Near the end of 1997, Reed Elsevier and Wolters Kluwer proposed a merger of their operations that upset more than a few librarians. Elsevier’s publishing "empire" was about to grow even larger. The fear of runaway academic journal prices was palpable in places where the mere mention of the company’s name causes people to grab for their wallets.
More than a year and a half later, much has changed in the academic publishing markets. Although the proposed Elsevier/Kluwer deal failed after facing regulatory scrutiny, consolidation continues at a rapid pace. A half dozen major transactions involving science, technical, medical (STM), or legal publishers have occurred over the past 18 months. At the same time, new web-based technologies are transforming the production and delivery of scholarly research articles. These events have provided me with a unique opportunity to assess the economic behavior of academic publishers and libraries, first at the U.S. Department of Justice (DOJ) and now as an Assistant Professor at Georgia Tech. Last fall I reported some preliminary results from this effort in ARL.2 My intention here is to briefly revisit that article and then describe my subsequent progress.
Normally, the most important issue in merger analysis is determining the extent of the relevant market. A narrowly defined market, where the number and type of products included is small, makes it more likely that a merger of two sellers in this market is anticompetitive. Conversely, in a broadly defined market the likelihood of harm is not as great. For any given merger the choice of market definition depends on whether market power could be exercised by a hypothetical monopolist in the defined market.3
In the case of publishing, the DOJ typically assumed that products consisted of book or journal content and that the appropriate market definition included books or journals containing very similar content. So, for example, if two publishers proposed a merger, antitrust officials would focus on the extent of overlap between the companies’ content. If the market definition suggested that the two firms’ share of a given market was "too" high, i.e., a postmerger price increase was likely, then a divestiture might be required; if the extent of the overlap was large across the firms’ product range, the DOJ could sue to block the entire transaction. Of course, since any journal is at best an imperfect substitute for any other journal, markets were defined fairly narrowly.4 And because most companies’ journal assets were highly differentiated, most if not all mergers appeared to be harmless. As a consequence, over the past decade or so antitrust activity in the academic and legal publishing markets has been very quiet. Even the controversial Thomson/West merger in the mid-1990s resulted in few meaningful divestitures.
So when the Elsevier/Kluwer deal was proposed in 1998, the companies probably anticipated little resistance from government authorities in the U.S. or Europe. Although the European Union would eventually register concerns about overlap among the two companies’ European legal products, their argument was based on the traditional approach to publishing markets. The product market was defined narrowly in terms of content. In the U.S., where these foreign-language legal treatises had no market, there was little opportunity for a conventional antitrust case. If markets were defined narrowly, STM journal content overlap between the two companies was fairly minor.
A characteristic of antitrust enforcement in the U.S. is that the population of staff attorneys and economists experiences a fair amount of turnover from year to year. Furthermore, even if an experienced crew is theoretically available, they may have pressing commitments on other cases. This is both a strength and a weakness: a weakness because good ideas may need to be rediscovered (or at least learned anew) by a new staff; a strength because the staff is more likely to adopt a fresh approach when the staff has no strong prior beliefs about a case. I like to think that the situation at DOJ at the time of the Elsevier/Kluwer deal reflected the latter set of circumstances. The staff had no experience with publishing markets. And thanks to discussions with dozens of concerned librarians, the staff grasped the need for a fresh approach.
Indeed, librarians across the country were concerned about persistent journal price inflation, especially among STM titles. In response, since their budgets were growing more slowly than journal prices, many university libraries had been forced to re-allocate dollars from monographs to journals, to postpone the purchase of new journal titles, and, in many cases, to cancel titles. As a consequence, libraries now need to rely more often on interlibrary loans to satisfy faculty demands. However, the most interesting thing we learned from these discussions was that library demand for journals was unlike most markets. Given a set of similar titles, libraries do not subscribe only to the journal offering the best value. Rather, journal cost per use is minimized across a broad field of study, e.g., biomedicine, subject to a budget constraint. The result is a demand for a "portfolio of titles" where the cost-per-use criterion is applied broadly. Thus, journal titles compete with one another for budget dollars over an entire field rather than across a narrow subfield, as intuition might otherwise suggest.
Based on this observation, we developed a portfolio theory of buyer and seller behavior in academic journal markets. Given libraries’ demand for portfolios of titles within broad fields of study, we demonstrated in a simple economic model that, all else equal, publishers set prices so that higher use (or quality) journals exhibit lower cost-per-use ratios. Thus, higher use journals (that have a lower cost per use) are purchased by most libraries. Conversely, lower use journals (that have a higher cost per use) are purchased by fewer, relatively high budget libraries.5
The intuition for this particular ordering is that higher use imparts a "cost advantage" that makes it more profitable for their publishers to price low and sell widely. Given this strategy, lower use or "high cost" journal publishers find it most profitable to price high and sell to fewer, relatively high budget libraries. Note that although the latter firms could match the "low cost" firms’ prices, this strategy is less profitable than targeting the smaller base of high budget customers.
Using this model it is also possible to show, in some cases, that mergers are profitable for journal publishers. A corollary is that the merged firm’s journal prices increase. The idea here is that the merged firm is able to internalize certain pricing externalities that the merging parties fail to consider when they act independently. Larger portfolio firms are better able to capture these benefits and therefore, all else equal, set prices at a higher level.6
Given these theoretical possibilities, we attempted to test these predictions with actual data. We collected information on literally thousands of STM titles from a variety of sources, for the period 1988–1998. We chose to focus initially on biomedical titles (see my working paper for details). By May 1998, we were able to show, using a so-called reduced-form econometric model,7 that a firm’s portfolio size was positively related to journal prices, and that past mergers were associated with higher prices. However, even after controlling for the effects of portfolio size and other variables, we still observed a substantial inflation residual.
One of the weaknesses of a reduced-form methodology is that unless the investigator has strong prior beliefs about the events that are being measured, the cause of a price change may be uncertain. For example, when we observe higher postmerger prices is this because the merged firm is exploiting greater market power or are the price increases due to unrelated increases in the willingness of buyers to pay higher prices, i.e., a decrease in the elasticity of demand? To help eliminate this uncertainty, economists are sometimes able to estimate structural econometric models that explicitly identify these separate factors. To estimate these types of models, price and quantity data are necessary. In other words, journal prices need to be supplemented by the number of subscriptions for each title. Fortunately, during our antitrust investigation we were able to collect subscription data from a variety of academic and medical libraries. However, unlike other types of information, such as journal price and citation data, library holdings are reported in a highly idiosyncratic fashion. Thus, creating a usable database is very labor- and time-intensive. As a consequence, estimation of a structural model was delayed until June 1999.
To estimate the structural model, I used holdings data from 194 medical libraries, selected randomly from among Medical Library Association members. This data includes some 60,000 subscriptions to ISI-ranked journals (those 8,000 journal titles that the Institute of Scientific Information considers the most significant in their contributions to scientific research). Libraries of all sizes are represented in the sample, some holding less than 10 subscriptions, while others report collections exceeding 1,300 ISI-ranked titles. The preliminary results support the portfolio theory.
Using the ISI-defined biomedical portfolio and the corresponding library holdings, I calculated the actual size of the various commercial publishers’ journal portfolios as well as the observed sample portfolios from the 194 medical libraries. Table 1 reports this information using both sources. It is clear from this table that significant variation in portfolio size exists in the industry.
Table 2 presents information by publisher on average subscription price, number of times cited, cost per use (price per citation), and number of articles published in 1988 and 1998. Though prices, citations, and article counts generally increased during the period, the rate of change for prices was far more striking, resulting in higher cost-per-use numbers by the end of the period. For example, Elsevier’s average journal price more than tripled during the period, while the corresponding citation and article counts increased less than 25%.
Table 3 provides the average number of subscriptions for ISI-ranked biomedical journals, by publisher, in 1988 and 1998.8 The average number of subscriptions declined only 1.5% from 1988 to 1998. Given that nominal prices increased dramatically over the sample period, the apparent inelasticity of demand indicated by these numbers is remarkable. It suggests that library serials budgets, whether through budget increases or a combination of budget increases and the cancellation of non-ISI-ranked titles, were sufficient to absorb most of the price increases of the ISI-ranked titles. At the same time, these numbers provide indirect support for the model of journal pricing that, all else equal, if library budgets increase, firms have an incentive to proportionally raise prices.
During the sample period (1988–1998) two significant mergers occurred: one between Pergamon (57 biomedical titles) and Elsevier (190) and the other between Lippincott (15) and Kluwer (75). To estimate the impact of these mergers on the prices of the biomedical journals being studied, a subset of data from the larger sample of medical libraries was analyzed. According to these empirical estimates, each of these mergers was associated with substantial price increases; in the case of the Elsevier deal the price increase was due solely to increased market power.9 For example, compared to premerger prices, the Elsevier deal resulted in an average price increase of 22% for former Pergamon titles, and an 8% increase for Elsevier titles. This asymmetry probably reflects the corresponding asymmetry in premerger journal portfolio size for the two firms. That is, Pergamon’s relatively small biomedical portfolio prevented it from realizing it could profitably set prices at the same level as Elsevier for journals in the same class. In the Lippincott/Kluwer merger, a 35% price increase in former Lippincott titles was due in part to increased market power, but also due in part to an apparent increase in the inelasticity of demand for the titles. That is, after the merger, Lippincott titles were even less likely to be cancelled.
These results also contain a likely explanation for the persistent journal price inflation observed in most academic fields.10 The sensitivity of library demand to price increases is very small by normal standards (a 1% increase in price results in a 0.3% decline in subscriptions). Given this inelastic demand, publishers have a strong incentive to increase prices faster than the growth rate of library budgets. Based on the structural model estimates for key biomedical journals, the average annual increase in journal prices net of price changes due to journal quality and costs was nearly 10%.
These new results are consistent with our earlier findings. Consider the policy implications. To date, the cumulative evidence indicates that conventional antitrust procedures are inadequate for evaluating mergers in academic journal markets. First, market definition needs to be focused on broad portfolios of journal titles rather than a narrow content-based concept in order to reflect the reality of libraries’ demand. Second, mergers involving relatively small companies can have substantial price effects (in 1991, Pergamon was not among the top five publishers in terms of portfolio size). Although antitrust policies in the U.S. and Europe have changed considerably over the past two decades in response to new developments in economics, the special case of academic publishing remains to be addressed. At least two options are available. On occasion, the Department of Justice and the Federal Trade Commission have adopted special antitrust guidelines for markets with unusual characteristics, e.g., for health care and intellectual property.11 In other instances, antitrust immunity has been granted to certain parties when important social objectives are threatened (access to scientific research certainly merits the label of an "important social objective"). For example, the DOJ, with congressional approval, could grant libraries permission to form a single nationwide buying consortium to counter the substantial market power of publishers.
In the meantime, this research project is still in its infancy. Important future objectives include (1) examining the impact of new journal entry on prices of incumbent journals, (2) contrasting the behavior of nonprofit and for-profit publishers,12 and (3) testing the robustness of this portfolio approach in other STM fields. Finally, I would like to thank the many libraries, librarians, and their associations for their invaluable assistance over the past year and a half.
© 1999 by Mark J. McCabe. The author grants blanket permission to reprint this article for educational use as long as the author and source are acknowledged. For commercial use, a reprint request should be sent to the author mark.mccabe@econ.gatech.edu.
A version of this paper was published as "Handling Medusa: The Impact of Publisher Mergers on Journal Prices: An Update," Against the Grain 11, no.4 (Sept. 1999): 58–61.
Mark J. McCabe, "The Impact of Publisher Mergers on Journal Prices: A Preliminary Report," ARL: A Bimonthly Newsletter of Research Library Issues and Actions, no. 200 (Oct. 1998): 3–7. http://www.arl.org/resources/pubs/br/br200/br200mccabe.shtml.
According to the horizontal merger guidelines (http://www.usdoj.gov/atr/public/guidelines/horiz_book/hmg1.html), antitrust authorities "…will delineate the product market to be a product or group of products such that a hypothetical profit-maximizing firm that was the only present and future seller of those products (‘monopolist’) likely would impose at least a ‘small but significant and nontransitory’ increase in price."
For example, suppose two publishers of economics titles were merging. If one owned a series of labor economics journals and the second firm specialized in industrial organization, it is not likely that antitrust concerns would be raised because the market for labor economics journals is seen as separate from that of industrial organization journals.
For a more extensive discussion of this model, its predictions, etc., see my working paper entitled, "Academic Journal Pricing and Market Power: A Portfolio Approach" (July 1999). This paper can be obtained in Portable Document Format at [2000 revised version: 7/10/2001] http://www.prism.gatech.edu/~mm284/JournPub.PDF.
When acting independently, publishers set prices to maximize the profits of their respective titles, and ignore the external impact upon their competitor’s profits. If products are differentiated, as in the case of scholarly journals, then this externality is "positive," i.e., if one firm raises its prices, then other firms have an incentive to raise their prices as well. Since mergers allow firms to internalize this positive effect, higher prices are observed. The intuition here is that a positive feedback process is at work inside the merged firm: raising the price of journal A makes it profitable to raise the price of journal B, which in turn makes it profitable to raise again the price of journal A.
For a description of this methodology, see McCabe Oct. 1998, 5–6.
These numbers exclude titles that commenced publication after 1988. Including these newer titles would tend to lower the reported 1988 figures relative to the later 1998 numbers.
See detailed statistical analysis of these mergers in McCabe July 1999, rev. 2000.
Note that these inflationary trends are not restricted to commercial publishers; in the case of biomedical journals, nonprofits have raised prices as well. See Table 4 and endnote 12.
See the Department of Justice’s antitrust guidelines at http://www.usdoj.gov/atr/public/guidelines/guidelin.htm.
Descriptive data on nonprofit publishers in the biomedical field indicate that in 1998 the average nonprofit journal provides one and one-half times the number of papers with four times the number of citations for less than one-half the cost of commercial titles (see Table 4). These differences between nonprofit and commercial titles may reflect variation in a number of factors, including market power, pricing strategies, timing of entry, the choice of product space, etc. Regardless of the ultimate explanation(s) and their implications for social welfare, it appears that the two modes of publishing behave very differently.
Mark J. McCabe, "The Impact of Publisher Mergers on Journal Prices: An Update," ARL: A Bimonthly Report, no. 207 (December 1999), 1–5, http://www.arl.org/resources/pubs/br/br207/br207jrnlprices.shtml.