"It needs to be tested. Hopefully, someone will step up to bat and do just that."

Call me crazy, but what about 'us'?

What kind of resources would it take for MyDD (or a group including MyDD) to finance our own poll?

This would be a project that, if realistic, I would financial support (to the degree that I can). Do others feel the same way?

Yeah exactly, i thought mydd was going to do a poll?  I was in favor of the NV poll.  If you want it done right, do it yourself.

"The theory is simple: Hillary Clinton is weaker among all self-identified Democrats than she is among the Democrats who will actually make up the primary / caucus electorate."

Am I going crosseyed, or did you get this backwards, Chris?

Other than that, fascinating post.

"The theory is simple: Hillary Clinton is weaker among all self-identified Democrats than she is among the Democrats who will actually make up the primary / caucus electorate."

Isn't your theory actually the reverse of this?

"It is entirely possible that likely Democratic caucus-goers in Iowa are more engaged in the campaign relative to all self-identified Democrats than they are nationwide."

This absolutely correct.  I worked in Iowa for the 2004 campaign. My daughter was organizing county caucuses for John Kerry so I went out to help numerous times.

They are far more committed and engaged in the political process than primary voters.  It's an enormous time commitment between the caucus itself as well as attending the various candidate events. They take it very seriously.

John Kerry came up fast on Howard Dean and Edwards was coming up fast on John Kerry.  He has very grounded support there.

I can not reiterate it more, but disciplined organization is the key to Iowa for anyone.  

Chris - You make a very good point about the squishiness of primary polling results. There are some very good comments about this on the website Pollster.com.  Basically they point out that there isn't a very good way to screen for the most likely primary voters, and it's ten times more difficult in caucus states like Iowa. So the poll results are pretty useless right now.

However, if we could find a polling company who has a solid method for identifying the most likely primary voters, we should do a poll in one or more of the first states, which will have an inordinate impact on candidate fortunes going into the big kahuna of Feb 5th.

The cost of doing a poll like this is at least in the $10,000 range, probably including buying the sample.  Of course, if you created one from the ground up with 100 MyDD people volunteering to each get 5-6 responses off of the sample list, it could be possible. . . .

How do you screen for likely voters?  You ask them:  "As you may know, there will be a presidential primary in your state on ( ).  Since this is a primary, most people will not vote.  How about you?  Would you say you are very likely to vote, somewhat likely to vote, chance 50-50 you'll vote, or are you not likely to vote?"
Then you only take the likely voters and tell the rest thank you for your time.

A couple other ways to screen for likelihood of voting is to ask people if they can identify where their polling place is (this is how Gallup does it, but this early out that just isn't decided yet), or get their previous track record if they've voted in past primaries (this is problematic because turnout will likely be larger this time around because of the extra interest created by the candidates and the very competitive field).

The problem you run into this early asking about likelihood of voting is it's WAY too far in advance.  Another problem is cost.  Assuming Chris is right (I think he is), you would have to screen out that extra 5-10% of your sample who are only marginally likely to vote and find even more responses.  This costs both time and money, something most media polls simply don't want to do.  

And considering the average journalist reads most polls about as well as I do a schematic to build a nuclear reactor, you get garbage in, garbage out.   Less effective poll + Less informed analyst = "HRC and Giuliani may as well be anointed the nominees right now"

Most media polls start to screen for likely voters come October of an election year, so I doubt we'll get any good public information for another 6 months at least that tests this theory.

From Public Policy Polling, a regional IVR polling company in North Carolina...

"At Public Policy Polling we go about primary polling completely differently. We don't use subsets of a general electorate poll, but conduct completely separate primary polls. We sample only households that have a history of voting in party primaries. That way we know we are only reaching those likely to vote in a primary.

Therefore, I suggest that the best primary polling would be those conducted using a sample based on voter history. Unfortunately, that type of sampling is unavailable nationally-- only about 3/4ths of states have that information available."

http://publicpolicypolling.blogspot.com/ 2007/04/do-national-polls-inflate-hillar y.html

It seems obvious that Caucus states select for the most engaged and politically-interested people. Also, older people, and the party-faithful, i.e. people who vote every time. Finally, non-presidential election years select even more towards the core. Chrs's stats show Obama's support droping off as the data is restricted to caucus goers, which may be a result of Obama receiving higher support from younger voters.

Colorado has both a caucus AND a primary. This gives you a pretty good way to measure these effect (from past election cycles). This history shows that the Party Establishment (elected officials, donors and Party officials) are more moderate, the Party faithful are very Liberal, maybe even California-esque, and the Primary voters are closer to the general public.

An Insurgent, even highly-liberal candidate can get a lot of support in the caucus, enough to get first listing on the Primary slate. But then the more moderate candidate shows better in the runoff, i.e. the Primary. You also see this effect in the Party Platform.

Republicans

The Republicans have this same problem with their extreme right-wing. McCain is clearly talking to this base. All his nut-case comments make sense if we realize he is trying to lock-in the die-hard 30% who still believe in Bush. That 30% is probably over 50% of the Republican Party base at the Primary level, and 66% in the caucus situation.

supporters are the most energized about a candidate as exemplified by crowds and donors, sure young voters generally don't show up but it's usually because they don't connect with any of the candidates, not because they are lazy ect...they did show for Kerry in 2004 and look at Jesse Ventura in Minnesota's governers race, Obama should actually do better in caucuses because the most likely voter is someone who care enough to donate, volunteer and attend a rally. I wouldn't bet against all these "Obama voters" not showing up becuase in addition to the support the Obama people also have the money now to organize thier voters as I saw firsthand last night.

Actually I think your caveats are quite important for the support of your theory in these numbers.

self-identified Democratic
Clinton 29.6%  
Edwards 26.3%
Obama   20.9%
Other   23.2%

Likely Democratic caucus goers
Clinton 28.5%
Edwards 34.2%
Obama   19.3%
Other   18.0%

Difference
Clinton -0.9%
Edwards +7.9%
Obama   -1.6%
Other   -5.2%

Obama suffers a slightly larger drop then Clinton. But only Edwards has any significant movement mainly coming from other sources.

If it was Clintons data that was polluted then all other candidates should improve their numbers or you'd expect a significant drop with her support. But all candidates except Edwards stay stable, the minor candidates as well. (Each under 2%) This points to Edwards as under represented in polls of self-identified Democratic not Clinton being over represented.

I still find your theory quite persuasive but this poll points to quite another and interesting anomaly