I'm not sure about the justification for the X-Y figure, but it seems like larger numbers would tend to make this kind of effect unlikely. If you assume that in

I'm not sure about the justification for the X-Y figure, but it seems like larger numbers would tend to make this kind of effect unlikely. If you assume that in- and out-migration are not in equilibrium, but that the probability of migrating is uniformly distributed over people of all levels of education, then X-Y should be zero. Obviously, it would be better to use net numbers (arrivals of highly educated – departures of highly educated) and not attempt to normalise the numbers.