There's been some discussion lately about how to get people to be better at refereeing. Referees are a vital part of the profession, and they're an all-volunteer corps, and they are, by and large, not awesome. It takes tons of time to get your paper refereed, there are often no helpful comments, and there is a strong temptation to free-ride.
An idea that gets bounced around maybe once a year is charging authors a submission fee and then use the funds raised to compensate the referees. This will give the referees an incentive to do their jobs and have the side-effect of reducing frivolous submissions. This guy, whose wikipedia entry suggests that he is a pretty accomplished person and probably not a dumb-dumb, suggests a sliding scale based on desired turn-around time. He says, "a submitter's willingness to pay a higher fee, some economists might say, serves as a signal of the submitter's confidence in the quality of the paper."
Now, I'm no economist, but I would have thought that a willingness to pay a higher fee under such a system would serve most straightforwardly as a signal of how much of a hurry the submitter was in. I would have thought that there would also be a significant relationship between a willingness to pay a higher fee and the submitter's access to disposable income. I'm sure that confidence in the quality of the paper is on the list somewhere, but it's not right at the top or anything. (Not that author confidence of quality is a super-reliable indicator of actual quality, anyways.)
This is the fundamental problem with all such proposals: they disadvantage graduate students and junior faculty, who don't make as much money as senior people, and for whom the marginal utility of an extra publication is greatest. That is, such proposals are unfair.
So here's the idea. Charge people a fee. But the fee isn't for submission; it's for publication. You wouldn't want just anyone who submits to pay this fee. And the fee isn't monetary, so it's not taking food out of the mouths of Professor Newbie's kids. The fee is the job of refereeing itself. The journal accepts your paper and part of the permission-to-publish contract includes an agreement that the author serve as a referee for that journal. There is wiggle room concerning how many times would be fair. But I assume it would be at least two, and maybe more. The jobs could be spread out over a year or two. And the journal would have some leverage over its referees, since it could withhold the paper until (some subset of) the author's refereeing duties had been discharged.
Does this proposal have some problems? Absolutely. But I submit that if it's not better than the current system, nothing is.