Arnold Kling reviews the positive externality argument of education and isn't buying it:
the positive externality is the benefits that accrue to me from your education. I think that those benefits tend to be pretty small. You get a higher income, and most of those benefits flow to you. I get some of the benefits, because you are more likely to pay taxes and less likely to require government transfers, so that my tax obligations can be correspondingly reduced.
In other words, he's arguing that the externalities associated with a college degree are pretty small, implying that the subsidy should be pretty small as well. But he goes on to note that the externalities may be negative, which would imply that higher ed should be taxed, not subsidized:
If the higher income that you get from education is due to its signaling effects, then that is a classic negative externality. The investment in the signal is wasteful, and your investment forces others to make a wasteful investment.
On the whole, the case for taxing education rather than subsidizing it is really quite plausible.
For those of you not familiar with "signaling," it is the idea that a degree is useful because it signals that you have a set of desirable characteristics (good work ethic, punctuality, creativity etc.) that are not learned in school, as opposed to being useful because it certifies that you learned X, Y and Z in school. Since people with these characteristics are more likely to get a degree, the value of a degree comes primarily from signaling, and because signaling is costly (tuition, usually subsidized as well), and encourages others to spend resources signaling, higher education has negative externalities. If that is the case it should be taxed.