Open Access: Doing the numbers

One question that has been repeatedly (and heatedly) debated since 1994 — when Open Access (OA) advocate Stevan Harnad first posted his "Subversive Proposal" — is the questions of costs. That is, what are the essential costs of publishing a scholarly paper? To date no one appears to have come up with an adequate answer.

For OA advocates this is all rather unsatisfactory, since many believe that one of the primary reasons for embracing OA is that it will resolve the journal price inflation problem that has now plagued the research community for several decades. The hope has been that OA will somehow squeeze out all unnecessary costs, and resolve the so-called "affordability problem".

As OA has grown in popularity, however, so it has become increasingly evident that Gold OA publishing could prove just as expensive (and inflationary) as the traditional subscription model. When OA publisher Biomed Central (BMC) started operating in 2001, for instance, it set its article-processing charge (APC) at $525; today BMC charges from $1,700 to $1,900 to publish a paper. Similarly, when Public Library of Science (PLoS) launched its first journal it charged an APC of $1,500; today it charges from $2,100 to $2,750.

The same inflationary effect has been evident with the institutional membership schemes that many OA publishers now offer as an alternative to charging an APC — leading to some controversy, and a very public withdrawal from the BMC scheme by Yale University.

But does this mean that OA publishing will turn out to be just as expensive as traditional subscription publishing? We don't know, not least because it is still not possible to say with any authority what it costs to publish a scholarly paper, let alone how much it costs to undertake each of the individual components of that process. Without this information we can't know whether OA publishing is likely to be more or less expensive, or whether current costs are reasonable and fair.

What we do know is that there is growing concern that the research community may end up simply moving from a situation in which it cannot afford to access all the scholarly papers it needs to, to one in which it cannot afford to publish all the scholarly papers it produces.

People have, of course, tried to crunch the numbers. In 1997, for instance, mathematics professor Andrew Odlyzko estimated that it was costing the research community around $4,000 to publish a paper. In reality, he concluded, the task could be done for as little as $300 to $1,000.

More recently (last month) the UK-based Research Information Network (RIN) estimated the "average total publishing and distribution costs per article" to be around £4,000 ($7,800) today. The report added that moving from a subscription-based publishing model to an OA publishing model would see a fall of £2.91 billion in the subscription prices paid by libraries, but that these savings "would be offset by an increase of £2.92 billion in the charges that the academic and research institutions of which they are part (or their funders) would have to meet in author-side publication fees". The end result, RIN concluded, would be that "academic institutions at a global level would need to fund an additional £10 million from the move to author-side payment."

But the problem with much of this number crunching is that it is generally the product of little more than back-of-the-envelope guestimates, not informed analysis, since it is invariably done by people who are able only to look through the window of the scholarly publishing business, not by those actually working in the industry. And it is only the latter that have access to the necessary data to make accurate assessments of costs. Unfortunately, most publishers are extremely reluctant to share any of their data with the outside world.

Recently, for instance, I posted a question on the Liblicence mailing list asking if anyone had done any research to establish the costs of a) implementing peer review and b) distributing a paper electronically — which some would argue are the only two essential costs in an OA environment. One of those who replied to my post was publishing consultant Joe Esposito, who responded, "All the figures Richard Poynder is looking for have been developed and redeveloped by commercial (and some not-for-profit) publishers over the years. Doing this analysis is simply part of what it means to run a business. Of course, this information is proprietary."

However, since it is primarily public money that is used to fund scholarly publishing, and historically publishers have been criticised for squeezing as much as a 35% profit margin out of the process, one might question whether that is a good enough answer.

Fortunately, at least one publisher is prepared to be more transparent: When I asked the American Physical Society (APS) how much it costs APS to publish a paper, the organisation's treasurer/publisher Joe Serene not only produced a figure, but agreed to break it down for me as well.

In total, Serene said, in 2007 it cost APS approximately $1,500 to publish the electronic version of a paper (with all print-related costs excluded), roughly 20% ($300) of which can be apportioned to each of the following functions:

• Editorial costs (including peer review)

• Electronic composition and production

• Journal information systems, "which support everything from manuscript receipt through electronic posting, mirroring, and archiving of the published papers"

• Central publication management

• Essential overhead expenses

Serene cautioned, however, that these financial categories are not functionally independent. For example, he said, APS receives approximately 35,000 manuscripts per year, and an effective central publication management system is essential for efficient (or simply non-chaotic) operations.

Do these costs help us assess whether OA publishing will be cheaper than the traditional subscription model? Since the answer will partly depend on whether you believe that all the above functions are necessary in an OA environment, and whether some or all could be reduced or streamlined, it should certainly help to have the figures broken down in this way. This in turn might lead people to want to explore in more detail what the various functions consist of, and how they are currently carried out, but at least we now have some real-life figures in the public domain to match against the various tasks associated with publishing a paper.

On the downside, having the figures from just one publisher is not enough in itself. What would help would be for other publishers to be as transparent as the APS. OA advocates would then be much better equipped to debate the issue in an informed and constructive manner.

One thing to note in the above figures, by the way, is that authors wishing to opt for the APS' "Free to Read" OA option are charged a $975 APC for articles in Physical Review A-E, and a $1,300 APC for Letters in Physical Review Letters. Serene points out, however, that these charges were purposely set below cost in order to encourage initial use of Free to Read, with the understanding that they would have to be raised if a significant number of authors were to chose this option; so far the use of Free to Read has been very low.

Perhaps the take-home point here is that either everyone has consistently underestimated the true costs of publishing a scholarly paper, or publishers (both traditional subscription publishers offering an OA option and pure OA publishers) still have some way to go in reducing their costs if OA is to prove more affordable than the subscription system.

Consider that at a workshop held at CERN in 2001 participants concluded that the cost of editing and processing an article could fall as low as €500 ($775 at today's rate) in an OA environment. (Although admittedly that estimate did not take into account any of the overheads associated with running a publishing organisation).

More significantly perhaps, as we noted above, within their short lifetimes both BMC and PLoS have increased their prices considerably — in the case of BMC rates have nearly quadrupled in some instances. This is all the more striking when you consider that when I spoke to BMC founder Vitek Tracz in 2006 he predicted that costs would fall. As he put it, "More and more of what we do for authors today they will be able to do for themselves in the future, and as we develop more tools to allow them to do it themselves, so what we charge them will be less and less."

So the question remains: Can OA reduce the costs associated with scholarly communication? If so, how, and when? If not, what are the implications of this for the "scholarly communication crisis?" These are important questions. But without accurate numbers to crunch we really cannot answer them adequately. Wouldn't it be great therefore if other publishers decided to be as "open" as APS in discussing their costs?

One thing is for sure: If OA ends up simply shifting the cost of scholarly communication from journal subscriptions to APCs without any reduction in overall expenditure, and inflation continues unabated, many OA advocates will be sorely disappointed. And if that were to happen, then we can surely expect to see calls for a more radical reengineering of the scholarly communication system.

I will close by pointing out that some OA advocates respond to any discussion about the costs of OA publishing by arguing that most OA journals don't actually charge an APC today. Others, meanwhile, insist that it is far too early to worry about Gold OA, since researchers can quite easily continue publishing in subscription journals and then self-archive their papers on the Web themselves — thereby achieving OA at no cost to them or their institutions (leaving aside the subscriptions that their institutions currently pay in order to buy access to research produced by other institutions). But there are reasons for arguing that these responses are not entirely satisfactory — as I hope explain in a future post.

In the meantime, I welcome comments from others, either via the comment button below, or by email to richard.poynder@btinternet.com.