<The only accurate way to poll the Iowa caucuses starts with the party's voter file, which provides some data on past caucus participation. Four years ago the Democrats charged about $65,000 for this vital tool. Not one public pollster purchased this database, despite the fact that it is impossible to develop a useful sample without it. Yet poll after poll purported to tell us which candidate stood where in the caucuses.>

The article from by Mark Mellman goes on to say that the voter file is extremely important because it gives you information such as whether the registered voter previously participated in a Democratic caucus.  It's a starting point for the campaigns.  They'll contact these people first and then work from there.  

For historical perspective, in the 2000 caucus 44% of caucus goers had participated in previous Iowa caucuses.  In the 2004 caucus, 55% of caucus goers had participated in previous Iowa caucuses.  While we don't know how many caucus goers will be repeats this year, it is fairly safe to say that about half the people that come out for the caucus in January of 2008 will have done it before.

Of course, this leaves the other half of the caucus goers who will be first timers.  Who are these people?  Well, before we can try to figure that out we should again see what the percentage of registered voters in Iowa actually participated in each of the past contested caucuses.  In 2004, 5.7% of eligible Iowa voters particpated.  In 2000, it was 3.5% and in 1988 it was 5.9%.  So, we are looking at about 6% of the eligible voters in Iowa who will actually particpate next January.  That's roughly 130,000 Iowans.

How do you try and figure out who those 130,000 people are?  Well, the campaigns know by getting the voter file that about 60,000 or so will be repeat caucusers.  How about the others?  Well, Richardson's poll lists who they think are "likeliest".  It was those who participated in the 2004 caucus, voted in the 2006 primary and said they definitely plan to attend the caucus in January.  This at least gives us an idea that the campaigns think that the people who voted in the 2006 primary are possible caucus goers even if they didn't caucus in 2004 and it seems that once they contacted those people who voted in the 2006 primary, they asked them if they planned to attend the caucus in January.

I would guess that these 2006 primary voters who said that they plan to attend the caucus are the ones that comprise Richardson's "likely caucus-goers"  while the ones who caucused in 2004, voted in 2006 and said they definitely plan to attend the caucus are his "likeliest caucus-goers".  Remember he mentions that this "likliest" group is about 40% which closely mirrors that 50% of caucusers who have participated previously.

Here's the link to their poll.

http://www.richardsonforpresident.com/ne wsroom/pressreleases?id=0139

Now, let's compare that to how the public polling companies are polling Iowa.  Most of them at this point are using some form of "likely voter" model. The question is what do they constitute as a "likely" voter?  Well, let's first look at ARG, since 1)they've polled Iowa the most compared to any other polling firm and 2)it seems to have results fairly different from the others.  Their model is derived from randomly calling Iowa's registered voters and getting 600 registered voters (80% which are Democrats and 20% Independents)who say they definitely plan to participate in the caucus.  Now you may ask why 20% of Independents are included, but that is about how many Independents caucused in the Democratic caucus in 2004.  So, methodology wise, I don't have a problem with that.

The problem, I think, that this methodolgy has is it asks the respondent if they plan to attend.  This far out from the caucuses, you are going to get a lot more people that say they will attend the caucuses, but who will not, in the end, attend the caucus.  Remember, in historical context, it's only 6% of eligible voters who will partipate.  Now, you compare this methodology with someone like Research 2000 where they ask the respondents if they have previously attended a caucus and then asking them who they support.  This is a much tighter group to poll from and may account for some of the difference in their polls.

Finally, since all of these polling outfits are asking the respondent if they plan to attend or if they've attended in the past from the vast group of registered voters, the internal polls are already starting from a smaller pool of respondents by using the voter file and actually seeing who has particpated and/or voted rather than relying on the actual person telling the truth to a poller.

Where does this bring us regarding where I think the race in Iowa is at?  Based on this, I believe it is a lot closer to what Richardson's internal poll is showing that any of the public polls.  Edwards with a good, but not insurmountable lead, followed by Clinton, Obama and Richardson with Richardson being a lot closer to Clinton and Obama than any of the public polls are showing now.