1. A golden rule in poll analysis: don’t care too much about crosstabs. In polls with samples like this, n=600, the margin of error of those small subsamples – say, “very liberal” – is huge. There’s not much you can read in it. PPP did very well in the past cycle. Look at those polls and you’ll see the same: very accurate toplines, subsamples completely screwed up. Same for every pollster. But ask anyone of them and they’ll tell you to not focus on the subsamples because the MoE is so massive.

2. Ron Paul has the largest and most effective campaign in Iowa. They are well-staffed and have the muscle they lacked 4 years ago. He is a much more well-known figure – his Jay Leno appearance got huge rates. They’ve been phone-banking Iowa intensively for weeks. They have hundreds of college kids landing there as of today. That turnout from self-ID indies and democrats is not surprising at all. I think that turnout model is very plausible.

3. Newt’s implosion is pretty much a given at this point. See the Gallup tracking poll. And nobody else is surging, every candidate is benefiting a little. Which obviously makes Iowa a 2 men race between Romney and Paul. I’d say that the superior commitment of Paul’s voters will push him to victory.

4. I’d say that Michelle Bachmann is the only one that can win over Paul or Romney at this point (in Iowa). Perry and Gingrich negatives are just too high. People don’t care about Santorum. But likely caucus goers still like Bachmann personally and she has a good operation on the ground, so she’s the dark horse here.

I know some have been predicting here Ron Paul can never win the caucus but I suspect that, once again, they’ll have to eat some crow. I think it’s the most probable scenery as of now.

In 2008, it was difficult to find a LV poll with a sample of 50% Bush voters in 2004. Same for Obama these days. When politicians are very unpopular, many don’t like to admit they were suckers.

It’s somewhat heartening indeed, but let’s not forget this is a poll of caucus goers. The Republican voters who caucus tend to be the most conservative ones and the independents/democrats who plan to caucus are in a very large extent Ron Paul supporters. A general population survey would produce far different results.

It’s the sample size of the crosstabs. If the sampling fraction is small, the margin of error is going to be much larger.

There is a lot of random ‘noise’ with smaller subgroups like ‘self-ID very liberal voters’. The margin of error increases as the reduced sample size of a subgroup becomes smaller and smaller and I suspect that subsample is really, really small.

Okay, we’re talking about 15 voters, 3% of the poll sample. The MoE is so massive that your question makes no sense whatsoever. It’s not even wrong, it’s simply meaningless.

Obsessing over small subgroups in small sample polls is something journalists and analysts like to do but has no real value whatsoever.

… are against ethanol subsidies. Take out Ron Paul hardcore supporters and that number is even smaller. And as dpmapper suggests, those who support the subsidies are the ones who care most and dearly about the issue. Those are the ones with a much larger chance of not voting for you/donating against you (for your opponents) because of this issue. Being against ethanol subsidies is not a guaranteed you’ll get like-minded voters, just a guarantee you’ll lose the other ones. As McCain results showed.

(this is the reason why there are so many government handouts that benefit a very small group yet are so difficult to repeal. A few people getting millions is, politically, a more powerful group than millions of people paying a few cents).

The fact that they don’t vote for a candidate of your liking?

If you’re posting here, you care much about politics than 95% of the American population.

Anyway, google “Shy Tory Factor”. It’s a similar phenomenon.

The concept that if the subgroup sample is small, the MoE is so high that analyzes are meaningless is pretty easy to assimilate for a regular person. If you can’t understand such a simple thing, there’s no point in debating it any further.

And yeah, don’t compare me to Neil when it comes to polls. I think Neil readily admits he’s just an amateur and lacks academic training and professional experience on this issue.

This time from Insider Advantage:
http://www.realclearpolitics.com/docs/2011/InsiderAdvantage_IA_1218.pdf

(note: IA has a very poor record and some analysts have cast doubts on the robustness of the work they’ve produced)