on the one hand, this is the perfect opportunity to test what i'm saying, which is why i'm doing this. we have long term aggregate modelling in multiple states. we have a single, reliable poll taken within a few days of the vote. which one is more accurate? i claim that reading the single snapshot will be more accurate than the aggregate modelling, and you can throw all that out the window.
on the other hand, i need to bring my cognitive dissonance back in. this poll is suggesting - across the board - that clinton is doing better with self-identified liberals, while sanders is doing better with self-identified conservatives. this is patently ridiculous, and could only be explained in three ways:
1) clinton's attacks (on gun rights, for example) are backfiring & people just aren't listening to bernie at all. do gun rights really overpower health care? this is crazy.
2) this poll is a part of the rigging process, in which case the clinton campaign is trying to lower expectations ahead of a careful grind & halt operation.
3) it's just some kind of weird anomaly.
see, i can't analyze this thing under the assumption that it's a vast, right-wing conspiracy against sanders - even if my gut tells me it is. i need to analyze it on face value. but i can caveat it.
florida
he's actually not doing so badly. she's under 60. the polling seems to be polling women higher than the population, which maybe says something about the sample. if he keeps her under 60, it's a hole he can climb out of - if he wins big in illinois. but, there's reason to think he may be able to keep it under 55, too.
the flip side is turnout amongst old people. but, i have a criticism of the modelling on this point, too: while they are correct to point out that turnout is higher around the age of 65, they fail to model the decline in turnout that sets in as age increases. so, when they model 65-70 year olds as more likely to vote, this is correct - but when they model 85-90 year olds as also more likely to vote, this is wrong. they only gather data in a 65+ age category. they simply don't know how much of the over 65 sample is also over 80.
over the last few years, it has been publicized that polling firms have screwed up predictions by overweighting young people. i believe that there are examples (the last ontario provincial election) where they overcompensated by overweighting old people. you need to watch for that in florida.
but, that doesn't change the reality that a win, here, for sanders is keeping clinton to 60. and a big win is keeping her to 55.
north carolina
they're claiming she has a big lead amongst early voters. again: if hillary was going to rig this, how would she do it? but, still, again: if she's really only pushing 50% on the yet-to-vote, then an optimistic scenario suggests he can keep her under 55, if he can get that turnout up.
what i said about michigan was that you should expect her to win, but that it will be closer than expected. i'm saying the same thing about florida & north carolina. i think there's a higher level of expectation that she win these states. but make of it what you will.
it will be suspicious if north carolina looks like alabama (alabamastan? these are soviet polling numbers...) as the polls don't support that - and, perhaps the polls didn't really support the large margins in these other southern states, either. it will also be suspicious if she wins florida by a large margin, when the polls are steady-to-narrowing.
illinois
all of the factors put together (including overpolling women, again) make it clear: bernie sanders will win illinois. the question is by how much - 55? 60? 65? it's all about turnout. he needs delegates! so, if you're in illinois then get out and vote. how well he does in illinois will likely determine how seriously he's taken after tomorrow.
another thing to watch in illinois: this poll suggests that his support amongst blacks is statistically the same as his support amongst latinos. now, he may lose both. but, the point is that they're the same. that is massive, as it is the first sign of hillary's weird lock on blacks breaking up - and strongly suggestive that this might be about southern, conservative blacks rather than blacks altogether.
it will be suspicious if she wins illinois, by any margin. bernie ought to be favoured, here.
missouri
bernie sanders will also win missouri. and, note that they are again oversampling women. it's almost like they're doing this on purpose. again: get out and vote! it's about turnout.
it will be suspicious if she wins missouri, by any margin. bernie ought to be favoured, here.
ohio
ohio has 12% undecided. how are you undecided on hillary clinton? what that means is that you want to vote for sanders, but can't get past your programming - or don't want to admit it over the phone. advantage: bernie.
they oversampled women, again.
with all of the factors, i also think that bernie ought to be favoured to win ohio, but it will be closer. so, get out and vote! that's going to be the difference: turnout.
=====
so, again, it's not completely clear what this poll is.
either it's the decent, post-michigan poll i was waiting for - and it's confirming my analysis...
....or it's a part of hillary's faltering attempt to fix the nomination, and meant to lower expectations for a run on the polls that she expects will overpower her attempts to stuff ballot boxes.
either way, it's clear that bernie is surging and that his ability to win tomorrow depends on turnout.
i think they basically split the delegates for the day, and move on to the next fight. but, the momentum puts sanders in a stronger position to gain delegates in new york and pennsylvania.
but, it was always about california.