But he adds:The huge polling miss in Michigan was a cautionary tale: for all its excellence, FiveThirtyEight is only as good as the polls, which at least this time were way, way off. It’s too soon to dismiss them, but not too soon to explore other approaches, which is what Alan Abramowitz has just done. Using data for the Dem contests so far, he estimates a simple relationship based on just two factors — nonwhite share of the population, and north versus south — that accounts for the results quite well — and Michigan is not an outlier.If this model is right — and bear in mind that until Michigan polls were doing pretty well, so it may be a special case — next Tuesday won’t go as well for Clinton as the polls predict, with Sanders quite possibly winning three states.
What about later? There aren’t many more Southern states. On the other hand, after Tuesday almost half the pledged delegates will have been chosen, and to close the gap Sanders would have to win big in what remains. The back of my envelope says that he would need to win the rest of the primaries by something like a 14-16 point margin, which seems unlikely. So even with this model, Clinton remains a strong favorite.
But this could go on for a long, long time.-RW