Tim Gowers reports on a new set of open access math journals. There will be a Forum of Mathematics:Pi that is supposed to be a generalist math journal, and Forum of Mathematics:Sigma that would have “clusters” for different areas. It will be paid for by Cambridge University Press for the first three years, and will then be author-pays open access. Mainly online (though one will be able to buy a print version). No fixed number of issues a year, so not rate-controlled.
The author-pays model is economically a more sensible one than the reader-pays one, I think. In the current model the agent deciding on where to send the paper has no incentive to take into account the cost of the product, except in very indirect ways such as concern that fewer libraries would subscribe to expensive journals and lead to fewer people reading the paper. As a result, journals have little incentive to keep prices down, and would essentially do monopolistic pricing. In the author-pays model, which in many (most? almost all?) cases will be an institution-pays model, at least the author will see what they are paying, so there will be some downward competitive pressure on prices. The currently announced price per paper for this new journal (USD 750; waived for the first three years) does seem a lot more than the cost of some of our CS open access journals (e.g. JMLR), and one hopes it will come down with time once they figure out the actual cost of running such an open access journal.
Windows On Theory 
The author-pay model is discriminatory against authors in poor countries or institutions.
I agree. On the other hand, one could argue that the reader-pays model is discriminatory against readers in poor countries or institutions, who can’t afford subscriptions to expensive journals. The situation is getting better with more papers being online, but there is still a lot of papers that I find myself looking for which are only available behind a paywall.
Also, this new journal is going to be free for authors from “developing countries”. The list currently is not perfect, but it’s a start.
It has the potential to be discriminatory, but this can be avoided with some care. Acceptance will be entirely independent of ability to pay, and fee waivers will be available for anyone who does not have grant or institutional funds available. I hope this will eventually become a moot point, if/when the community manages to get a sponsoring consortium in place that will take care of all these fees (see http://scoap3.org for an example in particle pysics). However, in the meantime I believe we’ll be able to keep this from becoming a problem, and of course I would resign as an editor if it did become a problem in practice.
$750 per article seems like such an an over-estimate that I am surprised that this was not a sticking point to the editors.
It’s actually surprisingly low: PLoS ONE, one of the cheapest mainstream gold OA journals, charges $1350 per paper, other PLoS journals are $2250, and Springer/Elsevier charge $3000.
EJC and JMLR and the like don’t just cut out things like copyediting (with results ranging from great to terrible, depending on the authors); they also rely heavily on volunteer labor to deal with the non-scholarly side of running a journal. That’s great, and they are doing a wonderful service for the community, but it does not seem to be scalable or sustainable: it’s much easier to recruit a small number of volunteers to participate in an exciting experiment than to maintain several hundred times as many in the long run. Unfortunately, the costs of running a free journal are still high, even though little or no money changes hands.
Part of the goal with Forum of Mathematics is to be transparent about the costs. Other journals with publication fees seem to base the fees instead on how much they think they can get people to pay. I hope increased efficiency allows fees to come down, but I hope it’s not simply by shifting labor to faculty volunteers (which would be a net loss for universities).
“Part of the goal with Forum of Mathematics is to be transparent about the costs”
I think this is the key issue. Assuming there is some cost in publishing, the funds should come from somewhere. It makes more sense to charge the author than the reader, but the most important thing is that the financial part of the endeavor be transparent and ‘non-profit’.
$750 USD !? You’re kidding, right.
Sorry, but author-pays model does not seem to be reasonable. A contributor should be paid, not pay him/herself.
Here is Mihai Pătraşcu’s opinion from 2008 on a similar idea.
http://infoweekly.blogspot.co.il/2008/08/open-access.html
Just to get some numbers in: quoting from a comment on Tim Gower’s blog post http://gowers.wordpress.com/2012/07/02/a-new-open-access-venture-from-cambridge-university-press/#comment-19267 :
“$750 is two orders of magnitude above the cost of archiving a paper — based on arXiv’s cost of under $7 in 2010, and the Journal of Machine Learning Research’s estimated cost of $6.50 (sources linked at http://goo.gl/uvPD1).
However, that does not include the cost of paper formatting and copyediting.”
I also think $750 is a lot. On the other hand, three years of transparency in costs should allow a good understanding of the costs. Also, now that the authors see the price tag clearly (rather than the libraries paying for the journals with little visibility to the scientists) the community can hopefully make better decisions on how much we want each aspect of the cost, and whether we are willing to pay for it.
To steal an analogy from Kamal Jain (in a different context, so if doesn’t quite apply, it would be my fault, not his), the current model is like buying a car where someone else pays for the car. In such a setting everyone would want a Mercedes (or better still Tesla/Ferrari/) and there would be little market for a Honda: a small feature that can cost say $1000 that I value for $100 will still be sold and we have large economic inefficiencies. With transparent costs, hopefully we will be able to see how much we pay for leather seats, GPS, etc., and decide better whether we want to pay for those, and economic efficiency increases a bit.
Regarding the $6.50 figure, this is valuing volunteer time at zero. JMLR published about a hundred papers last year. Even if you assume an average of only two hours per week was spent by faculty volunteers doing work that would otherwise have been handled by paid staff (which is almost certainly a gross underestimate), and that those volunteers have an average salary of $100k with 50% overhead for benefits, office space, etc., then that’s already $75/paper in volunteer time. I don’t know how to get good estimates for how much it really is in practice – I doubt anyone knows because JMLR probably isn’t keeping track carefully – but it’s vastly greater than the monetary costs. If I had to guess, I’d estimate the non-scholarly volunteer time (i.e., not counting refereeing, editorial decision making, etc.) would be at least a couple hundred dollars per paper, and that’s only if they are extraordinarily efficient. The $750 figure for FoM is higher, but it covers a lot more.
Thanks Henry for this analysis. I was thinking of doing something like this myself.
I think the outrage of some of the commentators above is a good sign. I assume much more than $750 per paper is being spent currently by the cummunity (mostly through libraries) worldwide, and the fact that this number is largely hidden from us makes it easy for us to ignore it.
There is no real cost for “formatting” and the likes. It is redundant. A good journal can be made with extremely low cost, while a bad journal can have fancy formatting and heavy bureaucracy.
Take for example, ECCC. Does it cost a lot? No (AFAIK). Do we have formatting and cosmetic editing issues there? No. But it’s quality far transcends journals like TCS (sorry, Elsevier). And this is even without refereeing!! only basic filtering.