OA favors cash-rich disciplines such as molecular biology (where grants are big), but not cash-poor disciplines such as math or the humanities (where grants are small or non-existent). Under subscription-based models, librarians can in theory offset the discrepancy by picking journals according to the perceived need for the title – essentially the system is meritocratic rather than plutocratic.

The OA solution to the problem is of course waivers. But while I’m sure biomedical funding bodies are happy to pay a bit extra in OA fees to subsidize the odd ecology article in PLoS Biology, I can’t see them happy to fund an entire history journal.