The Occasional Pamphlet

Tightrope walker, sculpture, Berlin, 2008. Photo from beezerella at flickr.com. Used by permission.

The short answer?  Almost nothing.

The Compact for Open-Access Publishing Equity is a statement of commitment to “the timely establishment of durable mechanisms for underwriting reasonable publication charges for articles written by its faculty and published in fee-based open-access journals and for which other institutions would not be expected to provide funds.” Some institutions who were considering signing on to the compact at its launch held off because of a worry that it might cost a lot of money at a time when library budgets are under phenomenal pressure. I had predicted that the costs would be minimal in my PLoS Biology paper proposing the compact. There, I said

By design, the overall cost to a university of implementing the compact, in the short term, would be quite small. Hybrid open-access fees are explicitly eschewed, and true open-access fees tend to be found at present in just those areas of scholarship where grant support is most prevalent, reducing the underwriting load on the university substantially. Rough estimates based on the experience of the Berkeley Research Impact Initiative fall in the range of tens of dollars per faculty member per year.

But I can understand that some universities might have wanted to wait until there is some empirical evidence of this claim. That evidence is now available. Barbara DeFelice, Director of the Digital Resources Program at Dartmouth Library has compiled some statistics from the COPE signatory institutions about their OA fund expenditures, which she discussed at a recent ARL talk. I used her statistics to calculate approximate costs per faculty member per year. The numbers reveal that the outlays are even more manageable than even I had estimated, perhaps by an order of magnitude. (I’ve made these numbers quite conservative by counting faculty conservatively and estimating that each article funded cost $1,500 dollars. The actual average seems to be somewhat less. As more data is collected, I’ll try to make it available.)

For universities that run their OA funds in accordance with COPE recommendations (that is, no hybrid fees, no grant-funded articles), the costs come to not tens of dollars per faculty member per year, but single digit dollars. The outliers are Berkeley and Ottawa, both of which will cover hybrid fees (though Berkeley places tighter caps on fee per article) and will cover grant-funded articles (though they ask for grant funds to be used first).

The bottom line is that the direct costs of running a COPE-compliant open-access fund are trivial, and the administrative costs of dealing with handfuls of requests are trivial as well. Cost should not be an impediment to setting up an open-access fund in this way. In particular, harangues about open-access funds amounting to throwing away large quantities of valuable dollars can please stop now. For instance, Stevan Harnad likes to say things like “COPE is based on the illusion that there is enough money available in institutions today to pay for OA publication in all the must-have journals — Nature, Science, the American Physical Society journals, and all the other top journals — while continuing to subscribe to those journals (and we don’t as yet have OA for their contents, so it’s premature to cancel).” He either misunderstands the compact or willfully misrepresents it, since COPE-compliant funds need not, should not, and generally do not pay publication fees for the subscription journals he lists. COPE does not support “double-dipping”.

The reason that the costs of COPE-compliant open-access funds are so low is because demand for the funds is low because, in turn, there are very few quality OA journals charging publication fees, because, finally, to do so would be to put the journals at a systematic disadvantage in getting authors as compared to subscription journals that don’t charge fees. (This disadvantage is exactly what COPE is trying to remedy.) Here is how the numbers break down. Of the 5,000 or so open-access journals in the Directory of Open Access Journals, only a hundred or two are of the character that these universities’ researchers are likely to publish in them. For example, only a hundred or so are indexed by Thomson ISI for impact factors. Of these, the majority don’t charge publication fees, so can’t contribute to OA fund demand. Those that do charge a fee are overwhelmingly in the life sciences where grant funding is widespread, hence they also don’t generate demand on the OA fund.

If there is so little demand for an OA fund, why have it at all?  The goal of a COPE-compliant OA fund is not short term maximization of access to an institution’s output. (If it were, then hybrid fees would be appropriate to underwrite. But that goal can be accomplished much more cost-effectively by establishing good green open-access policies.) Rather, the goal of COPE is to provide the basis for an alternative business model, should a large number of institutions similarly commit.  If a large number of institutions were to commit to the compact, publishers would have a viable business model through charging publication fees in a way that they do not now have.

In essence, COPE is trying to establish a kind of safety net.  Safety nets are useful even when they are not used.  Safety nets allow people to take risks that we want to promote.  A publisher changing its business model is incurring such a risk.  We, the universities and research funders, may need collectively to build a very big safety net (university by university, funder by funder) to convince a publisher to take that risk. But given that the cost of each of our pieces of the safety net is so incredibly low, it is worth keeping our pieces up and encouraging others to add their pieces, in the hope that a big enough net will encourage the publishers to take the risk to walk the tightrope from the subscription model to the publication fee model. If we are successful, demand for the OA funds will grow as publishers will flip business model to an OA publication fee basis, thereby freeing up funds to pay those fees. If we are not successful, at least the costs are negligible.

(Thanks to Barbara DeFelice for collecting the data and making it available.)

[Update 11/16/2011: I’ve added a post on how funding agencies might best set up their part of the safety net.]

I recently had a conversation with someone (I’ll call him D) whose opinion I greatly respect, a staunch supporter of broadening access to the scholarly literature, who expressed a view I was quite surprised about. D is of the opinion that the publication fee business model for open access journals is economically flawed, so flawed that he thinks it is not worthy of support, for instance through university and funder commitment to underwrite publication fees as envisioned here and here and in nascent implementation.

D thinks that the only worthy alternative is what is sometimes called the “public access” business model, in which access is limited to subscribers but only for a short embargo period of, say, six months or a year. This model can effectively be mandated by having funders require public access for articles they fund.  The NIH policy is an example of a public access mandate, and the pending FRPAA legislation would broaden the mandate to essentially all US-government-funded research.

Certainly, I am a staunch supporter of public access mandates, and the shorter the embargo period the better. Harvard University is on record supporting the approach, as am I. But I also think that the publication fee model is viable, and indeed preferable in the long term.  Thinking about D’s argument has led me to write this post explaining my contrary view. Read the rest of this entry »

The University of North Texas is engaged in a laudable process of designing an open-access policy for their community. Draft language for their policy is now available at their site on open access; the most recent version is as of June 7.

They are pursuing a Harnad-style ID/OA policy, requiring deposit of articles but pursuing distribution only to the extent that publishers’ agreements allow. Although I prefer a Harvard-style approach, this is also quite a good and reasonable approach. I have some concerns though about the details of the UNT working out of the policy. In particular, they have incorporated into the statement of the policy some language from the Harvard-style policies, which doesn’t sit comfortably with the basic approach they have taken. Below, I argue that the intention of their policy can be more consistently specified, and in the process greatly simplified.

Here is the operative language in the current policy draft. (There is much more to the policy document, but the core is described in these paragraphs.)

In support of greater access to scholarly works, the UNT Community Members agree to the following for peer-reviewed, accepted-for-publication journal articles:

1. Immediate Deposit: Each UNT Community Member deposits a digital copy of his/her accepted manuscript, no later than the date of its publication. Deposit is made into the UNT Libraries scholarly works repository. Deposit of manuscripts provides long-lasting protection and preservation of scholarly works from loss due to computer failures.
2. Open Access/Optional Delayed Open Access: The author is encouraged to make the deposit available to the public by setting access to the deposit as Open Access Immediately Upon Deposit (the default). Upon express direction by a UNT Community Member for an individual article, the Provost or Provost’s designate (e.g., the Scholarly Communication Officer) will adjust the Open Access Immediately Upon Deposit requirement to align with the UNT Community Member’s request and/or to align with publishers’ policies regarding open access of self-archived works.
3. Licensing: Each UNT Community Member grants to UNT permission to make scholarly peer-reviewed journal articles to which he or she made intellectual contributions publicly available in the UNT Libraries Scholarly Works Repository for the purpose of open dissemination and preservation, subject to Open Access option selected above. Each UNT community member grants to UNT a nonexclusive, irrevocable, worldwide license to exercise any and all rights under copyright relating to his or her scholarly articles, in any medium, and to authorize others to do so, provided that the articles are not sold. The Provost or Provost’s designate (e.g., the Scholarly Communication Officer) will waive application of the license for a particular article upon express direction by a community member.

The first paragraph mandates deposit of the manuscript. The second specifies distribution consistent with publishers’ policies. So far, this is the ID/OA policy.

The third paragraph seems to specify rights retention à la a Harvard-style policy. In fact, a policy consisting of just paragraphs 1 and 3 would more or less constitute the Harvard policy. But why would broad rights need to be retained (paragraph 3) if no distribution beyond what publishers will allow (paragraph 2) is envisioned? Indeed, such broad rights retention may well lead to the possibility that the policy would inherit the problematic aspects of the Harvard-style policy (occasional requirements by publishers to get waivers of the license), without its advantages (wholesale rights retention for a broad swath of articles independent of publishers’ policies). There is an inherent tension between the second and third paragraphs that acts to the detriment of the policy. In summary, if you’re not going to use rights beyond what publishers allow, don’t retain them. It just muddies the water.

This might argue for merely dropping paragraph 3 and going for a straight ID/OA policy, and that would be one reasonable approach. (Another reasonable approach would be to drop paragraph 2 and go for a Harvard-style policy, but this is clearly not UNT’s intention.) However, there is one advantage to the rights retention aspect, namely that once the policy is enacted, no further effort will be needed for the university to acquire by default whichever rights the publishers do allow. But this advantage can be maintained without the broad language of the third paragraph, and in so doing, the entire policy can be simplified considerably.

My recommendation would be to simplify the third paragraph to grant only those rights needed to do what the policy envisions. Something along the following lines would work:

Licensing: Each UNT Community Member grants to UNT permission to make scholarly peer-reviewed journal articles to which he or she made intellectual contributions publicly available in the UNT Libraries Scholarly Works Repository for the purpose of open dissemination and preservation subject to publishers’ restrictions. In legal terms, each UNT community member grants to UNT for each of his or her scholarly articles a nonexclusive, irrevocable, worldwide license to exercise those rights under copyright that the author retains in any agreements with the article’s publishers. The Provost or Provost’s designate (e.g., the Scholarly Communication Officer) will waive application of the license for a particular article upon express direction by a community member.

Once this restricted license is in force, UNT would have rights to distribute as widely as the publisher allows but no more, as is clearly the policy’s original intent. At that point, the second paragraph (encouraging the author to allow distribution as broadly as possible) would no longer be needed. The university could already do so based on the license it was granted. If an author didn’t want the university to distribute the article for any reason (as envisioned in the second paragraph phrase “to align with the UNT Community Member’s request”), the waiver aspect of the licensing clause would already allow for this.

Presumably, the number of waivers generated under this approach would be minuscule, as publishers would have no incentive whatsoever to require a waiver for publication; the policy involves no license to the university beyond what a publisher would already allow.

While we’re adjusting language, I’d also recommend dropping the sentence providing motivation for deposit, which is appropriate for the explanatory material about the policy, but not its formal statement.

With these changes, the UNT policy becomes much simpler, more consistent, and would be what it was presumably envisioned as, an ID/OA policy with a built-in license to make sure that the university could distribute articles to the extent publishers allowed. This approach is an interesting variant open-access policy to be considered by other institutions in addition to the original ID/OA and Harvard-style approaches.

For completeness, the language I would propose (at least while hewing as close to the original UNT language as possible) would be something like the following:

In support of greater access to scholarly works, the UNT Community Members agree to the following for peer-reviewed, accepted-for-publication journal articles:

1. Immediate Deposit: Each UNT Community Member deposits a digital copy of his/her accepted manuscript, no later than the date of its publication into the UNT Libraries scholarly works repository.
2. Licensing: Each UNT Community Member grants to UNT permission to make scholarly peer-reviewed journal articles to which he or she made intellectual contributions publicly available in the UNT Libraries Scholarly Works Repository for the purpose of open dissemination and preservation subject to publishers’ restrictions. In legal terms, each UNT community member grants to UNT for reach of his or her scholarly articles a nonexclusive, irrevocable, worldwide license to exercise those rights under copyright that the author retains in any agreements with the article’s publishers. The Provost or Provost’s designate (e.g., the Scholarly Communication Officer) will waive application of the license for a particular article upon express direction by a community member.

I must mention that I am not a lawyer, and have not vetted my proposal with any lawyers, and am not making any claims about the legal force or appropriateness of the original or modified language. That is the job for the UNT General Counsel’s Office.

Sandy Thatcher feels “very uneasy about the massive postings of Green OA articles at sites like Harvard’s, which given that university’s great prestige may well lead to the widespread appropriation of those versions by scholars who find it easier to access them OA than to hunt down (and perhaps pay for) the final versions.” He should rest assured that we make every effort to make clear what version we are providing and where the version of record resides. We provide links to the version of record (when available) on the metadata page for each article (see here for a sample), and have even modified the DSpace software that runs our repository so that it provides users with links to the version of record on search results pages (like this) before they even get to the metadata page for the article. We provide citation information and links to the definitive version on the metadata page as well as on a front page added to the PDF for downloaded articles (for instance, this PDF). The PDF link is even clickable to go to the publisher’s site for the version of record. In short, we try to make it as easy as possible to “hunt down” the version of record.

Calling the mere use of an article in the repository an “appropriation” seems tendentious. To appropriate is “to take possession of or make use of exclusively for oneself, often without permission.” But in this case, there is nothing exclusive about the use of the articles, and permission is provided for. There is no inappropriate taking going on in DASH, or even in the Harvard OA policy, which allows for waivers of the license to Harvard. Publishers can feel free to institute and enforce policies to require waivers of the license for articles they publish if they fear that it might harm their business model — though few have done so. I expect many publishers appreciate that Green OA is not really the big problem for their business model.

On the other hand, I second the sentiment expressed by Dr. Thatcher that he “look[s] forward eagerly to the day when OA fully takes over the dissemination of scholarship…partly because it will solve the problem I have with Green OA now.” I agree that Green OA is a short term mitigation of an underlying problem that needs a fuller solution involving modifying the scholarly communication system in general.

Twenty-seven university presidents and provosts have posted an open letter in support of FRPAA. The list of institutions includes Harvard, Dartmouth, Princeton, Cornell, Duke, Stanford, Tulane, Rutgers, Indiana, two campuses of the University of Texas, and the University of California system and two of its individual campuses. The letter echoes an earlier letter supporting the 2006 version of the bill. The original posting is hosted at the web site of Harvard’s provost (one of the signatories), and Harvard’s Office for Scholarly Communication provides a copy as well.

An excerpt:

The United States Congress will have the opportunity to consider the Federal Research Public Access Act (FRPAA). FRPAA would require Federal agencies whose extramural research budgets exceed $100 million to develop policies ensuring open, public access to the research supported by their grants or conducted by their employees. This Bill embodies core ideals shared by higher education, research institutions and their partners everywhere. The Bill builds upon the success of the first U.S. policy for public access to publicly funded research – implemented in 2008 through the National Institutes of Health – and mirrors the intent of campus-based policies for research access that are being adopted by a growing number of public and private institutions across the nation.

We believe that this legislation represents a watershed and provides an opportunity for the entire U.S. higher education and research community to draw upon their traditional partnerships and collaboratively realize the unquestionably good intentions of the Bill’s framers – broadening access to publicly funded research in order to accelerate the advancement of knowledge and maximize the related public good. By ensuring broad and diverse access to taxpayer-funded research the Bill also supports the intuitive and democratic principle that, with reasonable exceptions for issues of national security, the public ought to have access to the results of activities it funds.

The broad dissemination of the results of scholarly inquiry and discourse is essential for higher education to fulfill its long-standing commitment to the advancement and conveyance of knowledge. Indeed, it is mission critical. For the land-grant and publicly funded institutions among us, it addresses the complementary commitment to public service and public access that is included in our charters. In keeping with this mission, we agree with FRPAA’s basic premise that enabling the broadest possible access to new ideas resulting from government-funded research promotes progress, economic growth, and public welfare. Furthermore, we know that, when combined with public policy such as FRPAA proposes, the Internet and digital technology are powerful tools for removing access barriers and enabling new and creative uses of the results of research.

The FRPAA bill — S.1373 in the Senate — has just been introduced into the House as HR.5037. The bill calls for federal agencies to “develop public access policies relating to research conducted by employees of that agency or from funds administered by that agency.” You can register your support for the bill.

Two years to the day after the Faculty of Arts and Sciences became the first school at Harvard to vote an open-access policy, the Harvard Business School enacted their own policy on February 12, 2010, becoming the fifth Harvard school with a similar policy. Under the HBS policy, Like the previous policies, faculty agree to provide copies of their scholarly articles for distribution from the university’s DASH repository and grant the university a waivable license to distribute the articles.

HBS is the second business school to fall under such a policy. MIT’s Sloan School of Management is covered by the similar MIT policy that was enacted March 18, 2009.

Correction (March 1, 2010): The HBS policy is the third OA policy of a business school, not the second, by virtue of the predating policy of the Copenhagen Business School of June 2009. Thanks to Stevan Harnad and Peter Suber for pointing out the error.

(Image of Baker Library at Harvard Business School via Wikipedia.)

Harvard’s provost Steven Hyman has submitted a response on behalf of the university to the White House Office for Science and Technology Policy’s RFI on public access policies. It should appear on the OSTP blog within a day or so, and is duplicated here as well. I am in strong agreement with the recommendations of Provost Hyman, and hope the White House process will lead to actual policies promoting open access to government-funded research. Read the rest of this entry »

Several publisher representatives have recently asked about why the Harvard open-access fund does not cover hybrid fees. I thought I’d explain my thinking on this issue, though I am certainly not doctrinaire when it comes to institutional underwriting of hybrid fees, and am perfectly in accord with institutions coming to different decisions on the issue.  In fact, as I mention below, I can imagine (counterfactual) situations in which I would support Harvard’s open-access fund underwriting hybrid fees.

In the case of the current Harvard policy, I’ll explain why for the time being at least we are not underwriting hybrid fees.  My comments extend the arguments I give in my PLoS Biology paper.

Hybrid fees would be worth supporting to the extent that (as their proponents claim) they provide an approach to smoothly transitioning to an open-access publication fee business model.  The argument is that as uptake increases in payment of the hybrid fees, revenue smoothly shifts from the reader side to the author side in a revenue-neutral way.  At the end, the journal is openly accessible, with hybrid fees (now become just publication fees) providing the revenue.  To the extent that the hybrid system works this way, there is no “double dipping” — using hybrid fees as a way to increase revenue rather than transitioning to OA in a revenue-neutral manner.  I’ll call such a model “true hybrid”.

There are two problems with supporting hybrid fee systems.  First, they may not be set up in the way just described as true hybrids, and determining whether this is the case may in fact not be possible.  Second, even if they were set up in this way, game-theoretic problems may inhere in the system that prevent them from realizing the transition.

Whether “double dipping” is going on is independent of when the hybrid option is set up.  In particular, whether a journal is hybrid from the start or a hybrid option is added well after the journal is founded is irrelevant.  What matters is the dynamics of how hybrid uptake affects subscription fees and revenues.  To be a true hybrid model, hybrid fees must be set at a rate no more than the true average revenue per article, so that universal uptake of hybrid payments is not a means to revenue enhancement.  Further, uptake of hybrid fee revenues must be directly offset by reductions in subscription fee revenues, so that incremental uptake of hybrid payments is not a means to revenue enhancement.

True hybrid journals must therefore have a degree of transparency that is difficult to imagine achieving.  The journal would have to demonstrate linearly decreasing subscription fees in direct proportion to the hybrid fee uptake.  Determining hybrid fee uptake is straightforward.  But determining subscription fee reduction is not.  Publishers practice price discrimination, bundling, and price changes over time, which separately and together make it impossible to tell what a subscriber’s costs would have been absent the hybrid fee discount.  The issue of transparency is not all or nothing.  Certainly some publishers, Oxford University Press for instance, may be more transparent than others in their hybrid discounting practice.  But determining whether appropriate reductions are taking place is difficult at best.

Even if there were some way to determine that a journal were a “true hybrid”, receiving no incremental revenue from its hybrid program, there is a more fundamental problem with the normal way in which hybrid programs are implemented, viz., that subscription revenue reductions are shared among all subscribers.  As institutions decide whether or not to underwrite hybrid journals, they are confronted with a kind of prisoner’s dilemma.

In a prisoner’s dilemma, a set of agents can either cooperate or defect.  If they all cooperate, all are better off.  But if any defect, all are worse off except for the defector.  In the case at hand, a cooperating institution underwrites hybrid fees, a defecting institution does not.  Given a status quo ante (as we have) of all institutions defecting, all would be better off if they cooperated, since they would pay the same amount (true hybrids being revenue-neutral) but would gain open access.  However, at the margin, an institution cooperating sees increased costs from the hybrid fees, but only a share (and a small one at that) of the subscription revenue reduction.  Defectors see a reduction in their subscription costs relative to the status quo ante at no cost to themselves.  Though both fully defecting and fully cooperating are equilibria,^[1] there is no transitional path from the former to the latter.  This is an inherent flaw in the hybrid system as traditionally conceived.  Empirical evidence accords with this analysis; uptake of hybrid fees has been exceptionally low.

It may be that institutions can be convinced to act temporarily against their financial interests in the expectation that others will join them and the cooperating equilibrium will be reached.  Especially in the current economic climate, I’m not sanguine about this prospect.  And without such temporary action against self-interest, the hybrid model is no transitional model at all.  If however, evidence accumulates that the transparency problem can be addressed and true hybrids can be identified and institutions appear increasingly willing to pay these fees against temporary self-interest, I can imagine that the Harvard fund might then change its policy for support.  Similarly, if changes to the hybrid model were made that eliminated the prisoner’s dilemma — I believe that this may be possible, and hope to talk about it in a future post — then again hybrid fees ought to be supportable for true hybrid journals.

^[1]Actually, full cooperation is not an equilibrium per se, as each agent still has an incentive to defect from cooperation even at that point. However, once all agents (or even most) were to cooperate, the hybrid system could be jettisoned for a true OA model that effectively removed the defecting alternative.

Pride does not wish to owe and vanity does not wish to pay.
—Francois De La Rochefoucauld

Open-access journal publishing has been criticized on a whole range of grounds as being unsustainable, unfair, or ineffective.  Perhaps the starkest criticism is that open-access journals amount to a vanity publishing industry, and will exhibit a “race to the bottom” in which journals compete to lower editorial standards to capture the revenue for publishing articles.  Is open-access journal publishing prone to the problems of a vanity press?

There are both theoretical and empirical arguments that the concern is unfounded. From a theoretical point of view, the prerequisites for vanity press are not found in scholarly publishing.  From an empirical point of view, current open-access journals display a pricing structure that does not indicate a vanity press industry, as we demonstrate below in a new analysis of OA publication fee data. Read the rest of this entry »