This seems to be an important bit of research. The U.S. Chronicle of Higher Education reports on new NHA research which finds that…
“It costs more than three times as much to publish an article in a humanities or social-science journal as it does to publish one in a science, technical, or medical, or STM, journal [ reports ] an in-depth study of eight flagship journals in the humanities and social sciences.” […] “It cost an average of $9,994 in 2007 to publish an article in one of the eight journals analyzed” […] first-copy costs — “collecting, reviewing, editing, and developing content” — added up to about 47 per cent of the total outlay among the eight journals studied
The National Humanities Alliance report The Future of Scholarly Journals Publishing Among Social Science and Humanities Associations (not yet online) was written during 2007-2009, and examined U.S. data from 2005 to 2007. The Chronicle journalist highlights three possible reasons for the difference…
* articles are significantly longer than in the sciences
* acceptance rates are far lower than in the sciences, at a pitiful 11%
* such journals include a wider variety of content than in the sciences…
“peer-reviewed research made up about 62 percent of what the eight journals published in 2007. The remaining 38 percent consisted of “other scholarly content,” including book reviews.” […] Such material does not come cheap, though; it must still be commissioned, edited, and put into production. It cost an [annual] average of $313,612 per journal in 2007, the study found.
On the “articles are longer” argument, I’m not sure that a simple word-count is a valid measure. Science articles are full of complex tables, formulae, diagrams, and it must take quite some time for a reviewer to mull these over. Similarly, I’m thinking that the acceptance rate may be so low because only the “top eight” most prestigious journals were surveyed — lesser journals may well have a far higher acceptance rate?