i've demonstrated repeatedly that there isn't even a really meaningful correlation. the real correlation is over age. so, the correlative basis of the logical error of saying "she wins because of black voters" isn't even true in the first place. so, it shouldn't be surprising to anybody that the models have consistently been 10-15 points off in making predictions. the racist models suggested she would do better in illinois than ohio, for example - something that was obviously absurd to anybody with a cursory understanding of the cultures of these places (ohio voted for bush!). and, it's likewise absolutely ridiculous to suggest that sanders will win west virginia, but i'll get to that in a moment.
as the data has come in, we've learned that sanders has been competitive with young blacks in less conservative states. he has not been competitive with older blacks. but, he has not been competitive with older whites, either. he has just broadly not been competitive with old people. that's the real connection. and, that's the reason florida was such a rout.
the real lesson to take away from the demographics is that young blacks are voting at a dramatically lower level, proportionally, than young whites. there is obviously a serious issue of disengagement with the traditional political system. if young blacks were voting at the same percentage as young whites, sanders would be carrying the black vote in the northern states. and, that's actually a serious issue. it's something the media should be strenuously focusing on. why aren't young blacks voting at the same levels as young whites?
but, we haven't had an opportunity, yet, to turn the models on their heads. i was hoping hawaii would be the first break, but the reality is that there wasn't any polling done there. the rust belt states should be interpreted as blurring the correlations; they haven't been interpreted as they should, but there wasn't anywhere where the predictions were dramatically wrong, either. so, the pseudo-science can maintain a line of plausibility by pointing to error.
again: i'm suggesting it's a proxy variable. so, you can accidentally get to the right answers so long as it continues to be the case that
(1) all the states with high black populations have overwhelmingly right-leaning bases of democrats.
(2) all the states with low black populations have overwhelmingly left-leaning bases of democrats.
you'd get it right every time. but by accident. you haven't even attempted to falsify it, yet. you haven't done the proper experiment. there are two types of states that have not yet voted:
(1) states with high black populations that are left-leaning. this would include maryland, dc and delaware.
(2) states with low black populations that are right-leaning. this would include kentucky and west virginia.
i need to reiterate: these tests have not been done. we do not know their outcome. but, in order to put any faith in the model, it is necessary to test them. one should not say that maryland should vote for clinton because it has a lot of blacks. one should say that we can only determine the relationship between skin colour and voting intention by controlling for ideology, and that maryland/delaware/dc are essential in building the dataset.
so, do not listen to these people. their models should not be able to predict the results in the following states:
1) new york [liberal, with blacks]
2) maryland [liberal, with blacks]
3) delaware [liberal, with blacks]
4) west viriginia [conservative, few blacks]
5) kentucky [conservative, few blacks]
6) new jersey [liberal, with blacks]
7) dc [liberal, with blacks]
however. once the new york primary is done, you will at least have some data. i am not likely to claim it will be very useful. but, the argument will at least be different.
again: you can't use what you have to predict new york. rather, the results of new york are a pre-condition to the coherency of the model.
http://www.realclearpolitics.com/articles/2016/04/01/sanders_hopes_look_dashed_but_sanders-ism_isnt_130156.html