The Iowa Gold Standard
A very happy new year to you all!
I want to expand a bit on Turkana's post below (I was working on this post at the same time and just saw his post up). As you may be aware, the latest Des Moines Register (DMR) poll is out and it shows Sen. Obama at 32%, Sen. Clinton at 25%, and Sen. Edwards at 24%. The MoE is 3.5%, which, when you add the percentage of people still undecided (6%) or likely to change their minds (34%), still shows a 3-way race whose outcome will entirely be decided by the final turnout on caucus day and second choice preferences. Turkana has mentioned this, but I want to emphasize that the DMR poll is unusual in a few respects and that is partly reflected by the fact that when you look at the last several polls out of Iowa, it is the only one showing what, at face value, seems like a meaningful lead for Obama. The remaining polls from the past few days have nominal "leads" of 2%, 1%, 7% and 4% for Clinton, 1% for Edwards, 1% for Obama and 1% (tied) for Obama and Edwards. Yet, I see that the DMR poll is being labeled the "gold standard" because a number of campaign and media pollsters consider it the gold standard. Let's just say, while DMR may indeed be more reliable than others in Iowa, I am always skeptical of labeling any pollster the "gold standard" because no pollster has a perfect record and it gets particularly difficult to poll multi-candidate primaries, let alone multi-candidate caucuses. (For example, even in the simplest 2-person race of all, DMR's final general election poll had Kerry nominally "leading" Bush in Iowa (although only within the MoE) and the final result was the opposite, by a small margin.)
David Yepsen of the DMR has explained some of the assumptions behind the DMR poll results. I will discuss those below the fold - along with some historical polling data - to show why the assumpions behind the DMR poll are very unusual by historical standards and that simply calling the DMR the "gold standard" and reflexively trashing others like Reuters/Zogby (not Zogby Interactive) may be premature.
Let's start with Yepsen's explanation (emphasis mine):
* Undecideds exist. There are 6 percent of likely Democratic caucus-goers and 4 percent of the Republicans who have no first choice. Their final decisions will be enough to change the order of finish in both parties. That fact alone should keep anyone from using these polls to forecast the outcome of the race.
* Last-minute developments won't be reflected. In 2004, 21 percent of those who showed up at Democratic caucuses decided who they'd support in the last three days of the campaign. This poll won't reflect those decisions because it came out of the field on Sunday night -- four days before people vote.
* Some support is soft. Of those who have decided on a candidate, 34 percent of the Democrats say they could still be persuaded to change their minds. Among Republicans, it's 46 percent.
* A lot of caucus-goers are first-timers. A whopping 60 percent of the Democrats say this would be their first time at a caucus. Some 40 percent of the Republicans say that.
* A lot of Democratic caucus-goers aren't all that Democratic. Some 40 percent of the Democratic caucus-goers say they are independents, and another 5 percent say they are Republicans. (Technically, they'll all have to re-register as Democrats to participate, but that can be done at the caucus site.) Put another way, 54 percent of the Democratic caucus-goers say they're Democrats. In 2004, it was 80 percent.
That will raise some eyebrows among party pros. While we all know the Democratic turnout will be large, the events just haven't attracted that many newcomers in the past. If that happens as the poll suggests, the caucuses will no longer be affairs of the party activists and faithful, but more like primaries.
Will some of these fair-weather Democrats not bother to show up? If they don't, it will change the outcome. If pollsters adjust the party identifications in 2008 to look like they did in 2004, Clinton could beat Obama 31 percent to 29 percent.
Or, in the unlikely event just the registered Democrats who say they'll definitely show up are counted, Clinton gets 33 percent, Obama gets 27 percent and Edwards gets 25 percent.
* Democrats are on fire. And what if 40 percent of the Democratic caucus-goers are independents and 5 percent are Republicans? That tells us the GOP is in real trouble, that a lot of Americans want change and that they are turning to the Democrats to find it.
* Second choices don't really change that much. On the Democratic side, much is made out of the fact that to qualify for any delegates (which is what the state party releases as the results), a candidate must have at least 15 percent of the total people attending the event.
If the candidate can't muster that, then supporters may realign with a candidate who does. So, if you assume that Bill Richardson, Joe Biden and Chris Dodd can't reach that 15 percent threshold, the question is: Where do their people go? The three front-runners each get a few, but it doesn't change the order of finish: Obama would win with 28 percent; Clinton would still finish second with 20 percent and Edwards would capture third with 19 percent.
* Young voters. The poll shows 17 percent of likely Democratic caucus-goers are under age 35. In 2004, there were 17 percent who were under age 29. So, maybe too much has been made of the so-called "youth vote." Obama's people have always said they would be gravy on his victory and they are. He gets more from boomers but loses with the old fuds. They like Clinton and Edwards.
To me, the most unusual aspect of this poll is the assumption that 40% of Democratic caucus-goers are likely to be independents. Turkana has discussed this, and while this may turn out to be true, I will remain skeptical until I see that happen. In fact, I'm not the only one who's a bit skeptical about this. As Marc Ambinder points out:
The turnout model is weird, and even the Obama campaign doesn't project a turnout universe of 200,000, including 40% independents. (OK, Gordon Fischer, Obama's Iowa co-chair does, but..) ...
This is very important because Obama is lagging among Democrats but leading significantly among independents in this poll (it's 39 O - 15 C - 24 E among Independents and 27 O - 33 C - 25 E among Democrats). Also as Beltway Dem points out at MyDD: "The model, according the The Washington Post, assumes that more than half of Hillary's and Edwards's support will be from first time caucus goers and that 72% of Obama's will come from that group." (Beltway Dem also has the results of two other polls released after DMR both showing Clinton nominally "leading" but not outside the MoE.)
Let's set aside the turnout model assumptions and look at the results of different polls prior to the multi-candidate 2004 Democratic Iowa caucus, catalogued by Pollster.com's Mark Blumenthal. What is apparent from the data is that:
- The only poll that got the 1st, 2nd and 3rd place choices right is the DMR Likely Caucus Goer poll - but their numbers, unlike the final results, showed a tie when MoE was included.
- The DMR Definite Caucus Goer poll had Kerry in first place but actually missed Edwards in second place but only within the MoE.
- Interestingly, Reuters/Zogby and Insider Advantage both had Kerry in first place, but were like the DMR Definite Caucus Goer poll in missing Edwards in second place but only within the MoE. In other words, it is really difficult to make the claim that Zogby or Insider Advantage were significantly worse than DMR in 2004. (As a footnote, Reuters/Zogby and Insider Advantage currently both show Clinton nominally "leading" in their latest polls).
Further, as Mark Blumenthal at Pollster.com notes about Zogby:
Now to be fair, Zogby did show Edwards rising at the rate of a little less than one percentage point a day over the week leading up to their Sunday release. However, on the final Zogby release, on Monday (interviews conducted Friday to Sunday, 1/16-18), Edwards suddenly jumped 3-points (showing Kerry at 25%, Dean 22%, Edwards 21%, Gephardt 18%). Given the 3-day rolling average, the Edwards number on the final night of interviewing had to be in the 25-30% range. Such a result is plausible, given that Edwards received 26% on the network entrance poll, and Edwards was certainly surging in the campaign's final week. But draw your your own conclusions as to why the Register caught more of the Edwards surge earlier.
Bottomline, the turnout will decide who wins. It's still a 3-way tie from a pre-caucus polling perspective as far as I can tell.