Monday, April 27, 2020
I seem to get many, many times more referee requests than the number of papers I generate. Part of the explanation is the fact that journals nowadays require more than one referee per paper; which automatically means that, on average, people are going to be asked to referee more papers than they write. But this still doesn't explain the magnitude of work being requested.
Another reason, perhaps, is that judgment gets passed up the hierarchy to a smaller number of people than that produced the work. If you are not at the bottom, then you're going to have more work to do. However, I don't think this fully explains the workload, either
The remaining culprit, that probably makes up the remaining difference, is that some people just write too many papers! Some people write 30 papers per year, in fact. Assuming two referees per paper, and then also the editor, that means that 90 people have to read them. And math papers aren't like ones in many other fields, that are written in plain English (maybe with some technical terminology, here and there); at least the ones *I* get asked to referee usually involve pages and pages of picky-tricky inequalities and analytic estimates. Sometimes the papers are 50 pages long, in fact; and a proper read (instead of a quick scan) would take many hours of dedicated work.
I suppose many areas of CS have it worse. Machine Learning, for instance, generates many times more papers than math, per author. However, Machine Learning papers are much easier to read; and that field seems to go out of its way to make papers reader-friendly, unlike math, where they are written more in service to "principle", rather than "people" -- neuroticisms about things like "right language" and "getting the most general form" are paramount. The paper must be made to resemble a legal document or religious text.