Yes, there definitely is - in my experience - but it actually isn't in the direction that you would probably expect it to be in. I don't know how relevant my sample is and of course the sample size is going to be somewhat reduced but in my experience the people without fall into two groups: those that are not motivated at all and those that are highly motivated but in a direction orthogonal to what the education system would expect. Everybody else gets educated normally and ends up somewhere in the middle.

So the really smart people are the high end of the educated ones and the high end of the non-educated people (and with educated I mean university degree and onwards), and between those the gap isn't all that large, it's just that there are many more of those than that there are university educated ones (at least, around me).

Now, obviously if your circle is exclusively composed of well educated people then you will come to a different conclusion, and once you go outside of tech/IT the distribution will likely be a completely different one again.

So that 'set' has many subsets with very different distributions between the subsets.

I've probably done 50 interviews and hired around 10 people. The people doing them with me and I used a decently uniform process to give all applicants the same set of quite varied questions. We also gave the interview process to a decent number of existing employees to see how an existing employee fares.

I started the process thinking education would be quite independent of performance. Our data from the process showed exactly the opposite.

Quality of education, especially at entry level, made a large difference in the process and also in the long term performance of the hires.

This question is somewhat irrelevant, considering you have more information to work from. Given the other constraints you can apply, is level of education still a useful proxy?