With all of the discussion over the pricing of journals, one thing that has come up is their supposed decreasing importance. With online repositories of working papers, scholars and others no longer need access to the journals to get access to the research. To be sure, publication gives with it certification and that helps but conceivably no journal is required for that per se; just a certifying authority.
It is in this context that I want to refer you to a new paper by Jeff Furman, Kyle Jensen and Fiona Murray on the role and effect of retractions. Here is a convenient write-up of their results. They find that when a paper is retracted, scholars stop citing it. Interestingly, the citation drop off isn’t to zero but between two thirds and three quarters less than what it might otherwise be. (How do they work out what it might otherwise be? A clever use of journal placements). What this suggests is that retractions and the ability to retract are important for altering the pool of knowledge that others draw upon.
But here is the thing: with the Internet, working papers online stay there pretty much forever. Certainly, it is hard to retract something from Google Scholar. But there is an issue with working papers as to how they are removed from search, in particular. One suspects that we need alongside Google Scholar, the retraction to sit there too. But unlike journals and even publishers, it is murky as to how we can influence the make-up of Google Scholar.
The point is that retractions bolster the case for journals as a means of dissemination over working papers and so it is something to think about.
As for the paper discussed here, here is the published version. Ironically, I searched and there is no working paper online for this one.
[Update: In response this post, a working paper is now available here.]
Digitopoly 
So, the certifying authority also needs to be able to retract certification. Retraction is a minor, if valuable function for journals. The fact that retraction is worth something does not mean that the antiquated journal system needs to be retained as a whole.
Also, someone has to say it: the sensible model for “working papers” is the physics pre-print archive, http://arxiv.org/ .
And this most certainly allows for retractions, as well as corrections. These are clearly dated and in both cases the original version is always still available.