Excerpted from this talk at the 2013 APA Eastern.
Let's say, for the sake of argument that, notwithstanding some exceptions, merit is a necessary condition for TT placement and advancement in university philosophy programs. But it doesn't follow from that that merit is a sufficient condition for that; there are many talented people who end up in precarious academic labor. But this injection of sheer luck into placement and advancement is hard to accept for some people; they want to think that those who end up in precarious labor deserved it somehow; the reason they didn't make it was some lack of merit on their part. In other words, some folks just don't want to accept that we have a tragic job system where bad things happen to good people.
There's a wrinkle here: if you don't win the early TT job lottery [this is a strong way to put the anti-"merit as sufficient condition" position, but what the hell, let's go with it], your work conditions are going to be such that your productivity will suffer and it will look, retrospectively, like you always lacked the merit that would have warranted your getting a TT job. But this lack of productivity is produced by external circumstance as much as – or better, more than – it is an exhibition of some inherent quality of the person. So we're back to our critique of the attribution error -- we have made network position into the property of a person.
In other words, there's nothing about you, Asst Prof X, that would let you show your merit in a precarious labor position. (Again, this is an extreme formulation, and probably should be reworked along population thinking lines – given a big enough sample, the odds of a good sized random sub-sample of early TT hires placed into precarious labor positions being able to gain the publications and citations that would allow a "merit" perception would be much lower than that of a similar random sub-sample of precarious labor people placed into a TT post getting those publications and citations.) Call this latter point the Trading Places hypothesis.