FWIW, your model is really badly broken if you didn't expect this- I would expect even most racist models (or, at least, my Turing-test-passing attempts at racist models) to predict this.
Why are you so confident that even most racist models would predict this? Prior to acquiring domain specific knowledge about trends or immigration policies in the United States it doesn't seem especially likely. Until Randaly provided the relevant link I knew next to nothing about the particulars of which racial groups successfully immigrate to the US at which education level. In the absence of such information it seems reasonable to guess that the patterns would tend toward following the trends for IQ among the groups as well as correlating with the distribution of educational qualifications for individuals in the respective countries. ie. I'd guess that East Asian groups would have more educational achievement than African groups.
I'd call the model possessed by myself of the recent past wrong and also ignorant of that and related pieces of trivia but I'd hardly call it "really badly broken". Perhaps you consider this and related information far more important or fundamental than I do?
I am Randaly; I didn't know that specific information before, but it did not surprise me. My understanding was that phenomenon of brain drain is fairly well known.
People want to tell everything instead of telling the best 15 words. They want to learn everything instead of the best 15 words. In this thread, instead post the best 15-words from a book you've read recently (or anything else). It has to stand on its own. It's not a summary, the whole value needs to be contained in those words.
I'll start in the comments below.
(Voted by the Schelling study group as the best exercise of the meeting.)