I found this interesting. It seems to have some validity since both Obama and Edwards have done substantially better in
Iowa and NH, the two states where voters are paying more attention than let's say California or Montana, than nationally versus HRC (although this is changing with Obama's recent national surge).
==April 24, 2007
The Inflated Clinton Poll Theory: Are Clinton And Obama Tied?
Democratic uber-blogger Chris Bowers's Inflated Clinton Poll Theory has crossed our wingspan several times, and it is worth considering. In the end, Bowers's provocative interpretation of some recent polling data may wind up explaining an Obama surge we don't yet see.
In essence, he's noticed that the Rasmussen robopolls tend to show a might tighter national race than polls conducted by live humans at a call center. Generally, Rasmussen claims to screen a much tighter selection of Democrats. Clinton and Obama are tied for the lead in those samples. In most national polls, Clinton has a high-single-digit-to-low-double-digit lead over Obama. Because Rasmussen reportedly draws from a more clued-in sample -- a sample paying close attention to the race -- Bowers notes that Clinton's lead over Obama evaporates.==
==Bowers responds:
" I can't derive an ought from an is, in this case. I don't know why Obama is doing better than Edwards in national polls--I just know that he is. I also know, from data in Pew and elsewhere, that he significantly rises among voters paying close attention, while Edwards does not. That is why I think Obama would benefit from a tighter voter screen more than Edwards."
There is conflicting evidence from the states in which we know voters are paying attention. John Edwards leads in Iowa, but Hillary Clinton still tops most New Hampshire polls. We've seen four South Carolina polls in two weeks, and Hillary Clinton has solid leads in two of them. We don't know enough about the methodology of these polls to say for certain whether a universe of likely primary/caucus goers has been accurately sampled. In Iowa, in fact, it is surprisingly tough to find a reliable sample.==
http://hotlineblog.nationaljournal.com/archives/2007/04/the_inflated_cl.html