Hal R. Varian
Dean, School of Information Management and Systems
November 21, 1996; Revised April 4, 1997
One can hardly fail to notice that there is considerable activity in electronic publishing. The widespread availability of Internet access in academic circles has led to significant interest in using it as a distribution medium for scholarly communication.
The 1995 edition of the Directory of Electronic Journals, Newsletters and Academic Discussion Lists contains over 700 electronic journal and newsletter titles. This is an increase of more than 75% over the 1994 listing, and an increase of more than 6 times as large as the 1991 edition.
Some of these journals are mainstream supplements to print journals (e.g., Science Magazine On-Line). Others are publications of professional societies (e.g., the Institute of Electrical and Electronics Engineers, the Association for Computing Machinery). Some are published by university presses (Chicago Journal of Theoretical Computer Science); some are published by commercial publishers (Elsevier).
One of the most impressive online publications is the Journal of Biological Chemistry, published by Highwire Press, which is in reality an office in the Stanford University main library. This journal is the core journal for scientists working in biochemistry and molecular biology. The print version consists of about 80 to 100 articles, comprising 1,000 pages per week. A yearly subscription to the print version costs institutions $1,150 per year and individuals about $165 per year. A subscription to the print + online bundle costs institutions $1,350 and individuals about $215.
Among large professional societies the Association for Computing Machinery has put the most thought into how to organize electronic journals. Their electronic publication plan is well worth reading and the AEA is watching their experiments with great interest. Recently the AEA appointed an Electronic Publishing Committee to investigate the issues surrounding this medium; currently this committee consists of the editors of the 3 AEA journals, along with Malcolm Getz, Dan McFadden, and Hal Varian. Although this article expresses my own views, it draws upon my experience as a member of this committee.
The AEA publishes three journals, The American Economic Review, the Journal of Economic Literature and the Journal of Economic Perspectives. The subscription rate for all three varies between $52--$73, depending on income, which entitles members to receive 4 issues of each journal plus the Papers and Proceedings from the annual meeting, which is published as a fifth issue of the AER. Institutional rates rates for the package of 3 journals are $130 per year.
Some of the AEA journals are available in electronic form now. For example, the contents of the Journal of Economic Literature are available in a variety of electronic formats.
In addition, electronic back issues of the American Economic Review are available from at least two different sources. University Microfilms (Ann Arbor, Michigan) provides electronic access to about 5 years of back issues of the AER on an institutional basis, and JSTOR is about to make all issues of the AER from its inception to 1990 available on line to both institutions and individuals. OCLC, an organization that provides online materials to libraries, has asked for permission to make AEA journals issues available to libraries through their online delivery system and this request is currently under consideration. The JSTOR project is particularly interesting, so I shall describe it in more detail.
JSTOR was conceived by William G. Bowen, President of the Andrew W. Mellon Foundation, former president of Princeton University, and labor economist. His original idea was to convert back issues of journals to electronic form in order to save shelf space in libraries. Some back-of-the-envelope calculations suggested that there were significant economies from such conversion.
It costs about 50 cents a page to digitize material and prepare a reasonably accurate ASCII version of the page for searching. This number only includes the variable costs; there are also significant fixed costs to acquire a complete collection, develop an indexing system, implement quality control and so on. The first 80 volumes of the AER consists of about 428 issues and 135,449 pages. The variable cost of scanning this was around $70,000 and the total costs was somewhere between $125,000 and $175,000. Note that this is only the cost of scanning; it does not include the cost for software development, transferring the contents to new media, and administration associated with keeping this material current and making it available to users.
Studies by Cooper  and Getz  suggest that the capital costs associated with the storage space for a single volume of a journal in a library is between $25 and $40, which implies that the 80 volumes of the AER cost between $2,000 and $3,000 to store in perpetuity. It follows that the savings due to freed up shelf space in about 100 libraries would cover the costs of digitization alone. When one adds in the additional costs of operation of JSTOR, it appears that the breakeven point is around 700-800 library subscriptions.
Space costs are, in fact, only part of the story. There are also huge cost savings in library staff time and user time involved in search, circulation, and retrieval. See William Bowen's speech, ``JSTOR and the Economics of Scholarly Communication'', for some rough calculations about the magnitudes of these costs. Extrapolating from the figures given in this speech, the total annual costs (including the amortized space costs) of storing the entire set of volumes of the AER to a typical library is about $400 per year.
These cost estimates, tentative though they are, suggests that there may be considerable savings to digitization. As a proof of concept, the Mellon Foundation sponsored a pilot project to convert back issues of ten journals in economics and history to electronic format. This project was carried out by a group of librarians and computer engineers at the University of Michigan. The economics journals represented are the American Economic Review, Econometrica, Journal of Political Economy, Quarterly Journal of Economics, and the Review of Economics and Statistics. All volumes of these journals published prior to 1990, along with 5 history journals, comprised about 750,000 pages of material. They were scanned into 600 dots per inch bit-images of each page. These bit images were then run through Optical Character Recognition (OCR) software to produce ASCII versions of the text. The ASCII version of the text is not suitable for viewing, but can be used to perform full-text searches of the corpus for words of interest. This combination of linked ASCII and bit-images offers considerable flexibility: the ASCII is searched and the bit image is viewed. The bit images can be viewed on the screen or downloaded and printed out at the desired resolution. The quality is essentially the same as a high-quality photocopy.
The trial users of the JSTOR system have been very enthused about it. This enthusiasm prompted the Mellon Foundation to establish an independent not-for-profit organization (also called JSTOR) to manage further development of the project. The stated objectives of JSTOR, take from their Web page, are:
Five colleges and universities took part in the pilot stage of the process: Bryn Mawr College, Denison University, Haverford College, Swarthmore College, Williams College, University of Michigan, and Harvard University. Subsequently, more universities were added: Cornell University, Emory University, Princeton University, Stanford University, University of California, Berkeley, University of Chicago, University of Delaware, University of Texas at Austin, University of Wisconsin, and Yale University. Members of these universities are being given free access while the system is being shaken out. The system will be available to subscribing libraries in the near future.
The American Economic Association has granted JSTOR access to all back issues (defined as everything 5 or more years old) on a non-exclusive basis. The reason for this decision was that the AEA makes no money from back issues now and that making these back issues more widely available would be socially valuable. JSTOR will make these materials available to libraries and other institutions on a cost-recovery basis. Of course there are significant economies of scale and scope in this operation and JSTOR is actively seeking additional journals and will market its file aggressively in order to exploit these economies. The processing of the American Political Science Review has just been completed, and various ecology and demography journals are in the process of being scanned. They intend to have 100 journal titles in 10-15 fields available within 3 years; the JSTOR Web site contains a list of the available journals.
Having back issues of journals available is highly convenient, but it would be even more convenient if current issues were available on line. This would allow a reader to do very quick and thorough literature reviews. Pulling up references could be done at the click of a mouse rather than the traditional scouring of the library. Having all issues online makes the both the back issues and the current issues more valuable.
It is worth emphasizing that online access of current journals does not preclude paper access. There are many ways in which paper access is more convenient: paper is portable, easier on the eyes, easier to browse, etc. Since the incremental cost of an incremental issue is about $2 for printing and mailing, it is likely that many AEA members will want both paper and on-line access to current issues.
Given the strong complementarity between current and past issues, it seemed sensible to try to investigate ways to make current issues available electronically. Technically this is easy: all of the AEA journals are now produced using PostScript which can easily be displayed online. The difficult issue is the cost recovery problem: how will online copies be paid for?
Currently the AEA sells institutional subscriptions to libraries at about $130 per year and individual subscriptions to members at around $60 per year. Most individual members can access the library version of the journal at some cost in terms of inconvenience. A subscriber at a university, for example, may have to walk to the university library in order to retrieve a current issue of the journal. Most subscribers apparently feel that the inconvenience cost of accessing the library copies is more than the $60 membership fee so that it is worthwhile to have your own subscription.
If libraries subscribe to online versions of the AEA journals they would like to make these versions accessible via the campus network. This means that individuals could access the AEA journals via their desktop on the same terms that they could access their own copies of the journals. The library subscription would no longer be more inconvenient to use than the individual subscription, which makes one wonder whether individuals would still find it worthwhile to continue to pay the subscription fee for personal access.
It is this fear that worries professional societies. If individual do not renew their memberships in the societies, the societies lose a significant source of revenue and a significant part of their membership base.
If we take the average membership fee to be about $60 per year and the variable cost of printing an mailing journals to be about $25 per year, the AEA would lose about $35 for each member who opted out. But it's this $35 per member that is used to fund the fixed cost of producing those journals. If that money disappeared, the Association would be forced to raise the price of the journals---presumably leading to even more defections.
There are several different ways to address this problem. The first is to go to a boundary solution and have either all library subscriptions or all individual subscriptions. Many for-profit journal providers have essentially priced individual subscriptions out of the market and sold only to libraries for years, so it is certainly viable to sell only to the institutional market. However, many professional societies would still be loathe to see their membership roles depleted, even if they were still able to cover the publication costs of their journals.
Another solution is to sell only to individuals---or, in other words, prevent libraries from mounting their electronic subscriptions on a campus-wide network. Presumably the library subscription would still be available within the physical confines of the library via the local network. This solution makes library usage just as inconvenient as it is now and so would likely lead to members continuing to pay their membership/subscription fees.
The third solution is by far the most interesting one. One way to view the problem is as a self-selection problem: how can we get members to reveal their true willingness to pay for a subscription when the ``same'' product is available at a lower price via the library? One answer is to differentiate the product---to make the members' subscriptions more valuable than the library subscription. As we've seen, one way to do this is the make the library subscription more inconvenient to use by requiring that it only be used on-site. Another way is to make the individual subscriptions more convenient to use by providing additional features that would be attractive to members.
Examples of such features would be things like the following.
Some items on this list require investment in creating appropriate technology. Others---like delaying access to the library copy---can be implemented trivially. The key issue is to find aspects of a subscription that the members value highly but are relatively less valuable to library users.
Note that several of these possibilities can be investigated on a campus-by-campus basis via controlled experiments. Others can be examined on a case-by-case basis.
The AEA Electronic Publishing Committee is currently considering these proposals. We are working with JSTOR to offer current issues online in a way that both enhances their value for our members and continues to recover the costs of producing the material.
The goal is to differentiate the product so that library users have high-quality access to the journals ... but not so high a quality that members choose to opt out of membership and access the library subscription. We are also happy to work with other intermediaries who provide online services to libraries. However, we intend to make sure that the key features that we choose to differentiate the individual and library subscriptions are not circumvented by third-party suppliers.
Suppose, for example, that we decided that the members subscriptions would have hypertext links in the bibliography while the library subscription would not have this feature. Third party suppliers of AEA content might well decide to compete by providing this feature. (Even non-profit firms compete!) If so, the value of the hypertext links is reduced as a self-selection device.
The trick in maintaining economic viability for the association will be in finding ways to provide high value to both the individual subscribers and the libraries while still maintaining subscribers in both segments of the market.
Once we figure out an economically sustainable way to offer electronic versions of the AEA journals to members and libraries, it is likely that the form of the journals themselves will change. Here I will offer three forecasts---or maybe I should say guesses---about the form those changes will take.
It is obvious to everyone that value of the AEA journals is that they filter information. In the AER and JEL this is done by peer review; in the JEP this is done by editorial discretion. When publishing was expensive, it made sense to filter ex ante. Expert opinion was used to determine whether or not material should undergo the expensive procedures involved in typesetting, printing, and mailing.
With the advent of ubiquitous computer networks, distribution has become cheap, while attention is still expensive. In the future everything will be published, in the sense of being readily available, but the need for filtering will be even greater than it is now. Filtering will be done ex post, after the material is published.
This means that there will be new ways to filter. One can collect citation data instantaneously. One can use ratings systems other than the old 0-1 decision of published or not. One can personalize ratings: I might be interested in the best articles of the year as determined by the Economic History Association. Economic historians probably have a pretty good idea of what these are, but dabblers like myself would like easy access to that information.
Groups like the AEA will see their primary value in filtering and certifying economic research. Peer review will likely remain a part of that process, but it will be supplemented by a variety of other mechanisms. See my Web page on recommender systems for some provocative examples of such systems and see Resnick and Varian  for discussion and overview of such systems.
Journal articles are now typically written in parts: there's an abstract, an introduction and summary, the body of the article itself, and an appendix. With electronic versions of articles, this is just the beginning. One can write articles in a way that they can be read by a variety of audiences for a variety of purposes.
One obvious possibility offered by this ``variable depth'' of electronic articles is that it allows for the inclusion of material too long for the print media: data sets, computer programs, detailed proofs, and so on could all be included with the electronic version. But I think that a more important possibility of variable depth articles is that it will allow electronic articles to become shorter.
CEOs don't like to read anything longer than 2 pages, and politicians don't like to read anything longer than a bumper sticker. But academicians don't like to write anything shorter than 40 pages. The most common advice from the AER editors to authors of accepted papers is probably ``Make it shorter!.'' This is partly due to space limitations, but is also due to limitations on the readers' attention spans. Authors who are forced to write shorter articles generally write clearer, more accessible, and more to-the-point articles.
With digital documents you can have the document available in all lengths. You might have the title and the one-paragraph abstract, just as you do now. With one click you can display the executive summary. Another click brings up the 4-page ``papers and proceedings'' version. A further click brings up the 12-page AER-length paper. And, if you dare, another click will bring forth the article that the author would really like you to read! Finally, if you want access to the data, computer programs, codebooks, etc. that support the research, you can access the electronic appendix.
This sort of flexibility could do a lot to broaden the readership of economics articles. There are lots of articles that would be great for undergrads, politicians, or CEOs to read. Unfortunately, the articles that are published are written for only one audience: professional economists. When electronic documents become widely available, we will no longer have to live with this one-size-fits-all model.
One of the most popular sections of the AER is the shorter papers and comments section. Here there is something of dialog between authors and commentators. Errors are found and corrected; additional evidence is brought to bear on a problem; new points of view are offered.
This kind of dialog can be very valuable in digging out the truth. At a broader level, the whole enterprise of academic communication is a kind of a dialog. You cite and you are cited as a way of recognizing your contribution to the discussion.
Electronic documents offer much more flexible ways to carry on this sort of dialog. The unit of academic discourse will become a ``thread.'' The citations will become hyperlinks that allow you to follow a discussion quickly and easily. And responding to a discussion will be trivial---although getting someone to read your contribution may well be difficult. Discussions can take place asynchronously, as in comments and replies, or synchronously, as in a chat room.
In a way, academic research will return to its roots: when scholars exchanged letters to report results of their research there was a real dialog going on. When this dialog was replaced with broadcast publication, something was lost. It is true that more people could be exposed to the thoughts of these scientists, but the sense of conversation and dialog was eliminated. Now, with the advent of digital documents, we may be able to have the best of both worlds: broadcast and dialog.
Everyone recognizes that one of the driving forces behind the academic publishing enterprise is the tenure system. Non-tenured faculty publish so that they can become tenured faculty. This is one of the contributing factors to the explosion of academic literature. I'm sure that I am not alone in feeling as though I am drowning in working papers, manuscripts and journals.
If electronic documents make publishing cheaper, this will only get worse. If tenure comes down to weighing 6 years of manuscripts on a scale, assistant professors will turn out as many manuscripts as they can. Herb Simon once said that ``wealth of information creates a poverty of attention,'' and it has only gotten worse since then. Something must be done.
Let me suggest that universities adopt a policy of putting a limit on the number of papers they will accept for purposes of tenure review. For sake of argument, let's say that tenure committees will only evaluate at your 5 best papers, not everything you've ever written. This policy seems like a win all the way around. The authors can focus themselves on doing a few serious pieces rather than lots of shallow ones. The review committee will have a manageable task. And the readers of articles will have fewer, but higher quality pieces to read.
Electronic access to back issues of the AEA publications will soon be widely available through JSTOR and other vendors. We are exploring ways to make access to current issues available electronically in a way that is revenue neutral. Once electronic versions of journals are available, many new possibilities present themselves for scholarly communication.
Michael Cooper , ``A Cost Comparison of Alternative Book Storage Strategies,'' Library Quarterly, 59 (3).
Malcolm Getz  ``Storing Information in Academic Libraries,'' unpublished note, October 17.
Resnick, Paul and Hal R. Varian , ``Recommender Systems,'' Communications of the ACM, 40 (3), March.