(This is the text of a talk — with only two slides! — given at the Researcher to Reader Conference in London, 20 February 2017.)

The question at hand is this: is there good reason to expect that open access (OA), and particularly OA of the Green variety, is likely to lead libraries and other customers to cancel their paid journal subscriptions? The context in which we have to ask this question is, quite frankly, one of dire economic straits in many if not most academic libraries. Around the world, and certainly in the US and UK, library budgets are largely flat, some are falling, and in those unusual circumstances in which annual budget increases actually do happen, they are hardly ever sufficient to keep up with price increases levied by publishers, especially journal publishers and especially those publishing in the STEM disciplines.

Of course, libraries have been saying this for years, even decades, and we have now cried “Wolf!” so often that most of the publishers I know have simply stopped taking us seriously. Why, they ask, if our circumstances are so dire, do we keep renewing our subscriptions? Why have publishers not seen massive (or at least sustained and growing) cancellations from libraries?

I believe there are two answers to this question: one that explains subscription persistence up until the past few years, and one that explains it since a few years ago.

Up until a few years ago, I think the persistence of library subscriptions can be substantially explained by the Big Deal, which did two things: first, it greatly reduced the unit cost of journals and articles for the individual library by adding large amounts of content to that library’s existing subscription base at a very low marginal cost increase. It is important to point out here, however, that the Big Deal does nothing to save libraries money. It doesn’t lower costs; it only increases the value realized in return for the cost, even as the absolute cost goes up. The Big Deal, in most cases, lowers tremendously the library’s cost per article – however, cost per article is not actually a cost measure; it’s a value measure. The cost measure — and the only cost measure — is the figure on the invoice, and the Big Deal in the 1990s made that number bigger, not smaller. However, the second (and more significant) thing that the Big Deal did to forestall cancellations was to make cancellation more difficult, by fencing off large swaths of journal titles and making them uncancellable. The emergence of the Big Deal, I believe, goes a long way towards explaining the persistence of journal subscriptions even as libraries’ collection budgets have gotten tighter and tighter – at least up until a few years ago.

Whenever anyone suggests that libraries might cancel their paid subscriptions to journals whose content is freely available online, they are shouted down – not by frightened publishers, but by outraged OA advocates.

It is worth pointing out here that hardly anyone ever believed that the Big Deal was going to be a sustainable arrangement for the long run. The Big Deal’s ongoing viability depends on a greater and greater portion of the library’s budget being redirected from other purchases into the Big Deal, and the inevitable end game is that the library ends up subscribing to nothing but one Big Deal. This inevitable scenario has already begun playing out in the last few years, as more and more academic libraries have begun shifting money out of their book budgets and into their serials budgets. This is the second explanatory factor for subscription persistence, the one that I believe explains that persistence over the past five years or so: in fact there have been journal cancellations, just not in massive numbers, and on most campuses I think it’s fair to say that just about every individual journal subscription that can be cancelled without the faculty burning down the library has been cancelled, and we are now buying fewer and fewer books in order to keep feeding the insatiable beast that is the Big Deal, as well as any truly essential standalone subscriptions that may remain. We are able to get away with this, for now, because so many of the books we buy get so little use, whereas journal content tends to be heavily used – this is particularly true in large research libraries, and is probably less the case at smaller institutions that focus more heavily on undergraduate education and on the arts and humanities.

However, even this gambit has started to lose its effectiveness, as we can all see it has to eventually. And indeed, the wolf about which we have been crying for decades now seems to be well and truly upon us. In the past year or so we have seen journal cancellation projects in the amount of $500,000 at the University of Wisconsin-Milwaukee, $350,000 at Towson University, $1.5 million at the University of Calgary, $468,000 at the University of New Mexico, $200,000 at the University of Missouri, and $135,000 at Colorado State University, just to name a few. The University of Maryland cancelled 2,000 journal titles but didn’t provide a dollar figure publicly; CalTech, similarly, recently cancelled 650 titles. At the University of Utah, where I work, a planned $300,000 journal cancellation was avoided at the last minute by the discovery of some available one-time money – but of course when it comes to subscriptions, one-time money only delays the inevitable, and not for very long.

Furthermore, Big Deals themselves may be losing their iron grip on library budgets. In recent years the conventional wisdom has been that libraries love to talk about cancelling their Big Deals, but almost always end up deciding to stay with them – either because the publisher offers new terms, or because the faculty revolt. That may be changing. A few weeks ago I posted a query to a couple of listservs. I asked that listmembers share with me the names of libraries that had publicly announced plans to abandon the Big Deal, expecting that the great majority of them would have changed their minds. The data is messy and I’m still sorting it out, but right now I can report that a surprisingly high number of research libraries have recently abandoned Big Deal packages from Springer, Wiley, Elsevier, Taylor & Francis, SAGE, Oxford University Press, and Cambridge University Press. I don’t know whether this is a trickle that will dry up or one that presages a flood — I only know that the data coming in so far suggest a growing willingness on the part of libraries to break away from their Big Deal arrangements.

This, then, is the context into which we have to place any discussion of OA and library subscriptions: tighter budgets and inexorably-rising journal prices leading to greater scrutiny of existing subscriptions, which inevitably means more cancellations. The number of cancellations grows as the money available to siphon away from book purchases runs out.

OA does have the potential to provide libraries with some relief from this situation. But only certain manifestations of OA will do that.

In this difficult circumstance, OA promises to come to the rescue. However, the rescue scenario is not simple.

OA does have the potential to provide libraries with some relief from this situation. But only certain manifestations of OA will do that. What does not help libraries, financially speaking, is the emergence of new Gold OA journals, whether supported by APCs or by institutional subvention. We certainly welcome the appearance of these journals, but they do nothing to relieve the financial pressure we are under. Why? Because the journal marketplace is a marketplace of complements, not a marketplace of substitutes. The emergence of, for example, a new Gold OA journal that competes with The Lancet does not lead our faculty and students to give us permission to cancel our Lancet subscription. What they will want us to do instead is add the new Gold OA journal to our online journal list, while continuing to support the Lancet subscription as well. Why? Because each of the two journals provides a unique set of content — they report on different studies of different medications and interventions performed and written up by different researchers. The emergence of a new Gold OA journal is a good thing, because it adds to the fund of publicly-available science — not because it saves the library any money.

The only OA scenario that can actually save money for the library (and individual subscribers) is one that allows them to cancel subscriptions. If the emergence of new Gold OA journals doesn’t provide us that opportunity, then what does? The answer, of course, would be an OA model that, instead of providing new and additional content on an OA basis, makes existing toll-access content available on an OA basis. What does that? Not Gold OA, but Green OA. By making content from toll-access journals freely available in repositories, Green OA has the capacity to free libraries and other subscribers from the necessity of paying for access to that content.

To repeat, then: the financial relief that OA can give libraries and other subscribers is relief from subscription bills. And there was a time, in my callow youth, when I thought that such relief was actually a major part of OA’s purpose – that a pillar of OA’s value proposition was that it is designed to make it so that libraries and other subscribers would have access to scholarship without having to pay for it. But apparently I was mistaken, because whenever anyone suggests that libraries might cancel their paid subscriptions to journals whose content is freely available online, they are shouted down – not by frightened publishers, but by outraged OA advocates.

What do these outraged advocates say when they’re shouting down those who make this suggestion? The most common arguments are these:

First, that there is “no evidence that Green OA negatively affects subscriptions.” This is one of the most reliable talking points of SPARC. But if you take the time to stop and think about it, it becomes clear that this answer is a non sequitur. It responds to a concern about the future by pointing out that the future hasn’t happened yet, and by implying (though not explicitly saying) that we have no reason to believe the future will be different. In other words, the implicit argument here is that since Green OA has not yet demonstrably affected the existing subscription base, to suggest that it will ever do so is unreasonable. But the conditions that have protected subscriptions up until now – a Green OA landscape characterized by inconsistent and delayed access to inferior versions of articles – are not, as I understand it, conditions that most OA advocates find acceptable. My understanding is that the OA movement is working to move us towards a future in which the landscape would be characterized by consistent and prompt access to versions of record, or at least pre-publication versions that don’t depart in any material way from the versions of record. This scenario is the one that poses a challenge to the ongoing viability of subscriptions. Of course there’s no data proving that this scenario leads to cancellations – you can’t gather data on a circumstance that doesn’t yet exist. But it’s also very difficult to imagine that as we get closer to this scenario, libraries and other subscribers will remain just as willing to pay for access as they were under the old scenario – one in which consistent and prompt access to versions of record was only available for a price.

The only OA scenario that can actually save money for the library (and individual subscribers) is one that allows them to cancel subscriptions. If the emergence of new Gold OA journals doesn’t provide us that opportunity, then what does?

And this is where the real problem with Green OA – successful Green OA, that is, as opposed to crap Green OA – becomes clear. Unlike Gold OA, Green OA is absolutely dependent on the ongoing vitality of subscription journals. Without those journals, there is no Green OA. And the more successful Green OA is, the less incentive anyone has to continue paying subscription fees.

This brings us to the second thing that OA advocates often claim when they are trying to get people to stop saying that Green OA will undermine journal subscriptions. There are some who predict that libraries will continue paying for access even though they don’t have to, for the Good of the Order. This, to me, seems like the scenario for which we would need to see some evidence before we decide to rely on it to keep the economic structure of journal publishing afloat. As we all know, remarkable claims require remarkable evidence. The idea that people will refrain from paying for free content is not a remarkable claim; the idea that they will pay for free content is a remarkable claim.

Now, there actually is some evidence that libraries are willing to do this. My library is one among many that willingly pay several thousand dollars a year in order to keep the arXiv afloat, despite the fact that it costs nothing to access. The temptation to be a free rider in this case is real, but our library resists it and so do many others. However, in our case (and, I’m certain, in the case of every other library that willingly contributes to the maintenance of the arXiv) we also receive many other appeals to contribute to projects and initiatives that contribute to the greater good by making high-quality scholarship or scholarly services freely available to all. In the great majority of cases, we decline to contribute — not because we doubt the goodness and value of these projects, but because our resources are limited and we have to make difficult choices, just as we do when choosing between desirable journal subscriptions.

There’s another important factor in our decisionmaking as well, and that is alignment with our host institution. The money we spend is the university’s money, not our own, and we are held accountable for the spending decisions we make. In the past, the accountability has been fairly loose; the university gives us a certain amount of money with which to buy content, and unless administration are given good reason to believe that we’re spending it very badly, they have been willing to trust that we’re spending it well. If, however, our university were to get wind that the library is spending significant amounts of university funding on content and services that actually cost nothing, we would expect to have to answer some questions. And if our answer to the question “Why are you spending so much money on journals that don’t cost anything” is “We’re doing it for the greater good,” our administration might reasonably start wondering whether the greater good is better served by continuing to supply the library with money or by refurbishing the university’s physics labs, providing more scholarships to underprivileged students, or hiring more faculty. Again: the money academic libraries spend is not their own. It’s university money, vouchsafed to libraries for the purpose of meeting university goals and priorities. If librarians take it upon themselves to use that money for purposes that do not, in administration’s view, directly support those goals and priorities, there is a strong likelihood that the money will no longer be forthcoming. All of us in the room may agree that university administrations ought to care very much about making scholarship freely available to the world. But what will affect the library’s future is not what administrators should want, but what they actually do want.

Now, granted that Green OA exists now, and has existed for years, and doesn’t seem to have had a dramatic effect on subscriptions, we should ask ourselves what it would take for that effect to occur. I would like to propose two possible future scenarios, one of which will tend to leave the subscription base more or less stable, and the other of which will tend to undermine it dramatically.

The low-cancellation scenario is one in which the future looks very much like the present: Green OA is distributed spottily and unreliably; providing public access to articles is largely optional rather than mandatory; free access is delayed by a year or more after formal publication; articles available on this basis are relatively hard to locate; and the versions available for free are generally considered to be at least somewhat inferior to the formally-published versions of record. This is the world in which we now live, and it is the one in which there is no strong evidence of subscription cancellations in favor of Green access – for the obvious reason that Green OA has not yet been successful enough to pose a threat to toll access, and therefore is not yet achieving its promise.

The high-cancellation scenario is one in which access to deposited articles is pervasive and reliable, either because authors have willingly adopted the practice in large numbers or the choice of whether to do so has been taken away from them by funders or employers or government; articles are freely available without embargo, and are easy to find; and the versions deposited are identical to the versions of record in every material way. This is a scenario in which readers get virtually all of the benefits of subscription without having to pay for a subscription.

If you are an OA advocate, ask yourself this question: which of these two scenarios is the one towards which you are working, and encouraging others to work?

In light of all this, it’s difficult to avoid the conclusion that those who advocate for Green OA as the preferred mode, and as the more or less universal solution towards which we should all be working, have simply failed to think through the implications of this goal. You cannot make something freely and easily available without undermining the commercial market for that thing. Examples from other marketplaces of circumstances in which providing free samples generates additional sales are not relevant here, because Green OA does not provide samples; OA is not the salesperson standing by the produce section handing out small bites of pineapple on toothpicks; OA is putting the whole produce section out on the sidewalk and inviting everyone to take as much as they want. On the other hand, the “freemium” model might possibly give us an idea of what to expect in a comprehensively Green OA world; this model is financially sustainable to the degree that it leads customers to buy an enhanced version of the free product. However, for this model to work in the context of scholarly communication, the OA version would have to be meaningfully inferior to the toll-access version. This is an arrangement more likely to be acceptable to publishers – but I’m not sure that it’s really what the OA community has in mind.

So, what if I’m right, and if Green OA really does increase the likelihood of journal cancellations the more it achieves its stated purposes? One rather difficult question remains: how might libraries go about it? Wouldn’t it be nearly cost-prohibitive to do the research necessary to determine which individual journals can be cancelled?

No, for two reasons. First, that kind of research is only cost-prohibitive in the low-cancellation scenario – that’s one reason it’s a low-cancellation scenario. To the degree that Green OA becomes pervasive and successful, the cost of identifying cancellable titles drops dramatically.

The second reason has to do with both journal cost and institutional priorities. Even today, living as we do under the low-cancellation scenario I’ve just outlined, there are factors that can make it cost-effective for a library to cancel subscriptions based on the existing Green OA landscape. The more expensive a journal is, and the less central it is to the curricular priorities of the institution, the more worthwhile it will be to investigate inferior but free access options. The investigation doesn’t have to be especially time-consuming and doesn’t have to be undertaken by professional librarians in order to produce a strong return on investment. Assign a student employee working on a service desk a list of high-cost journals, and ask him (as patron traffic permits) to take a random sample of issues from that journal and try to find Green OA versions of articles from those issues online. If two hours of that student’s time lead the library to cancel a $2,000 journal (or even a $500 journal), that is certainly time well spent. And if the investigation leads to no cancellations, the opportunity cost of the investigation is practically nil, because you were paying the student to staff that service desk anyway. But let’s take it a step further: what if you pay a student $10 per hour to spend ten hours a week investigating Green OA alternatives to toll-access journal content? If three months of that student’s time results in cancellations totaling $1,300 (or, in other words, one moderately-priced science journal), the library has saved more money than it spent investigating — and that’s only in the first year after the cancellation. The savings continue from year to year after that, since what was cancelled represented an ongoing annual expense.

To sum up then, I would encourage my colleagues who actively advocate for Green OA not to recoil from, or be outraged by, the possibility that it will lead libraries to cancel their subscriptions, but rather to embrace it. After all, if the promise of open access is not the promise of free access, then it’s difficult to see what the point is.