Green Tea and Velociraptors

The term ‘article processing charge’, or APC, is ubiquitous in discussions about Open Access. It refers to the author-facing charge levied by many publishers in order to make an article freely available on their website. Now, putting aside the fact that this system actively discriminates against less-wealthy authors and institutes, I think that the term APC itself is incredibly misleading. Furthermore, I believe that this misdirection occurs in favour of publishers, to the detriment of all other parties. Hopefully in this post, I can explain why, and offer a potential solution to it.

Definition of an APC

Wikipedia defines an APC as: “An article processing charge (APC), also known as a publication fee, is a fee which is sometimes charged to authors to make a work available open access in either an open access journal or hybrid journal.”

The average APC for hybrid journals has been calculated to be almost twice as high as APCs from full Open Access publishers, and APCs range from around $8 to $3,900. This charge must cover two main aspects:

1. The indirect costs of running a publishing business. This includes salaries, web-hosting and maintenance, advertising, and marketing, among other things.
   1. Note that this section also includes profits, which are removed from any of the internal processing and costs by shareholders. For example, with a profit margin of 30%, and an APC of $2100, $630 of that APC goes straight into shareholder pockets.
2. The direct costs of processing and publishing an article (e.g., editorial work, copy editing, typesetting).

The real cost of article processing

Now, I see several major issues with this current system.

• That the same APC is applied indiscriminately to all articles published in a journal, independent of the true indirect and direct costs associated on a per-article basis.
  • As an analogy, this would be like going to a mechanic, and being charged the same universal price to fix your car, irrespective of the actual work involved. An oil change would cost $2000, and so would having your engine replaced.
• In most cases, it is completely obscure what the cost breakdown is for an APC.
  • To use the above analogy, this is like getting the bill for your fixed car, with no idea what had actually been repaired.
• The APC is charged upon article acceptance, but the fee is decided on before an article is even submitted. It is therefore independent of anything which occurs subsequently to once the fee is set.
  • Again, this is like getting a bill from a mechanic for $2000, before they have even told you what the problem is or fixed it.

So it doesn’t make much sense at many levels.

Given the term “article processing charge”, we might logically expect this definition to explicitly be about the cost of processing an individual article. Especially so since they are levied on a per-article basis (although now also often in bulk, which makes even less sense). However, it is clear that an ‘APC’ covers both the direct processing costs of an article, and the indirect costs of running the entire publishing business. Therefore, the term APC is itself misleading in multiple dimensions.

Transparency?

There are some exceptions to the second point, however. One of these is Ubiquity Press, and they offer a completely transparent cost breakdown below. They state: “Our average base Article Processing Charge (APC) from only £400 (€450, $525) and this is the only thing we charge. In order to establish trust with authors, institutions and funders, we provide a transparent breakdown of how the APC is calculated.“

Ubiquity Press is completely sustainable based on this, and their APC is around 25% of what many of their competitors at the large publishing houses charge. Others such as eLife and the Royal Society have made their costs transparent too. eLife estimates the ‘fixed’ and marginal costs of publishing per article to be £3,085, with £1,798 based on marginal costs (including paying editors and other staff, and maintaining online systems), but an APC of $2,500. It would be great to create a database of publishers and journals that are appropriately transparent about their publishing costs.

Can it be done more cheaply?

Yes, absolutely. Let’s look at some simple examples.

1. The Journal of Machine Learning Research has an average per-article cost of $6.50. It does not charge an APC for this, and is supported by outside donations.
2. arXiv. Costs about $800,000 a year to run, with average costs coming to around $10/article (EDIT: for updated statistics, see here, thanks to Phil Gooch for pointing this out).
3. Discrete Analysis: “The absolute worst that could happen is that in a few years’ time, we will have to ask people to pay an amount roughly equal to the cost of a couple of beers to submit a paper, but it is unlikely that we will ever have to charge anything.” (Supported by the University of Cambridge for now, costs around $10/article).

All of these include the direct and indirect costs. There are many more examples than these few too.

Björn Brembs also estimates that the maximum cost for a scholarly article should be no more than $100: “..15 years ago a publicly financed solution was developed (SciELO) that publishes fully accessible articles at a cost of between US$70-200, depending on various technical details.”

Another study from 2016 showed the following: “We found that an end-to-end scholarly communication solution can be provided exclusively by a combination of vendor services. Based on price data from 15 vendors, we found that the marginal cost of scholarly communication was between $69 per article and $318 per article. We found that these costs were not impacted by the different input formats used by authors (e.g., Microsoft Word or LaTeX).” (bold emphasis from me).

The only times in which higher estimates are made (usually in the ballpark of $5000), this is based on the total number of published articles divided by the total amount of revenue publishers derive. This is not an estimate of the cost of article processing, but of the average amount of revenue generated per article. Therefore, it is an estimate of how much it would cost to sustain that revenue stream through ‘APCs’, which would be decoupled from the cost of article processing.

How much can be saved if the true cost of publishing was known?

From Nature News: “Data from the consulting firm Outsell in Burlingame, California, suggest that the science-publishing industry generated $9.4 billion in revenue in 2011 and published around 1.8 million English-language articles — an average revenue per article of roughly $5,000.”

So, back of the envelope calculation. What would happen if the global research community wanted to publish 2 million articles in 2018, and utlised systems modelled on arXiv, Discrete Analysis, and JMLR, which expose the true cost of publishing at different scales (around $10). That would cost in the region of $20 million in total. Which means, in theory, we could save around $9.38 billion a year by switching to an efficient Open Access publishing system. Here, the APC would be a real APC, reflecting the per-article processing cost, including direct and indirect costs. It would be about $10 per article.

Even if we multiply this by a factor of 10 (e.g., to represent scale), the total cost would be $200 million, and we would save $9.2 billion a year. This would be to have 100% Open Access too, and overcome the slow growth that the system is currently experiencing: “We estimate that at least 28% of the scholarly literature is OA (19M in total) and that this proportion is growing, driven particularly by growth in Gold and Hybrid.” (Piwowar et al., 2018).

Just to reiterate. It is eminently possible to achieve 100% Open Access, while saving research institutes and the public purse more than $9 billion every year.

Pre-empting the backlash to this

Now, people are going to argue against this price calculation, usually using several common angles. Let’s deconstruct them here to save some time:

• “This doesn’t take into account the cost of rejecting articles.”
• “This doesn’t take into account the cost of managing peer review.”

Well, folks, if we’re spending 99% of our global publishing budget on rejecting science (on solid or unsolid grounds, this ratio cannot be known due to editorial opacity) and managing a process where virtually the entire workforce are volunteers, then we are doing a pretty terrible job.

Another common counter-argument will be something like:

• “APCs are based on what the market can bear.”

Well, this is actually quite deceptive, as there is no actual market for research articles. Markets work when there is price variation based on competition. However, that does not exist for scholarly publishing, as every single research article is unique, and therefore comprises a mini market in itself, with its owner having a mini monopoly on it. For example, if a research article costs $40 to access in one venue, you have to pay that, as that article is only availble there; you cannot go to another vendor and pay a lower price for a similar article, because each research article is unique.

This argument is also just another way of saying ‘We charge you as much as we can get away with. We like money.’

An extension of this is usually something like:

• “APCs reflect market value to researchers, based on the prestige/brand of a journal.”

Seeing as this is decoupled from any real sense of scholarly value (e.g., real world impact), and the facade that corporate journal brands should mean anything in a rational research ecosystem, this point is also null. In fact, if what we are paying for is prestige (which research suggests we are), then we are basically being duped, and deserve to be in this pretty crap state of affairs. But upon recognising this, we should also get ourselves out of it, pronto.

Where is all the money going?

Another serious question arises from this. How much more are the ‘indirect costs’ for publishers that charge $3000+ per article, and what are those costs assocaiated with. We know that the true cost of article processing (i.e., the direct costs) are around $100, even at a high end. So what about the rest?

Lets take Elsevier journals as an example. We don’t know a breakdown of their direct or indirect costs. But lets be really generous and say that their direct processing costs are $500 per article. In the UK, their average APC is £1,863.26 (see Fig. 9 of this JISC report from 2016), which is $2609.41 at current rates. In 2016. Their adjusted operating profits in 2016 were 40%,  so around $1043.76 of each APC went straight to shareholders.

So $2609.41 – $500 – $1043.76 = $1,066.41. (APC minus processing costs minus profit = remainder).

So if this remainder represents an estimate of the indirect costs, which amount to around 8 times that for a Ubiquity Press APC, how do/can they justify this? Alternatively, if we use the cost breakdown from UP, that around 62% of the APC comprises direct costs (i.e., processing), then that is still around $1600 for Elsevier, or 4 times the total APC for UP.

Either way you look at it, even after profits are accounted for, a lot of money is going somewhere. Some possible outlets include:

• Lobbying politicians, which they do, and spend millions each year on.
• Inefficient and outdated systems (contributing to high direct and indirect costs).
• Bloated salaries (their CEO takes in about $14 million a year alone).

Now Elsevier aren’t alone in this. The same can be said of Wiley, PLOS, Taylor and Francis, and more, with adjusted calculations. If we compare it to journals that are even more cost effective than those published by Ubiquity Press, the story gets worse. Until we get transparency, we have to assume that these publishers are charging more for their own inefficiency. Clearly, APCs again do not reflect the true cost of processing an article. And even if they did, they would represent inefficiency again, if they were anything above $500 or so.

APCs are misleading, and combining this with the lack of transparency means that everyone suffers, except for publishers.

What needs to be done

The whole concept of APCs then, is largely redundant and misleading. It does not reflect the true cost of publishing, it does not reflect an efficient system of publishing, and it certainly does not reflect the actual cost of processing an article (with exceptions). There is also clearly no APC market, and therefore no pressure for an APC to reflect any of these things.

It is time for research funders and policymakers to recognise this, and implement some way of stopping it. Policymakers (e.g., at the national level) should be accountable to the wider public, not the whims of private corporations, and base scholarly publishing policies based on this. Their incentive to do this is simple: Getting more value for their money. Funding bodies that do not take into account the true variation in APCs, and the true cost of publishing, are doing an incredible disservice to research, and those who provide the funds for them to distribute.

Ideas for practical steps that can be taken:

• If a publisher is not 100% transparent about their direct and indirect costs, it cannot receive an APC. Transparency will help to create a functioning market.
• Based on the total range of APCs gained from this, a market cap could be implemented at some fraction of the highest APC (maybe 10%, tops). Any journal with an APC above this cannot be a recipient of an APC. This will actually force a system of price competitiveness to come into existence.
• There is enough money in the system, and enough evidence out there, to start experimenting with new more-efficient models.

Barriers to change

The only people who wouldn’t want this are those in higher up positions at many publishers. Most who work in publishing probably want to offer more efficient, cost-effective alternatives. But feel they are powerless to create change within an organisation. I can’t imagine the suggestion of a 90% revenue cut going down too well in board room discussions. Which in itself is exactly why this issue exists. A for-profit entity has no incentive to reduce its revenue or income.

Similarly, many researchers will feel they have little power in this issue (if they are even aware of it), as they are captive to the ‘publish or perish’ system. Also, most researchers are not accountable for paying these APCs themselves, and therefore insensitive to prices. On a personal note, I have spent around $30,000 of taxpayers money on APCs during my few years as a junior researcher, and do not feel great about it. I was not accountable to anyone, and I feel it is difficult to ask junior researchers to be the ones in charge of disrupting this system.

However, steps can be taken. Awareness is the first simple one for us all. For people with the power to change this, it is your civic duty to do so. The longer people with the power to do so let this abuse continue, the more they label themselves as irresponsible.

An alternative term to APCs – the SPF

So, are APCs misleading, as currently used? Yes. They do not tell us about the actual cost of article processing, and they do not actually relate to the explicit cost of article processing. I suggest that the term ‘APC’ be much more strictly used to reflect the cost of article processing, and only under conditions when these numbers are known. 

For the majority of publishers, this means that they would not be able to use the term APC anymore to reflect their author-side fees. This includes publishers like PLOS, Elsevier, PeerJ, F1000, Wiley, Taylor and Francis, Sage, Springer Nature, and the American Chemical Society, etc.

Instead, for these journals, I recommend that we call what they call an APC something different. One idea for this could be the Sustaining Publishing Factor, or SPF. This would be simply based on their “APC” divided by the estimated true cost of publishing.

For example, PLOS ONE has an “APC” of $1,495; but this is not an actual APC, as we have no idea what their indirect or direct costs are. Hopefully the arguments above have made that clear. However, if we estimate that the actual cost of article processing is around $100 again, then PLOS ONE would have an SPF of 14.95. Nature Communications, with an “APC” of $5200 (before tax) would have an SPF of 52.

Here, the higher the SPF, the less value for money a journal offers, the more divergent they are from the true cost of publishing, and the more misleading their advertised “APC” is.

If a journal wants to change its estimated SPF, the only way to do so would be to be transparent about both its indirect and direct costs. Bring it on.

Further reading

• Wikipedia entry on APCs
• Ubiquity Press – Article Processing Charges
• Royal Society
• eLIFE – Setting a fee for publication
• Nature News – Open Access: The true cost of science publishing
• Björn Brembs – How much should a scholarly article cost the taxpayer?