In the comments to this post we discussed the signalling theory of education, which has previously been discussed on Less Wrong. The signalling theory says that education doesn't make you more productive, but constitutes a signal that you are productive (since only a productive worker could obtain a degree at a prestigious university, or so many employers think).

Such signalling can be very socially wasteful, since it can lead to a signalling arms race where people spend more and more money on signals that don't increase their productivity (like peacocks' tails). Now an important question is how one could rein in such signalling arms races. One way is by prohibiting employers to consider educations that are irrelevant for the job. That is, if your education is a pure signal of the abilities you had before you started the education, and doesn't increase your productivity in any way, employers would not be allowed to consider it when deciding between you and other applicants. The downside of this is, though, that it means more regulations and that it could be seen as illiberal.

Another hope is the increased use of big data in recruiting. Whereas previously, employers used crude heuristics such as which university you went to, they now have access to constantly improving algorithms which pick out precisely which applicant features that predict productivity and which don't.

Now suppose that what university education you went to is in fact a less accurate signal than some other feature. Then employers would fight over the applicants that have this other feature, rather than those with the university education. This would lead to people being less keen to obtain long and expensive university educations.

Of course new wasteful arms races could arise regarding these other features. Then again, I think we have reason to believe that these arms races would not be quite as wasteful as (I believe) the present educational arms races are. The reason people spend so much time and money on education as a signal is that it has proved to be so stable as a signal. People aren't going to spend as much time and money on a new signal, because they won't be as confident that it will continue to function as a strong signal. If what is taken to have signalling value is constantly shifted around, people would presumably be less willing to engage in signalling arms races.

These are just some loose thoughts. I'd be interested to hear if someone has any further thoughts on how to decrease wasteful signalling in education, or any other thoughts on the fascinating topic of signalling in general.