Friday, March 03, 2006

Elsevier and TCS

My post A Referee's Boycott generated quite a discussion in the comments, particularly about Elsevier. Paul Beame asked about why the EATCS still sponsors the Theoretical Computer Science through Elsevier. Don Sannella, editor-in-chief of TCS-B (Logic, Semantics and Theory of Programming), responded to Beame and earlier comments. Paul sent me a response to Sannella's comments. I'm reposting Sannella's comment followed by Beame's response.
Don Sannella's Comment

Regarding the relationship between EATCS and TCS: EATCS is in the process of changing its statutes to say that it supports the spread of the results of research and exchange of information through scientific publications, without specific mention of TCS or any other journal. This decision has already been made and approved by the membership; the only thing holding up its implementation is the fact that EATCS is legally a Belgian organization so revision of the statutes involve lawyers etc. I think this is an appropriate change (speaking also as a member of the EATCS Council); the previous situation was simply a result of the way that EATCS and TCS grew up together and were set up by the same people, starting at a time when there were very few journals.

Regarding criticisms of TCS:

  • Copyright: There is a lot of misinformation circulating about this issue. I have even caught one of the main advocates of open access publishing making plainly false statements in a public talk. I suggest that there would be more light and less heat if people would take the trouble to find out what the actual situation is before criticizing.
    I think the main practical issue is ability of authors to publish their work on their own websites. In this respect Elsevier's copyright agreement is not significantly different from the ACM's, or Springer's, unless there has been a recent change to these that I haven't noticed. There is an explanation of this aspect of the Elsevier copyright, by the Elsevier editor in charge of TCS, in the Bulletin of the EATCS number 75 (Oct 2001). The EPrints organization regards Elsevier as self-archiving-friendly ("green" status) and it reached that status before Springer did.
  • Price: I know that TCS is expensive, probably the largest item in any Computer Science library's journal subscription budget. But it is also very large, with 12000 pages published per year. If you look at the price per page (here are 2004 figures from the AMS for mathematics journals which are by the way substantially different than the price comparison given by Wim van Dam) the cost is $0.42/page which is comparable with other journals. This doesn't take the thousands of pages in ENTCS, which comes free with TCS, into account. The whole issue of journal price is complicated because the primary mode of access these days is electronic, and prices for electronic access are negotiated on a case-by-case basis. If you discuss the issue with Elsevier, the statistic they will give you is that the per-download price of an article in TCS (computed by taking the total cost of subscriptions and dividing by the total number of downloads, I think) is considerably less than $1. According to Elsevier, this is the figure that librarians care about, and the fact that it is a fraction of the cost of interlibrary loan is the key point.
  • Open access: The open access movement advocates journals that are free to readers. In this model, the author is the one who ends up paying; this fact is mentioned much less often and some people who advocate open access don't appear to be aware of it. (I know of one new open-access journal that is free to authors as well because the costs are covered by a university, at least for the moment. The point is that somebody needs to pay; running a journal is not a cost-free spare-time activity. See "Guide to Business Planning for Launching a New Open Access Journal" from the Open Society Institute.) There are major opportunities for unfairness in the editorial process with author-pays but otherwise the only problem I see is that with both models co-existing, few authors with an article that would be accepted by a "normal" journal will be willing to pay for publication in an open access journal. Springer has recently offered authors the choice of paying a fee in order to make a paper open access, or not paying and leaving it as paid access. I hope they publish statistics on how many authors decide to pay!
  • Academic Press versus Elsevier: "Academic Press had its flaws but they were not predatory in their pricing." Well, compare AMS's 2004 figure for Information and Computation ($1.07/page, Elsevier-owned) with its 2001 figure ($1.92/page, Academic Press-owned).
  • Quality of TCS: As editor-in-chief of TCS-B — which is admittedly probably not the main part of interest to readers of this blog — I am responsible for its quality. I think the quality is pretty good and improving. Opinions on this may vary of course. At least, it is not the case that the alleged decline in quality is because (as Paul Beame asserts) "TCS went to a highly distributed editorial board". The way that the TCS editorial board works has not changed since it was founded in 1975, as far as I know. I wonder where he gets his information. I am unhappy about the implied suggestion that the TCS editorial board members are not exercising proper editorial judgment.
Finally: I am not here to make excuses for Elsevier. My interest is TCS (and EATCS) and replying to some points above that are factually incorrect.
Paul Beame's Response

I am happy to hear about the EATCS change. Let me address the two main points, copyright and price, as well as open access journals.

Copyright I agree that copyright is no worse at Elsevier than at Springer (in fact Springer has gotten worse recently). Copyright transfer is apparently not required given the following text I received from Elsevier regarding a JCSS paper:

Recently, we sent you a Transfer of Copyright form relating to the above-mentioned. We note that you have not yet returned a completed form duly signed. In order to avoid any delay in publication, we ask that you do so immediately. Attached you will find a further copy of the form. Please return the completed and signed original of this form by mail or fax, or a scanned copy of the signed original by e-mail.

If we do not hear from you by return, the article will carry a line in place of the copyright line merely indicating that Elsevier published the article.

This sounds all right BUT when I have explicitly took advantage of the second option I noticed that when the article was published Elsevier still explicitly claimed copyright on it!

Price Thinking about things as price per page is exactly the problem. TCS was one of the top 2 or 3 theory journals and around 2000 pages annually until 1989 when it decided to go to bi-weekly publication and a much larger editorial board and upped its page count to 3500, raising its prices drastically overnight to keep the same price per page. The average quality declined markedly at this time as the good papers were swamped with more lower quality fare. TCS still publishes many good papers but it is nowhere near as high quality as it was in the 1980's when it got many of the top papers in the field.

Moreover TCS is just one Elsevier journal. Their behavior with others is part of the problem: In the early 90's I was deciding between publishing in Annals of Pure and Applied Logic (Elsevier) and Journal of Symbolic Logic (ASL). I was told that longer papers were more appropriate for APAL and so submitted there. I made the mistake of not checking prices: JSL was 12 issues a year, each over 300 pages, and cost $400 or so annually. APAL had 4 issues per year, each about 250 pages, and cost more than $2000. The quality of the two was similar.

I speak with librarians who have to purchase journals. The pricing for electronic journals that Elsevier sets are bundled in such a way that they feel forced to subscribe electronically to many journals that they do not want to purchase. The comparison with inter-library loan is absurd.

The price comparison should be with society-published journals such as the ACM and SIAM journals. These do provide the main office editorial staff that for-profit journals provide.

Open Access I agree that the long-term soundness of the open access model is not yet fully established. (There are some things that need to be paid for without voluntary investment beyond refereeing and it is not yet completely clear how to do this long-term.) However, if you want an example of an open access journal that does not seem to suffer from the flaws you describe, consider JAIR (the Journal of Artificial Intelligence Research) which has been operating for more than a decade and is one of the top couple of journals in AI.

(It may be too soon to tell about Theory of Computing is in its infancy but it already has a very high quality of papers.)

Why is it that Elsevier regularly emphasizes the comparisons with nascent open access journals but regularly ignores comparisons with high quality society-published journals such as SIAM and ACM journals?


  1. In general, you should not believe a single word you hear from Elsevier about is pricing. They have mastered the art of distorting the facts to obscure the truth.

  2. Six things that researchers need to know about open access:

  3. "Thinking about things as price per page is exactly the problem.": I agree that price per page figures have to be taken with a grain of salt, but I don't know a better metric for making comparisons. Do you?

    Here is one metric that may have some faults but is at least a start:

    I checked the last couple of years of TCS versus SIAM Journal on Computing on the ISI Journal Citation Reports website.

    In 2003 and 2004 TCS had only 10% more total citations of its articles than SICOMP did despite having between 6 and 7 times the number of articles!

    But it is also very large, with 12000 pages published per year.

    ... for SIAM Journal of Computing the figure is $0.27/page versus $0.42/page for TCS. Looks like the same ballpark to me.

    OK, so if the value of a journal is represented by how many times people cite its articles then we can make a comparison that is more meaningful. SICOMP is publishing between 1,500 and 2,000 pages per year versus the 12,000 that TCS publishes. Using the 50% higher per page cost of TCS, this conservatively makes TCS 9 times as expensive in total for only 10% more value.

    While I not saying that citations are the only indication of the worth of a journal I think this makes it clear that price per page is a very poor metric.

  4. I can only say that SICOMP is of not much relevance to the scope of TCS-B, which is "my" bit of TCS

    You are certainly right that Theory-B topics (named after volume B of the Handbook of Theoretical Computer Science) are not covered by SICOMP. What would you say are the main competitors for TCS-B in covering Theory-B topics? I can think of one relatively new society-published journal covering Theory-B topics: ACM Transactions on Computational Logic (ACM TOCL).

  5. I came to this discussion late, but it's one that I am very passionate about. I have two comments, but they are about scientific publishing in general rather than the specific case of Elsevier or TCS.

    1. what monetary costs are people referring to in producing an electronic journal? Has anyone written an analysis of what it costs to sustain an electronic journal? Note that I said a *journal*, not what it costs to maintain a staff for a society that backs a journal.

    2. I fully support the decision of a referee to refuse to referee for Elsevier. In my opinion commercial publishers have largely outlived their usefulness for publication of scholarly journals in computer science, and the direction to online subscriptions is detrimental to the health of science. It's time to start saying no to participating in the scam. That doesn't mean that we should stop refereeing - that is an absolute requirement for participation in the scientific process. It means that some of us will give our support by refereeing for journals that we feel are serving the field well. As a result, these journals will receive better service in refereeing, resulting in more timely and quality publication, and in the long run it means that they will overtake the more costly commercial journals.