The following chart shows key numbers for the 2004 Iowa caucuses. The three rows at the bottom are all five-poll averages.
2004 January Iowa Polling
Poll
Kerry
Edwards
Dean
Gephardt
Final Results, 1/19
37.1%
32.6%
17.4%
11.2%
Entrance Poll, 1/19
35%
26%
20%
11%
Certain Voters, 1/18
26.8%
20.2%
23.6%
17.8%
All Voters, 1/18
26.4%
22.2%
22.4%
18.2%
All Voters, 1/10
17.4%
11.8%
27.2%
22.0%
With polls currently showing three different leaders in Iowa based on which factors one looks at--previous caucus goers, most certain caucus goers, caucus goers who have definitely made up their minds--I produced this chart to help determine which factors are the most important ones. Here are my thoughts:
Four years ago, Dean held an advantage among the most certain caucus goers. Over the final five polls before the caucus, he was tied with Edwards among all voters, but three points ahead among the most certain caucus goers. Given that Dean’s Entrance poll numbers, 20%, were closer to his numbers among all voters, 22%, than his numbers among the most certain voters, 24%, it does not seem to me that an advantage among the most certain caucus goers should be taken all that seriously. Gephardt’s 7% collapse alone cannot explain the 8% increase for Kerry from the final polls to the entrance poll, nor can it explain the 5% increase Edwards had from the final polls to the entrance poll. The fact is that the entrance poll in Iowa was closer to the final five-poll average among “all voters” than it was among “the most certain voters.” That isn’t very good news for Edwards since he consistently does better among the most certain caucuses than he does among all caucus goers.
Gephardt received 11% in the entrance poll, and 11.2% in the final result. This number seems quite important to me, since many people seem to be assuming that candidates at under 15% in Iowa polls, mainly Biden (currently at 5%) and Richardson (currently at 8%), will see most of their support dissipate in the actual caucus when they are unable to reach viability. There doesn’t seem to be any guarantee of that at all. A candidate can clearly come in at less than 15% statewide, and still hit their projected number even with Iowa’s wacky viability system. There is no reason to write off the support of candidates close to 10%, but below 15%, as certain to flow to other candidates. This somewhat reduces the importance of the second place vote, or at least of the Richardson second place vote, another area where Edwards has consistently led.
Clearly, the second place vote still had a big impact, if one looks at the differences from the entrance poll to the final results. Edwards went up 6-7%, Kerry went up 2%, and Dean went down 3%. Given these shifts, it seems pretty clear to me that leading among second-place choices means more than the supposed organizational advantage of “most certain voters.”
A lot of voters clearly made up their minds very late, given the huge pro-Edwards and Kerry surges over the final week. There was even a large gap between the final polls which, on average, concluded on January 16th, and the entrance poll, conducted on January 19th, three days later. This makes me think that the trendline is extremely important (good for Obama), and the results among those who have supposedly already made up their minds isn’t very important (bad for Clinton).
Although it is not reflected in this table, Pollster.com has been blogging about the possibility that entrance polls exaggerate the number of younger, first-time caucus goers. Such a skew, if accurate, would be bad for Obama, since Clinton and Edwards (especially Edwards) do much better among older voters than he does. This would make the “attended a previous caucus,” to be a key number to look out for in polling. Rasmussen shows a large, anti-Obama swing among such voters, as Obama slips to 22% (from 25%) and third place (from second place) among previous caucus goers. Edwards moved from third to first.
So, in short, here is how I would rank the importance of the various factors in Iowa polling:
Trend, which currently favors Obama and is not so great for Clinton. This matters a great deal, and could account for more than a 10% swing.
Second place choice, which currently favors Edwards and is not so great for Clinton. This could account for a 5% swing.
Previous caucus attendance, which currently favors Edwards and is not so great for Obama. This could account for a 2-3% swing.
The factors that don’t seem to matter much at all are caucus goers who have supposedly made up their minds, which currently favors Clinton, and organizational, “most certain to vote” strength, which currently favors Edwards. All told, it would be my guess that Edwards and Clinton are roughly tied in Iowa right now, with Obama a little bit out in front. I don’t know if yesterday’s events will change any of this, but I do know that polling Iowa is extremely difficult. This should all be considered guess work. The huge shifts we saw in Iowa back in 2004 from one week out to a couple days out, from a couple days out to the entrance poll, and from the entrance poll to the final results show just how difficult to track all of this actually is. Dean went from a 15% advantage over Edwards to a 15% deficit, and Gepahrdt went from a 5% advantage on Kerry to a 26% deficit. Since the January 3rd caucus date will even leave us without the benefit of much polling during the two weeks before the caucus, 2008 might be even harder to figure out than 2004.
