Josh Marshall on Confusion over headline number in LAT Poll (Is Edwards in 2nd now?)

This really is a weird case. It looks like when the LA Times wrote the story about their poll they might not have understood how their own polling company defined the categories, and has since changed "very likely or certain" to just "likely." So nobody knows whether the correct "headline number" is CLINTON-29 OBAMA-26 EDWARDS-25, or CLINTON-31 EDWARDS-25 OBAMA-22.

...In Iowa the numbers for Democrats who plan to attend the caucuses is Clinton (29%), Obama (26%), Edwards (25%). With the 4% margin of error, that's basically a tie.

In yesterday's LA Times article it gave a separate set of numbers for "voters who said they are certain or very likely to actually participate in the Iowa caucuses." 'Very likely voters' is usually such a tight 'screen' that we somewhat discounted these results, which are better for both Clinton and Edwards. They're Clinton (31%), Edwards (25%) and Obama (22%). The same applies even more to 'certain'. And the more respondents you screen out, the higher the margin of error becomes.

But when we looked this morning at the actual poll internals it just says 'likely voters'. The Times article was also revised to remove 'certain'....

...In any case, the upshot is that based on what we know now, we'd give more weight to the likely voter number. It's still pretty much anyone's game, but the news is better for Clinton than we'd thought, and at least indirectly better for Edwards too.

http://talkingpointsmemo.com/archives/062159.php