also noticed some libraries are charged a technology fee while other libraries are not. For example, Oklahoma’s fee is $13,000 and Texas Tech’s is $4,800. Iowa State and Utah are not charged this fee. Closely scrutinize your library’s Web of Science Product and Services Detail table to ensure that your library is not paying an unnecessary fee.

There are many other comparisons that can be made from the pricing in the dataset and lessons to be gleaned. For example, what is the advantage of licensing directly versus going through a consortium? How prevalent are introductory pricing rates? And what are the trends and averages for inflation cap rates? Alex Lao, from Stanford Libraries, used the pricing from the dataset to create visualizations in Tableau, which are sure to provide additional insights.

Fifteen years ago, in the midst of the Great Recession, the Association of Research Libraries Board of Directors passed a resolution strongly encouraging its member libraries to refrain from signing agreements that restricted the sharing of pricing and terms.

But today, accessing library pricing and terms is as difficult and time-consuming as ever. It took the two of us many, many hours over a two-year period to surface the pricing for these three databases. While this data can provide many insights, research libraries subscribe to dozens of these types of resources. This dataset is only a flashlight glimpse into a darkened room.

If you believe having access to pricing data will support your negotiations, then please take a stand on removing confidentiality clauses. Otherwise, vendors, as we have shown, are happy to overcharge.

Note: This article has been updated to correct an error in data for the University of Illinois that appeared in Figures 3 and 4. The “Total Full Time Students,” “Total Library Expenditures,” and “Total R&D Expenditures” figures presented in those tables reflected only the University of Illinois Urbana-Champaign campus, while the Scopus contract applies to the entire University of Illinois system. As a result, the “Price Per Full Time Student,” “% of Total Library Expenditures,” and “% of Total R&D Expenditures” figures in those tables were inaccurate. Accordingly, we have removed “University of Illinois” from both tables. We regret the error.

References

Bergstrom, T. C., Courant, P. N., McAfee, R. P., & Williams, M. A. (2014). Evaluating big deal journal bundles. Proceedings of the National Academy of Sciences, 111(26), 9425-9430. https://doi.org/10.1073/pnas.1403006111

Brundy, C., & Thornton, J. B. (2024a). Academic library pricing dataset for SciFinder Scholar, Web of Science, and Scopus: 2018-2024. Journal of eScience Librarianship 13 (2): e959. https://doi.org/10.7191/jeslib.959

Brundy, C., & Thornton, J. B. (2024b). Academic Library Pricing Dataset for SciFinder Scholar, Web of Science, and Scopus, 2018-2024. Iowa State University DataShare [Data set]. https://doi.org/10.25380/iastate.26081797.v1

Thornton, J., & Brundy, C. (2021). Elsevier title level pricing: Dissecting the bowl of spaghetti. Journal of Librarianship and Scholarly Communication 9 (1): eP2410. https://doi.org/10.7710/2162-3309.2410