but, this is exactly the kind of boneheaded thinking that has pushed silver into such a corner with his models, this year. texas is texas. if you discard texas, because it's texas, note that florida & new york were closed primaries (and therefore skew towards upper class voters) whereas nevada and illinois were open (and are therefore both a broader representation of latin-speakers and a better predictor in california).
instead of arguing that clinton performs well with latin speakers, silver should be pointing out that closed primaries make it very hard for latin speakers to vote. what he's picking out is consequently not minority strength from clinton, but the efficacy of voter suppression laws in preventing poor minorities from taking part in elections.
http://fivethirtyeight.com/features/hispanic-voters-will-decide-bernie-sanderss-fate-in-california/
i want to be clear about what i think is actually happening.
1) a state elects a republican governor. like in texas. or florida. or arizona.
2) that state passes restrictive voter suppression laws that makes it hard for poor people, who happen to mostly be minorities, to vote. so, this is a class thing. but, poverty is racialized in large swaths of the united states, so it's easy to get confused.
3) the state runs a closed primary, which requires registration. registration is hard in states with voter suppression laws.
4) so, only wealthy minorities (with proper id) are able to vote.
5) exit polls indicate clinton won minorities by large margins, and the media concludes she has a large lead with minorities - when, in fact, this is just the consequence of laws that make it hard for minorities to vote.
i think this explains why sanders wins latin speakers in open primary states and loses them in closed primary states. it also goes a long way to explaining the ridiculous age discrepancy with black voters, but cannot be a total solution - there has to be some level of disenfranchisement or apathy amongst young black voters, as well.